Venture Autopilot DE Book a call

Example scenarios with model values

84 ventures. 84 honest answers.

This is what it can look like when a company brings us a venture. 36 times we build. 31 times a venture launches according to plan. 17 times we advise against it or show a narrower route. The market figures above each example are real and sourced.

Sector and niche make no difference. Law firm or bakery, workshop or warehouse, online shop or construction project: every venture goes through the same review.

Skilled trades: All 33 skilled trades examples

All examples are invented and show how we work. Companies, names, figures and results are model values and not a promise for your venture.

  1. 01Law firmBuilt
  2. 02Country innBuilt
  3. 03Construction companyBuilt
  4. 04New serviceAdvised against
  5. 05New business lineAdvised against
  6. 06New branchLaunched
  7. 07Own fashion brandAdvised against
  8. 08Pet food by subscriptionAdvised against
  9. 09Food onlineLaunched
  10. 10Furniture onlineAdvised against
  11. 11Selling through marketplacesLaunched
  12. 12B2B ordering portalBuilt
  13. 13Skilled trades businessBuilt
  14. 14AI in customer contactBuilt
  15. 15Introducing AIAdvised against
  16. 16Founding part timeLaunched
  17. 17Tax firmBuilt
  18. 18Medical practiceBuilt
  19. 19Care serviceBuilt
  20. 20Mechanical engineeringLaunched
  21. 21Natural cosmeticsLaunched
  22. 22Garden centreLaunched
  23. 23Second handLaunched
  24. 24City hotelLaunched
  25. 25Estate agencyBuilt
  26. 26HaulierBuilt
  27. 27ManufacturerLaunched
  28. 28Insurance brokerBuilt
  29. 29Staffing firmAdvised against
  30. 30BakeryLaunched
  31. 31Predictive maintenanceLaunched
  32. 32Quality inspectionLaunched
  33. 33Production planningBuilt
  34. 34Spare parts serviceBuilt
  35. 35Technical documentationBuilt
  36. 36PurchasingBuilt
  37. 37Shift handoverBuilt
  38. 38EnergyLaunched
  39. 39AutomationAdvised against
  40. 40In-house production softwareAdvised against
  41. 41Goods receiptBuilt
  42. 42Order pickingLaunched
  43. 43StocktakingLaunched
  44. 44Returns warehouseLaunched
  45. 45FulfilmentBuilt
  46. 46Site selectionLaunched
  47. 47New build or existing stockAdvised against
  48. 48Construction partnerBuilt
  49. 49Business premisesLaunched
  50. 50Office buildingAdvised against
  51. 51Micro apartmentsAdvised against
  52. 52Renovation roadmapLaunched
  53. 53Construction projectBuilt
  54. 54Heating engineer: heat pumpsBuilt
  55. 55Plumbing and heating: emergenciesBuilt
  56. 56Electrical firm: solar packageLaunched
  57. 57Electrical: contract inspectionsLaunched
  58. 58Roofer: survey by droneLaunched
  59. 59Painter: measuring from photosBuilt
  60. 60Tiler: 3D bathroom plannerAdvised against
  61. 61Carpentry firm: own softwareAdvised against
  62. 62Building: site diary, variationsBuilt
  63. 63Refrigeration: deadlines, reportsBuilt
  64. 64Joinery: enquiry with dimensionsLaunched
  65. 65Metalwork: job documentationBuilt
  66. 66Car workshop: approval by photoBuilt
  67. 67Body and paint: claim fileLaunched
  68. 68Glazier: own online shopAdvised against
  69. 69Stonemason: seeing the memorialLaunched
  70. 70Interior fitter: own 3D plannerAdvised against
  71. 71Contract cleaning: coverBuilt
  72. 72Chimney sweep: visits and routeBuilt
  73. 73Landscaping: care by contractLaunched
  74. 74Bakery: bake plan per branchLaunched
  75. 75Butcher: party cateringBuilt
  76. 76Hairdresser: bookings, waiting listBuilt
  77. 77Optician: lens subscriptionLaunched
  78. 78Dental laboratory: order intakeBuilt
  79. 79Orthopaedic workshop: estimatesBuilt
  80. 80Plumbing and heating: apprenticesLaunched
  81. 81Joinery: ready for handoverLaunched
  82. 82Painting contractor: e-invoicingBuilt
  83. 83Roofer: founding a portalAdvised against
  84. 84Pool building: enquiry in MarchBuilt

The inbox is sorted in advance. Client confidentiality shapes the architecture.

Assumed: German commercial law firm with 14 lawyers and around 600 incoming messages a day by email and beA.

The situation in figures

  • Companies in Germany that considered AI and do not use it: 62% cite legal uncertainty, 60% data protection21
  • 57% of companies with 20 or more employees use AI18

Built in the example

The request in the example

“We want AI in our inbox. But not one client word may end up somewhere we no longer control.”

Example with model values

Processing inside the firm's network

  1. Court via beAOrder with deadline for a statement
    Deadline matter98%
    Deadlines mailboxLawyer reviews
    File 0412/26
  2. ClientDocuments for the purchase agreement
    Ongoing matter96%
    Corporate law team
    File 0388/26
  3. Opposing partySettlement proposal
    Pleading93%
    Case handler
    File 0291/26
  4. New contactEmployment law enquiry
    New matter91%
    Client intake
    Conflict check
  5. Service providerOctober invoice
    Invoice99%
    Accounting
    Receipts
  6. UnknownPlease call back
    Unclear54%
    Front officeA human decides
    Not filed yet

Two routes to the model. The firm decides.

What the review shows

  • The office sorts mail by hand for around three hours every day.
  • Deadline matters wait four hours on average before a human sees them.
  • Every third attachment is filed twice or not at all.

What gets built

  • Every message is classified by matter, type and urgency.
  • It goes to the responsible mailbox and into the matter file of the document management system.
  • A suspected deadline always goes to a human. The firm still calculates, records and monitors deadlines. A human still reviews the full inbox every day.
  • Standard cases get a draft reply. Nothing is sent before approval.

Assumed result in the example

less sorting work per day in the office, assumed
Around 2 hrs
target until a suspected deadline reaches the responsible human
5 min
routing decisions are recorded with their reason
Log

Law and standards in view

  • § 203 StGBService providers are bound to secrecy as persons assisting the firm.
  • § 43e BRAOContract in text form with instruction on criminal liability. Access only as far as the service requires. For services from abroad, the protection of secrets there counts.
  • GDPR Art. 28, 32, 35Data processing agreement and technical safeguards. We supply the technical details for the firm's impact assessment before the start.
  • AI Act Art. 4, 6Measures to build staff AI literacy. The firm classifies the system, we supply the technical description. Nothing leaves the firm unless a human has checked it.
  • § 50 BRAOMatter files stay complete. Retention periods run in the document management system.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

We supply the technical details and the documentation. The legal and professional conduct assessment is made by the firm itself.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The kitchen runs. The front desk answers. The owner is home by eight.

Assumed: family run inn with 28 rooms, 90 seats and 19 employees.

The situation in figures

  • 245,555 legal units in hospitality in Germany, reporting year 202444
  • 108 insolvencies per 10,000 companies in hospitality in 2025, 69 on average across all sectors22
  • Bureaucracy costs an SME 32 hours a month on average7

Built in the example

The request in the example

“I want to know where my time goes. And I want my evenings back.”

Example with model values

The owner's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    1 hrs free
  • Sun
    2 hrs free

13 hours a week returned in the example

  • Kitchen

    Order suggestion from reservations, stock and weather. The head chef approves. Hygiene logs on a tablet instead of paper.

  • Front desk

    Enquiries from all channels in one place. Draft reply with availability and price within minutes. A human confirms.

  • Campaigns

    Annual plan with twelve occasions. Texts, images and mailing are prepared. Emails only to guests who gave prior express consent.

What the review shows

  • Orders are placed from memory: three suppliers, three slips of paper.
  • Room enquiries arrive through five channels and wait until after dinner service.
  • Promotions are written the night before the holiday.

Assumed result in the example

a week returned to the owner, assumed
13 hrs
target until a room enquiry is answered, six hours before
12 min
for all orders instead of three slips
1 list

Law and standards in view

  • Regulation (EC) No 852/2004The kitchen's own checks stay complete and are documented digitally.
  • KassenSichV (cash register ordinance)The till with its certified technical security system (TSE) stays untouched. Data is only read.
  • § 7 UWG, GDPREmail advertising only with prior express consent. Unsubscribe link in every message.
  • AI Act Art. 50A human approves every reply. If a guest later writes directly with an AI, the guest is told.
  • ArbZG (Working Hours Act)A human writes the duty roster. The system checks it against rest periods and maximum hours before it goes out. Where a works council exists, it has a say.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

First the review, then the technology. We only automate what the review proves to be a time sink.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The site reports on its own. The site manager builds.

Assumed: renovation and fit-out contractor with 35 employees and 40 active sites.

The situation in figures

  • 104 insolvencies per 10,000 companies in construction in Germany in 2025, 69 on average across all sectors22
  • Bureaucracy costs an SME 32 hours a month on average7

Built in the example

The request in the example

“My site managers are on the phone half the day. Customers ask about dates, material is missing, and nobody said so in time.”

Example with model values

  1. Quote
  2. Order
  3. Material
  4. Execution
  5. Acceptance
  1. Mon 07:10Order

    Tiles for the upstairs bathroom ordered after approval. Delivery on Thursday confirmed.

  2. Mon 07:12AI assistant

    Good morning Ms Keller, the tiles arrive on Thursday. Your bathroom will be tiled from Monday as planned.

  3. Tue 16:40Customer

    Can we still move the sockets in the kitchen?

  4. Tue 16:41AI assistant

    That is a change to the order. I am passing it to Mr Brandt. You will receive a quote with a price today.

  5. Tue 17:05Approval

    Site manager approves the variation quote: two sockets, €214.20 including VAT. The customer confirms in text form.

  6. Wed 08:00Reminder

    Door selection open for six days. Reminder sent to the customer.

  7. Fri 14:00Site diary

    Weekly report with photos of the work to the customer and into the project file.

What the review shows

  • A site manager makes around 40 calls a day. Half of them are status questions.
  • Material is ordered when it is missing. Two days of standstill per site on average.
  • Promises sit in chat threads and cannot be evidenced later.

What gets built

  • Orders are prepared from the construction schedule and placed after approval, with supplier confirmation.
  • Customers receive status, dates and reminders without asking.
  • An AI assistant answers questions. Anything touching money or contract goes to the site manager.
  • Every message lands in the site diary of the project.

Assumed result in the example

fewer per site manager per day, assumed
20 calls
standstill from missing material, none in the model month
Less
promises are in the project file with a date
Evidenced

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • §§ 631, 650b BGB (German Civil Code)Changes are always offered by a human. A variation only takes effect once the customer agrees in text form. The AI prepares.
  • § 7 UWGMessages concern the running order. Advertising only with consent.
  • GDPR Art. 6, 28Customer data only for the order. Photos show the work and no people. Providers are bound by contract.
  • Site diaryEvery promise is in the project file with date and sender.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Scale
  5. Built in the example

The AI writes, reminds and prepares orders. Decisions about money and contract stay with a human.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€380,000 is budgeted. The review says no and shows another way.

Assumed: fitness chain with nine studios and 21,000 members.

The situation in figures

  • 57% of companies in Germany with 20 or more employees use AI18
  • Companies that considered AI and do not use it: 62% cite legal uncertainty21

Advised against in the example, with counterproposal

The request in the example

“We want an AI health check as an add-on subscription: upload blood values, get analysis and advice in the app.”

Example with model values

  • Riskhigh

    Legal hurdles

    • Software that assesses individual blood values and gives advice is as a rule a medical device or an in vitro diagnostic (Regulations (EU) 2017/745 and 2017/746). Under today's classification a notified body is usually involved.
    • Health data under Art. 9 GDPR. AI in medical devices can count as high risk (AI Act Art. 6).
    • Advice on individual findings touches the German rule that only licensed practitioners may diagnose and treat (§ 1 HeilprG).
  • Riskhigh

    Finance

    • Certification and quality management: 12 to 18 months before the first euro of revenue.
    • In the model none of the three scenarios breaks even within 36 months.
  • Riskhigh

    Competition

    • Many health insurers offer health apps at no extra charge.
    • Providers with CE marked products have been on the market for years.

Do not start

The counterproposal

Same goal: more revenue per member and longer retention.

  • Training companion without diagnosis: goals, training data, class recommendation. Heart rate, weight and other body data only with explicit consent.
  • Health checks are offered by a medical partner under its own responsibility. The studio only points members to it and sees no findings. A law firm checks the payment model first.
  • Test in two studios over six weeks, with a clear stop rule.
  • Planned service, company budget
    €380,000
  • Test of the counterproposal, company budget
    €28,000

Assumed result in the example

stays in the budget in the example until the risk is clarified
€352,000
to the decision paper (planning value)
5 working days
duration of the test in two studios
6 weeks
  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

We name legal hurdles as a risk. The legal review is carried out by a law firm of your choice.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Their own logistics platform sounds like the future. The numbers in the model say: do not build it.

Assumed: building materials dealer with 60 vehicles and four sites.

The situation in figures

  • 51% of logistics companies in Germany use AI, 54% lack the expertise20
  • 133 insolvencies per 10,000 companies in transport and storage in 202522

Advised against in the example

The request in the example

“Our route planning costs us money every day. We will build our own AI platform and then sell it to other dealers.”

Example with model values

Cumulative cash balance of the new line € thousand

Advised against
  • Zero line: the money invested is back
  • Favourable −120
  • Expected −640
  • Unfavourable −1,100

Why we advise against it

  • Market: mature route planning is available at monthly prices below their own maintenance cost.
  • Edge: data from 60 vehicles gives no advantage over vendors with thousands of fleets.
  • Sales: selling software to competitors is a second company with its own team.
  • Focus: management would be missing from the core business for two years.

What happens instead

  • Standard route planning software, pilot with twelve vehicles in four weeks. Drivers and the works council are involved before the start.
  • The real problem sits in order intake. That is where the improvement starts.
  • The budget stays in the core business.

Assumed result in the example

scenarios reach the zero line within 36 months in the model
0 of 3
shortfall in the middle scenario of the model after 36 months
€640,000
pilot with standard software and twelve vehicles
4 weeks
  1. Pre-flight check
  2. Validate
  3. Advised against in the example

A no from Validate costs €4,900 net, planning value five working days. A no after two years costs the business line.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

New region, new services, 90 days. Every week shows whether the course holds.

Assumed: heating and plumbing company with 45 employees and one site so far.

The situation in figures

  • Around 564,000 skilled trades companies in Germany, reporting year 202423
  • 72% say they have too much to do to deal with digitalisation19

Launched in the example

The request in the example

“We are opening a branch 80 kilometres away, with heat pump maintenance contracts and an energy check as new services. How do we start right?”

Example with model values

  1. Day 1 to 30Build

    Team, page, listings, first campaign

  2. Day 31 to 60Demand

    Two channels compared, quotes within 24 hours

  3. Day 61 to 90Reinforce

    Budget to the better channel, maintenance contracts

Enquiries per week: plan and assumed run

Enquiries per week
34Plan 28
Quote rate
66%Plan 60%
Jobs
46Plan 40
Maintenance contracts
31Plan 25
Cost per enquiry
€38Plan €45
  1. Day 30Channel B stays below plan in the example. The rule halves the budget.
  2. Day 60Channel A carries in the example. The rule doubles the budget and a second fitter joins.
  3. Day 90In the example the run is above plan. The next cycle is being planned.

What the review shows

  • Analysis: demand, competitors and prices in the region.
  • Forecast: three scenarios with unit costs per job and per maintenance contract.

What gets built

  • Product: three service packages with fixed total prices, own page, booking flow.
  • Plan: 90 days in three stages. Five metrics per week and a rule for what happens on deviation.

Assumed result in the example

in the model run after 90 days, plan 40
46 jobs
for maintenance in the model run, plan 25
31 contracts
to the course in the example, both before day 61
2 corrections
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Scale
  5. Launched in the example

The plan is not a promise. It is a measuring instrument: whoever measures every week can reinforce or correct early.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€120,000 is budgeted for a fashion brand. The review says: not like this, but narrower.

Assumed: textile retailer with two shops and 14 employees.

The situation in figures

  • €15.1 billion online clothing sales in Germany in 2025, up 3.5%1
  • 44.2% of fashion sales already happen online3
  • 56.7% of online sales ran through marketplaces in 20253

Advised against in the example, with counterproposal

The request in the example

“We want to build our own fashion brand online. The market is huge, after all.”

Example with model values

  • Riskhigh

    Competition

    • The biggest online market is also the most crowded.
    • Whoever sells through marketplaces pays fees and stands in direct price comparison.
  • Riskhigh

    Returns

    • Fashion is sent back more often than most other goods.
    • In the model a return rate of 35% eats up the margin of the base collection.
  • Riskraised

    Capital

    • Stock in six sizes and four colours ties up the budget before a single item is sold.

Do not start like this

The counterproposal

Same goal: an own brand with its own customers.

  • A niche instead of a range: workwear for women in the skilled trades.
  • Pre-order instead of stock: only what is ordered gets produced.
  • Test over six weeks with 200 pre-orders as the threshold.
  • Planned brand, company budget
    €120,000
  • Test of the niche, company budget
    €18,000

Assumed result in the example

stays in the budget in the example until the niche is proven
€102,000
pre-orders as the threshold for the start, assumed
200
to the decision paper (planning value)
5 working days
  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

A big market is no reason to start. A proven edge in a niche is one.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Pet food subscriptions are growing. For this retailer only one of three variants carries in the model.

Assumed: pet shop with three branches and 22 employees.

The situation in figures

  • €1.97 billion online pet supplies sales in Germany in 2025, up 5.0%1
  • Up 9.8% year on year in the second quarter of 20262
  • 43% of households have at least one pet5

Advised against in the example, with counterproposal

The request in the example

“Everyone talks about food by subscription. Is it still worth it for us?”

Example with model values

Cumulative cash balance of the subscription € thousand

Only with a niche
  • Zero line: the money invested is back
  • Special food for existing customers 30
  • Broad range −310
  • Price war −590

What the review shows

  • In the broad range price decides. Large retailers have the advantage there.
  • Shipping heavy bags costs 14% of the basket in the model.
  • In the example the branches know 4,000 owners whose pets need special food.

What gets built

  • Subscription only for special food after advice, with pick-up in the branch or delivery.
  • Start with the existing customers of the three branches.
  • Stop rule: below 300 subscriptions after four months the trial ends.

Assumed result in the example

variants reaches the zero line after 24 months in the model
1 of 3
as the threshold after four months, assumed
300 subscriptions
existing customers with a matching need, assumed
4,000
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Advised against in the example, with counterproposal

Growth in the market does not mean room for everyone. The review looks for the variant in which the retailer has something others lack.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Regional food online: one delivery area, 90 days, five metrics.

Assumed: association of six farm shops and producers in one region.

The situation in figures

  • €4.1 billion online food sales in Germany in 2025, up 5.5%1
  • Up 8.9% year on year in the second quarter of 20262
  • Only 3.2% of food retail happens online3

Launched in the example

The request in the example

“Many customers ask for delivery. Should we build a joint online shop?”

Example with model values

  1. Day 1 to 30One delivery area

    Two postcodes, two delivery days, six producers

  2. Day 31 to 60Rhythm

    Fixed boxes, reorder with one click

  3. Day 61 to 90Expand or stop

    Second area only if delivery cost per order is below plan

Orders per week: plan and assumed run

Orders per week
152Plan 140
Basket
€48Plan €45
Delivery cost per order
€6.90Plan €7.50
Repeat customers
41%Plan 35%
Waste
5%Plan 6%
  1. Day 30Demand is just below plan in the example. A third delivery day is dropped.
  2. Day 60Fixed boxes carry in the example. Single items are reduced.
  3. Day 90Delivery cost is below plan in the example. The second area is being planned.

What the review shows

  • The last mile decides: in the model delivery only pays from a €45 basket.
  • A shop for the whole region would be too expensive to deliver.

What gets built

  • Joint ordering page with fixed boxes and two delivery days.
  • Start in two postcodes with a rule for expanding or stopping.

Assumed result in the example

per week in the model run after 90 days, plan 140
152 orders
delivery cost per order in the model run, plan €7.50
€6.90
at the start instead of the whole region
1 area
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Lots of headroom in the market and expensive delivery: that is why the plan starts small and measures every week.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

A nationwide furniture shop for €300,000: none of the scenarios carries in the model.

Assumed: furniture store with its own joinery, one showroom and 30 employees.

The situation in figures

  • €5.6 billion online sales of furniture, lamps and decoration in Germany in 2025, up 3.3%1
  • Down 2.1% year on year in the second quarter of 20262

Advised against in the example

The request in the example

“We want to sell online nationwide. The showroom alone is no longer enough.”

Example with model values

Cumulative cash balance of the furniture shop € thousand

Advised against
  • Zero line: the money invested is back
  • Favourable −490
  • Expected −930
  • Unfavourable −1,350

Why we advise against it

  • Market: the product group is receding in 2026 while others grow.
  • Cost: freight shipping and taking back large furniture cost 18% of sales in the model.
  • Edge: the strength of the business is made to measure work with advice. Nationwide it counts for little.

What happens instead

  • Configurator for made to measure furniture on its own page, with advice within 100 kilometres.
  • Visibility in the region instead of nationwide shipping.
  • In the example the budget for this is below a tenth of the planned shop.

Assumed result in the example

scenarios reach the zero line within 36 months in the model
0 of 3
shortfall in the middle scenario of the model after 36 months
€930,000
radius for the counterproposal
100 km
  1. Pre-flight check
  2. Validate
  3. Advised against in the example

The wish is right, the route is not. The review separates the two and costs €4,900 net, planning value five working days.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The specialist shop goes where people buy. And keeps its customers.

Assumed: specialist bicycle retailer with workshop, two sites and 6,000 items.

The situation in figures

  • 56.7% of online sales in Germany ran through marketplaces in 20253
  • €92.3 billion net online retail sales in 2025, up 3.9%3

Launched in the example

The request in the example

“Our own shop is barely growing. Should we go onto the marketplaces?”

Example with model values

  1. Day 1 to 30Selection

    400 items with margin after fees, clean product data

  2. Day 31 to 60Learn

    Prices, returns and reviews per item

  3. Day 61 to 90Own customers

    Inserts and service lead buyers to the own shop

Orders per week: plan and assumed run

Orders per week
78Plan 70
Margin after fees
19%Plan 18%
Returns
7%Plan 8%
Purchases in the own shop
24%Plan 20%
Listed items
400Plan 400
  1. Day 3060 items stay below the minimum margin in the example. They are removed.
  2. Day 60Accessories carry better than complete bikes in the example. The range follows.
  3. Day 90In the example the run is above plan. A second marketplace is being reviewed.

What the review shows

  • Of 6,000 items only 400 carry after fees and shipping in the model.
  • Whoever sells only through marketplaces does not know the buyers.

What gets built

  • Selection and listing of the 400 items with checked product data.
  • Weekly report per item with one rule: whatever falls below the minimum margin goes.
  • Service and workshop as the reason to order direct next time.

Assumed result in the example

per week in the model run after 90 days, plan 70
78 orders
margin after fees in the model run, plan 18%
19%
of purchases in the own shop, assumed
24%
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The marketplace is a channel and not a business model. What is measured is margin after fees, not sales.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The order arrives at 6:40 am. Inside sales takes care of the exceptions.

Assumed: building services wholesaler with 1,800 business customers and 55 employees.

The situation in figures

  • Manufacturers and wholesalers in Germany sold €549 billion through online shops and marketplaces in 2025, up 7.9%4
  • 56.7% of online retail sales ran through marketplaces in 20253

Built in the example

The request in the example

“Seven in ten orders arrive by phone or email. Our inside sales team types all day.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot customers
  4. Rollout
  5. Operation
  1. Mon 06:40Customer

    Reorder as last week, plus 20 boxes of item 4471.

  2. Mon 06:40AI assistant

    Order recorded. Item 4471 is only available from Wednesday. Shall I split it or deliver everything on Wednesday?

  3. Mon 06:42Customer

    Split it.

  4. Mon 06:42Order

    Two deliveries created in the ERP system, prices under the customer's framework agreement.

  5. Mon 08:15Approval

    Inside sales approves the volume discount for the 20 boxes.

  6. Fri 12:00Report

    Weekly overview: orders in the portal, open queries, delivery rate.

What the review shows

  • In the example inside sales records orders by hand for five hours a day.
  • Eight in ten orders are repeats with small changes.
  • Every customer has its own prices. That is why standard shops fail.

What gets built

  • Ordering portal with customer prices and availability from the ERP system.
  • An AI assistant takes orders in free text and asks when something is missing.
  • Discounts and special cases are approved by inside sales.
  • Start with 40 pilot customers, then rollout in groups.

Assumed result in the example

of pilot customers' orders through the portal, assumed
60%
less recording per day in inside sales, assumed
3 hrs
pilot customers in the first stage
40

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Prices and discountsWhatever deviates from the framework agreement is approved by a human.
  • GDPR Art. 28Providers are bound by contract. Customer data stays within the order process.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Scale
  5. Built in the example

Existing customers, recurring orders, measurable effort: this is where building pays off most readily.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Quotes, appointments, invoices: the master has his evenings back.

Assumed: electrical contractor with twelve employees and 900 jobs a year.

The situation in figures

  • Around 564,000 skilled trades companies in Germany, reporting year 202423
  • 72% say they have too much to do to deal with digitalisation. 4% use AI19
  • Bureaucracy costs an SME 32 hours a month on average7

Built in the example

The request in the example

“During the day I am on site. I write quotes and invoices in the evening and on Saturday.”

Example with model values

The master's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    2 hrs free
  • Sun
    2 hrs free

14 hours a week returned in the example

  • Quotes

    Draft from measurements, photos and price list. The master checks and approves.

  • Appointments

    Customers book free slots themselves. Reminder the day before, a cancellation fills the gap.

  • Invoices

    Draft from timesheet and material on the day of acceptance. Sent after checking.

What the review shows

  • A quote takes four days in the example. By then every third customer has agreed elsewhere.
  • Invoices go out 19 days after acceptance on average.
  • Arranging appointments costs an hour on the phone every day.

Assumed result in the example

a week returned to the master, assumed
14 hrs
target until the quote, four before
1 day
target until the invoice, 19 before
3 days

Law and standards in view

  • AI Act Art. 50Where a customer writes with an AI, the customer is told.
  • Quote and invoicePrices and invoices are checked and sent by a human.
  • GDPR Art. 6, 28Customer data only for the job. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The three processes that cost the business the most time are automated. Nothing more at first.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The enquiry arrives at 7:12 pm. So does the answer.

Assumed: property management with 2,400 flats and nine employees.

The situation in figures

  • 57% of companies in Germany with 20 or more employees use AI18
  • 72% name customer contact as a field of use18

Built in the example

The request in the example

“The phone rings non-stop, and it is always about the same twenty questions.”

Example with model values

  1. Analysis
  2. Knowledge
  3. Pilot
  4. Operation
  1. Tue 19:12Customer

    The heating in the stairwell is cold.

  2. Tue 19:12AI assistant

    I am the property management's AI assistant. I have logged a fault for your building. The emergency service will contact you by 9 pm today.

  3. Tue 19:13Job

    Fault passed to the heating contractor, building data and access attached.

  4. Tue 19:15Customer

    Can I reduce the rent because of this?

  5. Tue 19:15AI assistant

    That is a legal question. I am passing it to your property manager. She will contact you tomorrow.

  6. Wed 09:30Reply

    The property manager replies in person.

What the review shows

  • Six in ten enquiries are standard cases in the example: fault, appointment, document.
  • Every third enquiry arrives after office hours.
  • Legal questions and money questions need a human.

What gets built

  • An AI assistant answers standard cases from the management's own documents.
  • Faults go straight to the tradesperson with the building data.
  • Law, money and complaints go to the responsible person.

Assumed result in the example

of enquiries are standard cases in the model
60%
target until the first answer, evenings included
2 min
answers questions on law, money and complaints
Human

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Legal questionsThe assistant gives no legal information. It passes the question on.
  • GDPR Art. 6, 28Tenant data only for the matter at hand. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI answers what is in the documents. Where law or money is concerned, a human takes over.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€250,000 is budgeted for an AI platform. The review says: clarify first, then one pilot.

Assumed: mechanical engineering company with 180 employees.

The situation in figures

  • Companies in Germany that considered AI and do not use it: 72% lack knowledge, 62% cite legal uncertainty, 60% data protection21
  • 57% of companies with 20 or more employees use AI18

Advised against in the example, with counterproposal

The request in the example

“We have to do something with AI now. But we lack the people, and nobody tells us what we are allowed to do.”

Example with model values

  • Riskraised

    Knowledge

    • In the example nobody in the company has run an AI project.
    • A platform for everything presupposes knowledge that is yet to emerge.
  • Riskopen

    Law

    • It is open which data may go into which model.
    • Works council and data protection are not yet involved.
  • Riskopen

    Data

    • Quotes, drawings and calculations sit in four systems.

Clarify first

The roadmap

Same goal: AI where it measurably saves time.

  • Three fields of use with benefit and risk instead of a platform for everything.
  • Open legal questions as a list for the company's law firm.
  • A pilot in quote preparation over eight weeks.
  • Planned AI platform, company budget
    €250,000
  • Clarification and pilot, company budget
    €32,000

Assumed result in the example

stays in the budget in the example until the pilot shows results
€218,000
fields of use with benefit and risk instead of one platform
3
duration of the pilot in quote preparation
8 weeks
  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

We name legal questions as open points. The legal review is carried out by a law firm of your choice.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Ten hours a week, 90 days: prove demand first, then build.

Assumed: physiotherapist with her own practice who is founding a digital exercise programme alongside her work.

The situation in figures

  • Around 690,000 founders in Germany in 2025, 70% part time, 44% with a digital offering6
  • 57% name winning customers as a hurdle6

Launched in the example

The request in the example

“I have the idea and ten hours a week. Should I have the app built now?”

Example with model values

  1. Day 1 to 30Test

    One page, one promise, two target groups

  2. Day 31 to 60Correct

    The better target group, twenty conversations

  3. Day 61 to 90Decide

    Pre-orders as the threshold for building

Waiting list sign-ups per week: plan and assumed run

Sign-ups per week
86Plan 80
Cost per sign-up
€5Plan €6
Pre-orders
52Plan 40
Conversations with practices
24Plan 20
Own hours per week
10Plan 10
  1. Day 30The first target group stays far below plan in the example. Switch to practices instead of patients.
  2. Day 60Practices carry in the example. The offering becomes a tool for the practice.
  3. Day 90The threshold of 40 pre-orders is exceeded in the example. The build is being planned.

What the review shows

  • Building the app would tie up the budget before a single customer has paid.
  • It is unclear who pays: patients or practices.

What gets built

  • A page with waiting list and pre-order instead of an app.
  • Two target groups compared, with a rule for switching.
  • Building only starts from 40 pre-orders.

Assumed result in the example

pre-orders in the model run, threshold 40
52
of target group in the example, on day 30
1 switch
own time per week, as planned
10 hrs
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The product is rarely the problem. The first paying customer is. That is why the plan measures demand first.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The receipts sort themselves. Advice stays with the tax adviser.

Assumed: German tax firm with two advisers, eleven employees and 420 clients.

The situation in figures

  • Bureaucracy costs an SME in Germany 32 hours a month on average7
  • 20% of SMEs use AI8

Built in the example

The request in the example

“At month end we drown in receipts and queries. There is no time left for advice.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot
  4. Operation
  1. Mon 08:05Customer

    Attached are the receipts for October. Is anything missing?

  2. Mon 08:05AI assistant

    I am the firm's AI assistant. 46 receipts have arrived. The statements of the second bank and two incoming invoices are missing.

  3. Mon 08:06Pre-entry

    Receipts read and assigned to the client file. Booking suggestions are ready for review.

  4. Mon 10:30Review

    The clerk reviews the suggestions and books them.

  5. Mon 11:15Customer

    Can I deduct the new company car in full?

  6. Mon 11:15AI assistant

    That is a tax question. I am passing it to your tax adviser.

What the review shows

  • In the example 40% of working time goes into entry and queries about missing receipts.
  • Every third client delivers incomplete documents.

What gets built

  • Receipts are read, assigned and checked for completeness.
  • Clients learn immediately what is missing.
  • Booking suggestions are reviewed and booked by a human. Tax questions are answered by the adviser.

Assumed result in the example

of entry time saved in the model
40%
target until the receipt list is complete, nine before
1 day
reviews every booking and answers every tax question
Human

Law and standards in view

  • § 57, § 62a StBerGConfidentiality also applies to service providers. They are bound by contract.
  • GDPR Art. 28, 32Data processing agreement and technical safeguards.
  • AI Act Art. 50The assistant identifies itself as AI in every conversation.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The professional conduct assessment is made by the firm itself. We supply the technical details for it.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The phone goes quiet. The appointments still fill up.

Assumed: group practice with four doctors and nine employees.

The situation in figures

  • 271,527 legal units in health and social work in Germany, reporting year 202444
  • Bureaucracy costs an SME 32 hours a month on average7

Built in the example

The request in the example

“Two colleagues are on the phone all morning. And every tenth appointment is missed without cancellation.”

Example with model values

The week at the front desk

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    0 hrs free
  • Sun
    0 hrs free

10 hours a week returned at the front desk in the example

  • Appointments

    Patients book and move appointments online. Free gaps are filled from the waiting list.

  • Reminders

    Reminder two days before. Cancellations arrive in time.

  • Phone

    An AI assistant takes appointment requests. Medical questions go to the practice team.

What the review shows

  • In the example 60% of calls are about appointments.
  • Every tenth appointment stays unfilled.
  • Prescription requests and questions about findings need the practice team.

Assumed result in the example

a week returned at the front desk, assumed
10 hrs
target for missed appointments, 10% before
4%
answers every medical question
Human

Law and standards in view

  • GDPR Art. 9, 28Health data only to the extent the appointment needs. Providers are bound by contract.
  • § 203 StGBService providers are bound to secrecy.
  • Medical questionsThe assistant gives no medical information. It passes the question on.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

Organisation is automated. Everything medical stays with the practice team.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The round is set at six. The manager is with the team by half past six.

Assumed: home care service with 38 carers and 210 clients.

The situation in figures

  • 271,527 legal units in health and social work in Germany, reporting year 202444
  • Companies that considered AI and do not use it: 60% cite data protection as the reason21

Built in the example

The request in the example

“Every morning I spend two hours replanning rounds. And in the evening everyone is still writing documentation.”

Example with model values

The care manager's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    2 hrs free
  • Sun
    2 hrs free

14 hours a week returned in the example

  • Rounds

    Suggestion for daily rounds from visits, qualification and travel times. The manager approves.

  • Absences

    When someone calls in sick, a replacement plan is ready for approval within minutes.

  • Documentation

    Carers dictate. The draft is checked and only then saved.

What the review shows

  • In the example planning rounds costs the manager two hours a day.
  • Documentation is written after the shift, 35 minutes per carer on average.

Assumed result in the example

a week returned to the manager, assumed
14 hrs
target for documentation per carer per day, 35 before
20 min
approves every plan and every entry
Human

Law and standards in view

  • GDPR Art. 9, 28, 35Health data, data processing agreement and impact assessment before the start.
  • Care documentationEvery entry is checked and confirmed by the carer.
  • Working hours, co-determinationPlans are checked against rest periods. Where a works council exists, it has a say.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The time goes back to the people who care. The decision stays with them.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The quote takes two days instead of nine. It is checked as before.

Assumed: manufacturer of conveyor systems with 140 employees and 500 quotes a year.

The situation in figures

  • 57% of companies in Germany with 20 or more employees use AI18
  • Companies that considered AI and do not use it: 72% lack the knowledge21

Launched in the example

The request in the example

“Our best engineers write quotes instead of building machines. And the customer waits nine days.”

Example with model values

  1. Day 1 to 30One product line

    Bills of material, prices and 200 old quotes as the basis

  2. Day 31 to 60Pilot

    Two sales engineers work with drafts

  3. Day 61 to 90Expansion

    Second product line if the error rate is below plan

Quotes per week: plan and assumed run

Quotes per week
21Plan 18
Days to the quote
2Plan 3
Corrections per quote
1Plan 2
Errors after sending
0Plan 0
Product lines
1Plan 1
  1. Day 30Custom builds stay out in the example. The draft covers standard machines.
  2. Day 60The engineers correct less than planned in the example.
  3. Day 90In the example the run is above plan. The second product line is being planned.

What the review shows

  • In the example a quote takes nine days. Every fourth enquiry is lost in that time.
  • 80% of a quote consists of building blocks from earlier quotes.

What gets built

  • Draft quote from the enquiry, bill of material and price list.
  • The engineer checks, corrects and approves.
  • Start with one product line and a rule for expansion.

Assumed result in the example

to the quote in the model run, nine before
2 days
per week in the model run, plan 18
21 quotes
approves every quote
Human
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Scale
  5. Launched in the example

The AI writes the draft. Responsibility for price and engineering stays with the engineer.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Three products, one target group, 90 days. Repurchase decides.

Assumed: natural cosmetics maker with eight employees, so far sold only through specialist shops.

The situation in figures

  • €3.5 billion online drugstore sales in Germany in 2025, up 4.6%1
  • 56.7% of online sales ran through marketplaces in 20253

Launched in the example

The request in the example

“We sell through retailers and do not know our customers. We want to sell direct.”

Example with model values

  1. Day 1 to 30Three products

    One target group, one promise, own page

  2. Day 31 to 60Repurchase

    Refill by subscription, samples in every parcel

  3. Day 61 to 90Channel

    Budget to the channel with the cheapest first purchases

Orders per week: plan and assumed run

Orders per week
164Plan 150
Basket
€41Plan €38
Repeat buyers
31%Plan 25%
Cost per first purchase
€12Plan €14
Returns
2%Plan 3%
  1. Day 30One product stays below plan in the example. It leaves the range.
  2. Day 60The subscription carries in the example. Refill becomes the default.
  3. Day 90In the example the run is above plan. The fourth product is being prepared.

What the review shows

  • In the model the business only pays from the second purchase.
  • The full range of 40 products would be too expensive to start with.

What gets built

  • Own page with three products and refill by subscription.
  • Weekly report with five metrics and one rule per product.

Assumed result in the example

per week in the model run after 90 days, plan 150
164 orders
repeat buyers in the model run, plan 25%
31%
at the start instead of 40
3 products
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Start small, measure repurchase, then expand. Specialist shops remain partners.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Ordered online, picked up in the afternoon. The shop stays the centre.

Assumed: garden centre with a DIY range, 45 employees and one site.

The situation in figures

  • €3.0 billion online sales of DIY goods and flowers in Germany in 2025, up 4.9%1
  • €83.1 billion online retail of goods in 2025, gross1

Launched in the example

The request in the example

“People look online and then buy elsewhere. Do we need an online shop?”

Example with model values

  1. Day 1 to 30Stock

    1,200 items with reliable stock online

  2. Day 31 to 60Pick-up

    Ordered by noon, ready from four

  3. Day 61 to 90Season

    Planting season as the first campaign to existing customers

Pick-up orders per week: plan and assumed run

Pick-ups per week
88Plan 75
Basket
€61Plan €55
Add-on at pick-up
27%Plan 20%
Stock correct
98%Plan 97%
Minutes per order
7Plan 8
  1. Day 30Plants stay in the shop for now in the example. Stock is too uncertain online.
  2. Day 60Pick-up carries in the example. A fixed pick-up point is created.
  3. Day 90In the example the run is above plan. Delivery in the area is being reviewed.

What the review shows

  • Shipping heavy and bulky goods does not pay in the model.
  • Seven in ten customers live within 20 kilometres.

What gets built

  • Ordering with reliable stock and same day pick-up.
  • No shipping in the first stage.

Assumed result in the example

per week in the model run after 90 days, plan 75
88 pick-ups
buy something extra at pick-up, assumed
27%
shipping cost in the first stage
€0
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Not every retailer needs shipping. Here the online channel brings customers into the shop.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Used children's clothing online: listing decides the margin.

Assumed: children's fashion retailer with three shops and 16 employees.

The situation in figures

  • The online second hand market in Germany grows to €10.5 billion in 2025, up 5.3%3
  • €15.1 billion online clothing sales in 2025, up 3.5%1

Launched in the example

The request in the example

“Our customers ask whether we take back worn items. Is that a business or just work?”

Example with model values

  1. Day 1 to 30Buying in

    Buying in through the own shops, checks against fixed criteria

  2. Day 31 to 60Listing

    Photo, size and condition in two minutes per item

  3. Day 61 to 90Loop

    Voucher for items bought in brings customers back

Items sold per week: plan and assumed run

Items sold per week
240Plan 220
Minutes per item
2Plan 3
Margin per item
€7Plan €6
Returns
10%Plan 12%
Vouchers redeemed
58%Plan 50%
  1. Day 30Small items under five euros do not pay in the example. They go into bundles.
  2. Day 60Listing takes two minutes per item in the example.
  3. Day 90In the example the run is above plan. The third shop starts buying in.

What the review shows

  • In the model an item only pays if listing stays under three minutes.
  • Buying in through the own shops costs no shipping.

What gets built

  • Listing with photo and AI suggestion for title, size and condition. A human confirms.
  • Voucher instead of payment for items bought in.

Assumed result in the example

per week in the model run after 90 days, plan 220
240 items
per item for listing, threshold 3
2 min
of vouchers are redeemed in the shop in the model
58%
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The market is growing, the margin is thin. That is why the plan first measures minutes per item.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

More guests book direct. The portals stay, but not for everyone.

Assumed: owner managed city hotel with 62 rooms and 24 employees.

The situation in figures

  • 245,555 legal units in hospitality in Germany, reporting year 202444
  • 108 insolvencies per 10,000 companies in hospitality in 2025, 69 on average across all sectors22

Launched in the example

The request in the example

“Eight in ten bookings come through portals. The commission eats our margin.”

Example with model values

  1. Day 1 to 30Own booking

    Booking flow on the own page, same prices

  2. Day 31 to 60Regular guests

    Invitation to former guests who consented

  3. Day 61 to 90Enquiries

    Draft reply for groups and celebrations within minutes

Direct bookings per week: plan and assumed run

Direct bookings per week
38Plan 32
Share direct
41%Plan 35%
Reply to enquiries
15 minPlan 30 min
Regular guests reached
1,050Plan 900
Booking drop-off
52%Plan 60%
  1. Day 30Bookings drop off at payment in the example. Payment by invoice is added.
  2. Day 60Regular guests carry in the example. The invitation becomes a quarterly letter.
  3. Day 90In the example the run is above plan. Group enquiries are being expanded.

What the review shows

  • In the example every third guest returns and still books through a portal.
  • Group enquiries wait a day for a reply on average.

What gets built

  • Booking flow on the own page with the same prices.
  • Invitation to former guests, only with consent.
  • Draft replies for groups, a human confirms.

Assumed result in the example

direct bookings in the model run, 20% before
41%
per week direct in the model run, plan 32
38 bookings
until the reply to a group enquiry in the model run
15 min

Law and standards in view

  • § 7 UWG, GDPREmail advertising only with prior express consent. Unsubscribe link in every message.
  • Price informationTotal prices with all mandatory components in the booking.
  • Portal contractsExisting contracts are reviewed by the hotel's law firm before the start.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

What is measured is the share of direct bookings. Every week shows which step moves it.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The enquiry arrives on Saturday evening. The appointment is set two minutes later.

Assumed: estate agency with seven employees and 60 active properties.

The situation in figures

  • 57% of companies in Germany with 20 or more employees use AI18
  • 72% name customer contact as a field of use18

Built in the example

The request in the example

“One listing brings 200 enquiries. We answer half of them and lose the best.”

Example with model values

  1. Analysis
  2. Knowledge
  3. Pilot
  4. Operation
  1. Sat 21:40Customer

    Is the flat in Gartenstraße still available? I would like to view it.

  2. Sat 21:40AI assistant

    I am the office's AI assistant. The flat is available. Viewings are possible on Tuesday at 5 pm and Thursday at 6 pm.

  3. Sat 21:42Customer

    Tuesday works.

  4. Sat 21:42Appointment

    Viewing booked, list of documents sent to the prospect.

  5. Mon 09:10Selection

    The agent reviews the enquiries and decides whom to invite.

What the review shows

  • In the example six in ten enquiries arrive in the evening and at weekends.
  • Nine in ten questions are answered in the brochure.

What gets built

  • An AI assistant answers questions from the brochure and books viewings.
  • The selection of prospects is made by a human.

Assumed result in the example

target until the first answer, weekends included
2 min
of enquiries receive an answer in the model, half before
100%
decides who is invited
Human

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • AGG (Equal Treatment Act)The AI makes no selection among prospects. A human decides on objective criteria.
  • GDPR Art. 5, 6Documents are only requested when the next step requires them.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI answers and schedules. Who gets the flat is decided by a human.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The enquiry becomes a quote in six minutes. The dispatcher decides the price.

Assumed: haulier with 45 vehicles and 31 office employees.

The situation in figures

  • 51% of logistics companies in Germany use AI, 54% lack the expertise20
  • 133 insolvencies per 10,000 companies in transport and storage in 202522

Built in the example

The request in the example

“We receive 150 enquiries a day by email. Whoever answers first gets the load.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot
  4. Operation
  1. Thu 15:52Customer

    Six pallets from Ulm to Kassel, collection tomorrow morning. What does it cost?

  2. Thu 15:52Entry

    Enquiry read from the email: route, quantity, date. Dimensions are missing.

  3. Thu 15:53AI assistant

    I am the haulier's AI assistant. Are these Euro pallets and how high is the load?

  4. Thu 15:55Customer

    Euro pallets, 1.60 metres.

  5. Thu 15:58Approval

    The dispatcher checks the price suggestion and sends the quote.

What the review shows

  • In the example a quote takes 50 minutes. Half of that is retyping.
  • Details are missing in every third enquiry.

What gets built

  • Enquiries are read and transferred into dispatch.
  • An AI assistant asks for missing details.
  • Price and commitment are approved by the dispatcher.

Assumed result in the example

target until the quote, 50 before
6 min
less retyping per week in the model
25 hrs
approves every price
Human

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Price and contractQuotes are sent by a human.
  • Driver dataDrivers and the works council are involved before vehicle data is used.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

No own planning system is built here. What is automated is the path from the email into the existing system.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The dealer configures himself. Price and drawing arrive within an hour.

Assumed: manufacturer of shelving systems with 90 employees and 300 dealers.

The situation in figures

  • Manufacturers and wholesalers in Germany sold €549 billion through online shops and marketplaces in 2025, up 7.9%4
  • 20% of SMEs use AI8

Launched in the example

The request in the example

“Every enquiry goes back and forth three times until dimensions and variants are right. That costs us orders.”

Example with model values

  1. Day 1 to 30Rules

    Dimensions, loads and variants of one product family

  2. Day 31 to 60Dealers

    20 dealers configure themselves and receive price and drawing

  3. Day 61 to 90Ordering

    The configuration becomes an order in the system

Configured enquiries per week: plan and assumed run

Enquiries per week
48Plan 41
Queries per enquiry
0Plan 1
Dealers active
23Plan 20
Hours to the price
1Plan 4
Wrong orders
1%Plan 2%
  1. Day 30Special dimensions stay with inside sales in the example.
  2. Day 60Dealers use the configurator more often than planned in the example.
  3. Day 90In the example the run is above plan. The second product family is being prepared.

What the review shows

  • In the example an enquiry needs three queries and two days.
  • Eight in ten enquiries can be covered by fixed rules.

What gets built

  • Configurator for one product family with price and drawing.
  • Special dimensions go to inside sales.
  • The order lands in the system without retyping.

Assumed result in the example

to price and drawing in the model run, two days before
1 hr
per week in the model run, plan 41
48 enquiries
active after 90 days in the model run, plan 20
23 dealers
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Scale
  5. Launched in the example

What gets built is what can be put into rules. The exception stays with a human.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Administration is done in draft. Advice gets the time.

Assumed: insurance broker with six employees and 3,200 customers.

The situation in figures

  • Bureaucracy costs an SME in Germany 32 hours a month on average7
  • 72% of companies using AI name customer contact as a field of use18

Built in the example

The request in the example

“We administer the portfolio instead of advising. Yet advice is what we earn from.”

Example with model values

The broker's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    2 hrs free
  • Sun
    0 hrs free

12 hours a week returned for advice in the example

  • Portfolio

    Change of address, certificate, payment method: the draft comes from the customer file. A human sends it.

  • Claims

    The claim is recorded in full and prepared for the insurer.

  • Advice

    The time gained goes into conversations. Advice and documentation stay with the broker.

What the review shows

  • In the example seven in ten cases are administration without advice.
  • Claims arrive incomplete and cost three queries.

Assumed result in the example

a week returned for advice, assumed
12 hrs
per claim as the target, three before
1 query
advises and sends
Human

Law and standards in view

  • § 61 VVGAdvice and its documentation stay with the broker.
  • GDPR Art. 9, 28Health details in contracts only for the case at hand. Providers are bound by contract.
  • SendingEvery message to customers and insurers is sent by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI takes over administration. The recommendation to the customer is given by the broker.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

An AI is meant to select applicants. The review says: the AI organises, humans select.

Assumed: staffing firm with 25 employees and 4,000 applications a year.

The situation in figures

  • Companies in Germany that considered AI and do not use it: 62% cite legal uncertainty, 60% data protection21
  • 57% of companies with 20 or more employees use AI18

Advised against in the example, with counterproposal

The request in the example

“We want an AI to rate the applications and propose the best.”

Example with model values

  • Riskhigh

    Legal hurdles

    • AI for selecting applicants can count as high risk (AI Act Art. 6, Annex III).
    • Discrimination under the Equal Treatment Act is hard to rule out in automated pre-selection.
  • Riskraised

    Data

    • CVs contain details that must not be used for selection.
  • Riskopen

    Benefit

    • In the example most time goes into scheduling and feedback, not selection.

Do not start like this

The counterproposal

Same goal: fill positions faster and lose no good applicant.

  • The AI organises: confirm receipt, arrange appointments, give feedback within 24 hours.
  • Selection is made by humans against fixed criteria.
  • Pilot with three client companies over eight weeks.
  • Planned matching, company budget
    €160,000
  • Organisation and pilot, company budget
    €26,000

Assumed result in the example

stays in the budget in the example until the pilot shows results
€134,000
target until feedback to applicants, nine days before
24 hrs
makes every selection
Human
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Advised against in the example, with counterproposal

We name legal hurdles as a risk. The legal review is carried out by a law firm of your choice.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Every branch gets the quantity it sells. In the model that carries in two of three scenarios.

Assumed: bakery with 14 branches and 120 employees.

The situation in figures

  • Around 564,000 skilled trades companies in Germany, reporting year 202423
  • 72% say they have too much to do to deal with digitalisation. 4% use AI19

Launched in the example

The request in the example

“In the evening we throw goods away, and at noon the pretzels are sold out. Each branch orders by gut feeling.”

Example with model values

Cumulative cash balance of the order forecast € thousand

Carries in the model
  • Zero line: the money invested is back
  • Favourable 95
  • Expected 34
  • Unfavourable −26

What the review shows

  • In the example 11% of goods come back in the evening.
  • On every third day a main item is sold out before noon.
  • Till data, weather and calendar are available but not used.

What gets built

  • Order suggestion per branch and item from till data, weather and calendar.
  • The branch manager approves and can change it.
  • Start in four branches, expansion after eight weeks.

Assumed result in the example

scenarios reach the zero line within twelve months in the model
2 of 3
target for returned goods, 11% before
7%
at the start instead of all 14
4 branches
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The forecast is a suggestion. The branch manager knows the shop and decides.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The machine reports before it stops. The start is with three machines.

Assumed: machining company with 28 machining centres and 85 employees.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • 57% of companies with 20 or more employees use AI18

Launched in the example

The request in the example

“Every unplanned stop costs us a shift. We want to know when a spindle will fail.”

Example with model values

Cumulative cash balance of predictive maintenance € thousand

Carries in the model
  • Zero line: the money invested is back
  • Favourable 150
  • Expected 66
  • Unfavourable −20

What the review shows

  • In the example seven in ten stops occur on three machines.
  • The controllers deliver data nobody evaluates so far.
  • Sensors on all 28 machines would not pay in the model.

What gets built

  • Evaluation of the existing controller data on three machines.
  • Warning to maintenance with its reason. A human decides on the intervention.
  • Expansion only once two predicted failures are confirmed.

Assumed result in the example

scenarios reach the zero line within 18 months in the model
2 of 3
at the start instead of all 28
3 machines
decides on every intervention
Human
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The forecast does not replace maintenance. It tells it where to look first.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The camera inspects the clear parts. The human inspects the borderline cases.

Assumed: injection moulder with 60 employees and 40,000 visual inspections a day.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • Companies that considered AI and do not use it: 72% lack the knowledge21

Launched in the example

The request in the example

“Three people look at parts all day. After six hours nobody sees every scratch any more.”

Example with model values

  1. Day 1 to 30Learn

    One part, 4,000 images, defect catalogue with quality assurance

  2. Day 31 to 60Shadow

    The camera inspects in parallel, the human decides

  3. Day 61 to 90Take over

    Clear cases automatically, borderline cases to the inspection station

Automatically inspected parts in percent: plan and assumed run

Inspected automatically
78%Plan 70%
Missed defects
0Plan 0
False alarms
4%Plan 5%
Parts
1Plan 1
Hours at the station per day
4Plan 5
  1. Day 30Shiny surfaces disturb the image in the example. The lighting is changed.
  2. Day 60In shadow mode the camera misses no defect in the example.
  3. Day 90In the example the run is above plan. The second part is being prepared.

What the review shows

  • In the example eight in ten parts are clearly good.
  • Defects cluster in the last hour of every shift.

What gets built

  • Camera and evaluation for one part, first in shadow mode.
  • Borderline cases go to the inspection station. Quality assurance approves the rule.

Assumed result in the example

of parts inspected automatically in the model run, plan 70%
78%
missed defects in shadow mode in the example
0
at the start instead of the whole range
1 part

Law and standards in view

  • CameraThe camera captures parts and no people. The works council is involved.
  • Quality responsibilityInspection rules are approved by quality assurance.
  • DocumentationEvery decision of the camera is stored with its image.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Shadow first, then take over. The rule for what passes automatically stays with quality assurance.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The plan is set at seven. After a disruption it is set again ten minutes later.

Assumed: contract sheet metal manufacturer with 70 employees and 300 open orders.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • 20% of SMEs use AI8

Built in the example

The request in the example

“I plan in a spreadsheet. When a machine stops, I replan for half the day.”

Example with model values

The production planner's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    3 hrs free
  • Sun
    0 hrs free

13 hours a week returned in the example

  • Sequence

    Suggested order sequence from deadlines, set-up times and material. The planner approves.

  • Disruption

    If a machine fails, a new suggestion is ready within minutes.

  • Information

    Sales see the delivery date themselves and stop asking.

What the review shows

  • In the example the planner replans for two hours a day.
  • Sales ask about delivery dates 20 times a day.

Assumed result in the example

a week returned to the planner, assumed
13 hrs
target until the new plan after a disruption
10 min
approves every plan
Human
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

First we check whether a standard system is enough. Only what it does not cover gets built.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The machine stops on Saturday. The right part is found in two minutes.

Assumed: manufacturer of filling machines with 110 employees and 1,400 machines in the field.

The situation in figures

  • Manufacturers and wholesalers in Germany sold €549 billion through online shops and marketplaces in 2025, up 7.9%4
  • 72% of companies using AI name customer contact as a field of use18

Built in the example

The request in the example

“The customer calls when the machine is down. Then we search old bills of material for an hour.”

Example with model values

  1. Analysis
  2. Catalogue
  3. Pilot
  4. Operation
  1. Sat 06:20Customer

    Our filling line is down. We need the seal on the filling head, built in 2014.

  2. Sat 06:20AI assistant

    I am the manufacturer's AI assistant. Seal kit 77 12 04 fits your machine. It is in stock. Shall I prepare it as an express shipment?

  3. Sat 06:22Customer

    Yes, please.

  4. Sat 06:22Order

    Order created at the customer price, warehouse notified.

  5. Sat 07:05On call

    The service technician confirms the part and approves the shipment.

What the review shows

  • In the example finding the right part takes 50 minutes on average.
  • Every fifth delivery is the wrong part.

What gets built

  • Catalogue per machine from bills of material and serial numbers.
  • An AI assistant finds the part and prepares the order.
  • The technician confirms before shipping.

Assumed result in the example

target until the right part, 50 before
2 min
target for wrong deliveries, 20% before
2%
confirms every part before shipping
Human

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Safety partsParts with a safety function are always approved by a technician.
  • PricesThe prices from the customer's contract apply.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Scale
  5. Built in the example

The knowledge sits in the bills of material. The assistant makes it accessible at the moment the machine is down.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The manual is drafted. Safety notices are still written by a human.

Assumed: special purpose machine builder with 95 employees and 40 machines a year.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • Bureaucracy costs an SME 32 hours a month on average7

Built in the example

The request in the example

“Every machine is different, and each needs a manual in five languages. Documentation delays delivery.”

Example with model values

The technical writers' week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    0 hrs free
  • Sun
    0 hrs free

10 hours a week returned per writer in the example

  • Draft

    Draft operating manual from design data and earlier manuals.

  • Languages

    Draft translation using the company's terminology. A specialist reviews.

  • Review

    Safety notices are written and reviewed by the writers themselves.

What the review shows

  • In the example every third machine waits for its manual.
  • 70% of the text repeats from machine to machine.

Assumed result in the example

a week returned per writer, assumed
10 hrs
target until the manual, 15 before
5 days
writes and reviews every safety notice
Human

Law and standards in view

  • Machinery lawSafety notices and risk assessment stay with the manufacturer and its specialists.
  • TranslationEvery language version is reviewed by a specialist before delivery.
  • Design dataThe data stays with the manufacturer. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI writes what repeats. What decides about safety is written by the writers.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Four quotes, one table. The buyer negotiates instead of retyping.

Assumed: pump manufacturer with 160 employees and three people in purchasing.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • 57% of companies with 20 or more employees use AI18

Built in the example

The request in the example

“Every supplier answers in its own format. By the time we can compare, half the day is gone.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot
  4. Operation
  1. Tue 09:00Demand

    Demand from planning: 4,000 castings, delivery in six weeks.

  2. Tue 09:02Enquiry

    Enquiry to five approved suppliers prepared.

  3. Tue 09:10Approval

    The buyer checks and sends the enquiry.

  4. Thu 14:00Comparison

    Four quotes in one table: price, date, payment terms, deviations from the drawing.

  5. Thu 15:30Award

    The buyer decides and negotiates.

What the review shows

  • In the example comparing quotes takes four hours.
  • Deviations from the drawing are often noticed only on delivery.

What gets built

  • Quotes are read and transferred into a comparison table.
  • Deviations from the enquiry are flagged.
  • Enquiry, award and negotiation stay with purchasing.

Assumed result in the example

target per quote comparison, four hours before
30 min
per demand in the comparison, two on average before
4 quotes
decides every award
Human
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI prepares the comparison. The decision on the supplier is made by purchasing.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

What the night shift knew, the early shift knows at six.

Assumed: beverage producer with three shifts, four lines and 130 employees.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • Companies that considered AI and do not use it: 72% lack the knowledge21

Built in the example

The request in the example

“The handover is on a slip of paper. When the experienced colleague retires, his knowledge goes with him.”

Example with model values

  1. Analysis
  2. Knowledge
  3. Pilot
  4. Operation
  1. Mon 05:50Handover

    Night shift: line 2 stopped twice, sensor on the capper replaced, batch 318 blocked.

  2. Mon 06:40Customer

    Line 2 reports fault 41 again. What was the cause last time?

  3. Mon 06:40AI assistant

    Fault 41 occurred nine times on line 2. Seven times the capper sensor was the cause, twice the conveyor. The sensor was replaced last night.

  4. Mon 06:55Decision

    The shift leader checks the conveyor and records the cause.

What the review shows

  • In the example every third fault is investigated from scratch a second time.
  • Four employees know the causes of the most frequent faults.

What gets built

  • Handover by voice or text, producing a log per line.
  • An assistant answers questions on faults from the company's own logs.
  • The cause is always recorded by a human.

Assumed result in the example

target until the likely cause, one hour before
15 min
hands over with a log, in the example
Every shift
records every cause
Human

Law and standards in view

  • Co-determinationThe works council is involved. Faults are evaluated, not people.
  • Product safetyBlocking and releasing batches is decided by quality assurance.
  • DataThe logs stay inside the company.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The knowledge of the experienced stays in the business. Faults are assessed, not people.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Two furnaces start at staggered times. The peak load drops, production does not.

Assumed: heat treatment shop with six furnaces and 45 employees.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • 20% of SMEs use AI8

Launched in the example

The request in the example

“Our electricity bill depends on a few quarter hours a year. We just do not know which ones.”

Example with model values

Cumulative cash balance of load management € thousand

Carries in the model
  • Zero line: the money invested is back
  • Favourable 92
  • Expected 40
  • Unfavourable −8

What the review shows

  • In the example the expensive peaks arise when three furnaces heat up at the same time.
  • The metering data is available in quarter hour intervals.

What gets built

  • Load forecast from the furnace plan and metering data.
  • Suggestion for staggered starts. The shift management approves.
  • Start with two furnaces.

Assumed result in the example

scenarios reach the zero line within twelve months in the model
2 of 3
lower peak load as the target, assumed
12%
at the start instead of all six
2 furnaces
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The saving depends on the tariff. That is why the review first calculates three scenarios.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€2.4 million for a fully automated hall. The review says: one cell first.

Assumed: turning shop with 55 employees and 16 machines.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • Companies that considered AI and do not use it: 72% lack the knowledge21

Advised against in the example, with counterproposal

The request in the example

“We cannot find people any more. We automate the whole hall, then the problem is gone.”

Example with model values

  • Riskhigh

    Finance

    • In the model the fully automated hall only pays at double the utilisation.
    • The financing ties the company for seven years.
  • Riskraised

    Order situation

    • In the example the parts change every two weeks. Every changeover costs a day.
  • Riskraised

    Knowledge

    • Nobody in the company has operated a robot cell.

Do not start like this

The counterproposal

Same goal: more parts with the same people.

  • One cell at the machine with the highest volume.
  • Operation in the night shift, daytime as before.
  • After six months the numbers decide on the second cell.
  • Planned hall, company budget
    €2,400,000
  • One cell, company budget
    €190,000

Assumed result in the example

stays in the budget in the example until the first cell delivers numbers
€2,210,000
until the numbers decide on the second cell
6 months
instead of the whole hall
1 cell
  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

The problem is real, the leap too big. The review turns it into a step that can be measured.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

In-house production software for €600,000: none of the scenarios carries in the model.

Assumed: plastics processor with 120 employees and 30 machines.

The situation in figures

  • 209,815 legal units in manufacturing in Germany, reporting year 202444
  • 20% of SMEs use AI8

Advised against in the example

The request in the example

“No system fits us. We will have our own production software developed.”

Example with model values

Cumulative cash balance of the in-house production software € thousand

Advised against
  • Zero line: the money invested is back
  • Favourable −380
  • Expected −850
  • Unfavourable −1,240

Why we advise against it

  • Market: mature systems exist for machine connection and feedback.
  • Operation: in-house software needs developers permanently, whom the company does not have.
  • Need: nine in ten requirements in the example are covered by a standard system.

What happens instead

  • Standard system for machine data and feedback.
  • Only the one missing evaluation gets built.
  • The budget stays for machines.

Assumed result in the example

scenarios reach the zero line within 36 months in the model
0 of 3
shortfall in the middle scenario of the model after 36 months
€850,000
requirements are covered by a standard system in the example
9 of 10
  1. Pre-flight check
  2. Validate
  3. Advised against in the example

A no from Validate costs €4,900 net, planning value five working days. We like to build. But only what does not exist yet.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The delivery note is photographed. Only what deviates gets counted.

Assumed: fastener wholesaler with 9,000 storage locations and 28 warehouse employees.

The situation in figures

  • 107,052 legal units in transport and storage in Germany, reporting year 202444
  • 51% of logistics companies use AI, 54% lack the expertise20

Built in the example

The request in the example

“In the morning six lorries are at the ramp. By the time everything is booked it is noon, and sales cannot see the goods.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot
  4. Operation
  1. Mon 07:15Delivery note

    Photo of the delivery note read: 14 lines, supplier and order number recognised.

  2. Mon 07:15Match

    Twelve lines match the order. Two differ in quantity.

  3. Mon 07:18Check

    The warehouse worker recounts the two lines and confirms.

  4. Mon 07:19Booking

    Goods receipt booked, deviation reported to purchasing.

What the review shows

  • In the example every delivery note is typed in by hand.
  • Nine in ten deliveries match the order.

What gets built

  • Read the delivery note from a photo and match it with the order.
  • Deviations go to a human.
  • Booking happens in the existing system.

Assumed result in the example

target per delivery, 25 before
4 min
target for goods being available, noon before
9 am
checks every deviation
Human
  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

No new warehouse system. What is automated is the step from paper into the existing system.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The most frequent items move to the front. The routes get shorter, the shifts calmer.

Assumed: mail order dealer for workshop supplies with 22,000 items and 40 warehouse employees.

The situation in figures

  • 107,052 legal units in transport and storage in Germany, reporting year 202444
  • €83.1 billion online retail of goods in 2025, gross, up 3.2%1

Launched in the example

The request in the example

“Our people walk their feet off, and in the season we only manage the orders with overtime.”

Example with model values

  1. Day 1 to 30Measure

    Routes and picks from the existing warehouse data

  2. Day 31 to 60Relocate

    The 300 most frequent items to the front

  3. Day 61 to 90Routes

    Orders are bundled into rounds

Lines picked per hour: plan and assumed run

Lines per hour
71Plan 67
Distance per order
104 mPlan 120 m
Mispicks
0.3%Plan 0.4%
Relocated items
300Plan 300
Overtime per week
12Plan 20
  1. Day 30In the example 60% of the walking is for 300 items.
  2. Day 60Relocation takes effect faster than planned in the example.
  3. Day 90In the example the run is above plan. The routes are extended to the second hall.

What the review shows

  • In the example the most frequent items are spread across the whole hall.
  • Every order is walked individually.

What gets built

  • Evaluation of the existing warehouse data and a suggestion for new locations.
  • Orders are bundled into rounds.

Assumed result in the example

per hour in the model run, plan 67
71 lines
distance per order in the model run, 150 before
104 m
overtime per week in the model run, 35 before
12 hrs

Law and standards in view

  • Co-determinationThe works council is involved. Routes and items are evaluated, not individual employees.
  • Employee dataNo performance comparison between people.
  • Occupational safetyHeavy items stay at grip height.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Routes and items are measured. Nobody is compared with a stopwatch.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Ten minutes of counting per shift. The standstill in January is gone.

Assumed: parts dealer with 14,000 storage locations and 30 employees.

The situation in figures

  • 107,052 legal units in transport and storage in Germany, reporting year 202444
  • Bureaucracy costs an SME 32 hours a month on average7

Launched in the example

The request in the example

“For stocktaking we close for three days. And by March the stock is wrong again anyway.”

Example with model values

  1. Day 1 to 30Rule

    Counting happens where a deviation is likely

  2. Day 31 to 60Routine

    Ten minutes of counting per shift instead of a standstill in January

  3. Day 61 to 90Evidence

    Agreement with tax adviser and auditor

Storage locations counted per week: plan and assumed run

Locations per week
445Plan 400
Stock correct
98%Plan 97%
Minutes per shift
9Plan 10
Shortages in dispatch
0.6%Plan 1.0%
Standstill days
0Plan 0
  1. Day 30In the example eight in ten deviations concern small parts.
  2. Day 60Counting runs along in every shift in the example.
  3. Day 90In the example the procedure is agreed with the auditor.

What the review shows

  • In the example the year end stocktake costs three days of sales.
  • Shortages are only noticed by the customer.

What gets built

  • Daily counting suggestion for the locations with the highest risk of deviation.
  • Counting on the handheld, deviation booked immediately.

Assumed result in the example

of standstill for stocktaking in the model, three before
0 days
of stock is correct in the model run, 93% before
98%
counted per week in the model run, plan 400
445 locations

Law and standards in view

  • § 241 HGB (German Commercial Code)Whether the procedure is recognised as stocktaking is clarified by the company's tax adviser and auditor.
  • EvidenceEvery count is recorded with time and location.
  • Co-determinationThe works council is involved.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Technology is the smaller part. What matters is that tax adviser and auditor support the procedure.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The return is sellable again in three minutes. The refund goes out the same day.

Assumed: online retailer for sportswear with 35 employees and 300 returns a day.

The situation in figures

  • €15.1 billion online clothing sales in Germany in 2025, up 3.5%1
  • 44.2% of fashion sales already happen online3

Launched in the example

The request in the example

“There are 40 pallets in the returns area. The goods are missing from the shop, and customers are waiting for their money.”

Example with model values

  1. Day 1 to 30Sort

    Three routes: back to the shelf, refurbish, discard

  2. Day 31 to 60Recognise

    Photo and suggestion for condition and route, a human confirms

  3. Day 61 to 90Refund

    Refund on the day of receipt

Returns processed per day: plan and assumed run

Returns per day
360Plan 330
Minutes per return
3Plan 4
Back to the shelf
74%Plan 70%
Days to the refund
1Plan 2
Backlog in pallets
2Plan 5
  1. Day 30In the example the backlog is 40 pallets. Sorting comes first, then recording.
  2. Day 60The condition suggestion is right in nine of ten cases in the example.
  3. Day 90In the example the run is above plan. The refund goes out on the day of receipt.

What the review shows

  • In the example a return sits for nine days before it is back in the shop.
  • Seven in ten items are without defect.

What gets built

  • Recording with photo and suggestion for condition and route. A human confirms.
  • The refund is triggered with the confirmation.

Assumed result in the example

per return in the model run, seven before
3 min
to the refund in the model run, nine before
1 day
backlog in the model run, 40 before
2 pallets

Law and standards in view

  • § 357 BGB (German Civil Code)The refund goes out within the statutory period. The target in the example is far below it.
  • Co-determinationThe works council is involved. No performance comparison between people.
  • ConditionDeductions for loss of value are decided by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Every day in the returns area is tied up stock. What is measured is the time until back in the shop.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The retailer asks on Friday evening. The stock figure is there within the same minute.

Assumed: fulfilment provider with 60 retailer customers and 50 employees.

The situation in figures

  • 107,052 legal units in transport and storage in Germany, reporting year 202444
  • 51% of logistics companies use AI, 54% lack the expertise20
  • 133 insolvencies per 10,000 companies in transport and storage in 202522

Built in the example

The request in the example

“The retailers ask about stock and shipments all day. Two people do nothing but answer.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot
  4. Operation
  1. Fri 17:40Customer

    How many of item 2210 do you still have, and when does the replenishment arrive?

  2. Fri 17:40AI assistant

    I am the warehouse's AI assistant. Stock is 412 units. The container is announced for Tuesday, available from Wednesday.

  3. Fri 17:42Customer

    Can you handle Monday's promotion with 600 parcels?

  4. Fri 17:42AI assistant

    That is a capacity commitment. I am passing it to your account manager. She will reply by Monday eight o'clock.

  5. Mon 07:30Commitment

    The account manager commits and schedules a second packing line.

What the review shows

  • In the example eight in ten enquiries are information already in the system.
  • Commitments on capacity and price need a human.

What gets built

  • An AI assistant answers questions on stock, goods receipt and shipments from the warehouse system.
  • Every retailer sees only its own data.
  • Commitments are given by account management.

Assumed result in the example

target until the information, evenings included
1 min
of enquiries are information from the system in the model
80%
gives every commitment
Human

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Client separationA retailer never receives another retailer's data.
  • CommitmentsCapacity, price and dates are committed by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI gives information. What the warehouse promises is promised by a human.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Three plots, 60 flats. Only one of the sites carries in the model.

Assumed: residential developer with 22 employees planning its next project of 60 flats.

The situation in figures

  • From January to August 2026 a total of 169,400 dwellings were approved in Germany, 12.1% more than a year earlier9
  • Prices for building new residential buildings were 5.5% higher in August 2026 than a year earlier11
  • Condominiums in the second quarter of 2026: 0.4% cheaper than a year earlier in the seven largest cities, 2.0% dearer in other large cities12

Launched in the example

The request in the example

“We have three plots on offer. Where should we build?”

Example with model values

Cumulative cash balance per site € thousand

Site B carries in the model
  • Zero line: the money invested is back
  • Site B, mid-sized town with inflow 700
  • Site A, metropolis −900
  • Site C, rural district −1,800

What the review shows

  • Site A: highest prices, but plot and construction time tie up the most capital in the model.
  • Site B: medium prices, short marketing period, inflow according to the municipality's official statistics.
  • Site C: cheap plot, in the model demand is not enough for 60 flats.

What gets built

  • Comparison of the three sites in one model: plot, construction cost, construction time, marketing.
  • Three scenarios per site for interest rates and sales duration.
  • Decision paper for management and the financing bank.

Assumed result in the example

sites reaches the zero line within 36 months in the model
1 of 3
for interest rates and sales duration, three per site
9 scenarios
to the decision paper (planning value)
5 working days

Law and standards in view

  • No investment adviceThe example shows a profitability calculation for a company. It is not a recommendation to buy a property.
  • Building lawDevelopment plan and permit prospects are checked by the developer's architect and law firm.
  • FinancingFinancing is decided by the company's bank.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launched in the example

The sites in the example are invented. For a real venture we calculate with the data of the actual plots.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

48 new flats are planned. The review finds them in the existing portfolio.

Assumed: housing cooperative with 40 buildings, 1,100 flats and 18 employees.

The situation in figures

  • Prices for building new residential buildings in Germany were 5.5% higher in August 2026 than a year earlier11
  • Net cold rents were 1.9% higher in October 2026 than a year earlier13
  • Just under 60% of the roughly 19.8 million residential buildings were built before the first thermal insulation ordinance of 197716

Advised against in the example, with counterproposal

The request in the example

“Our waiting list gets longer every year. We will build a new block with 48 flats.”

Example with model values

  • Riskhigh

    Construction cost

    • In the model the rent the new building needs is above what can be achieved at the site.
    • New build prices rise faster than rents.
  • Riskraised

    Time

    • In the model 30 months pass from permit to moving in.
  • Riskraised

    Existing stock

    • In the example twelve of the 40 own buildings need a new heating system within five years.

Do not start like this

The counterproposal

Same goal: 48 additional flats in the own portfolio.

  • Adding storeys to four buildings with 28 flats in total.
  • Converting two attics and one commercial unit: 20 flats.
  • Heating renovation runs in the same construction phases.
  • Planned new build, company budget
    €14.4 million
  • Added storeys and conversion, company budget
    €8.9 million

Assumed result in the example

in the existing stock instead of a new build, in the example
48 flats
stays in the budget in the example
€5.5 million
are renewed in the same step
12 heating systems

Law and standards in view

  • Building lawWhether storeys may be added is clarified by architect, structural engineer and building authority.
  • SubsidiesWe name subsidy programmes as an option. The application is supported by an energy consultant.
  • Tenancy lawConsequences for existing tenancies are checked by the cooperative's law firm.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

The question was where to build. The better question was what is already there.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Seven quotes, one table. Who builds is decided by people.

Assumed: metal fabricator with 80 employees building a new production hall.

The situation in figures

  • 388,229 legal units in construction in Germany, reporting year 202444
  • 104 insolvencies per 10,000 companies in construction in 2025, 69 on average across all sectors22
  • Construction prices for commercial buildings were 5.5% higher in August 2026 than a year earlier11

Built in the example

The request in the example

“We build once in 20 years. How are we supposed to know whom to do it with?”

Example with model values

  1. Scope
  2. Enquiry
  3. Comparison
  4. Award
  5. Construction
  1. Mon 09:00Enquiry

    Scope for the hall prepared for six contractors and two general contractors.

  2. Mon 09:20Approval

    The managing director checks the list and sends the enquiry.

  3. Fri 16:00Comparison

    Seven quotes in one table: price, construction time, payment schedule, exclusions, references, credit report.

  4. Fri 16:05Note

    Two quotes exclude the floor slab. One bidder has a conspicuous credit report.

  5. Tue 10:00Award

    Managing director, architect and bank decide after two meetings.

What the review shows

  • In the example the quotes are not comparable: each bidder excludes something different.
  • After adjustment the cheapest quote is the third most expensive.

What gets built

  • Uniform scope with the architect.
  • Quotes are read, exclusions and deviations flagged.
  • Award and contract stay with management, architect and law firm.

Assumed result in the example

in one adjusted table
7 quotes
found that shift the price by 9% in the example
2 exclusions
decides the award
Human

Law and standards in view

  • Construction contractThe contract is reviewed by the company's law firm.
  • CreditworthinessReports come from credit agencies. The assessment is made by a human.
  • No recommendationWe do not recommend a contractor. We make quotes comparable.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

We do not say whom you should build with. We make sure you compare like with like.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Buy, build or keep renting: three routes calculated over twelve years.

Assumed: packaging manufacturer with 65 employees whose lease expires in two years.

The situation in figures

  • SMEs in Germany invested around €218 billion in new equipment and buildings in 202517
  • Construction prices for commercial buildings were 5.5% higher in August 2026 than a year earlier11

Launched in the example

The request in the example

“The hall is too small and the lease is running out. Should we buy, build or rent somewhere else?”

Example with model values

Cumulative cash balance compared with continuing to rent € thousand

Buying carries in the model
  • Zero line: the money invested is back
  • Buy and convert an existing hall 620
  • Build new 130
  • Rent larger premises −560

What the review shows

  • In the model renting is cheapest in the first six years and most expensive over twelve.
  • A new build ties up two years of management time.
  • In the example three suitable existing halls are for sale within 25 kilometres.

What gets built

  • One model with three routes and three scenarios each for interest rates and growth.
  • Requirement profile for estate agent and architect.
  • Documents for the meeting with bank and tax adviser.

Assumed result in the example

over twelve years in one model
3 routes
in which buying overtakes renting in the model
Year 8
to the decision paper (planning value)
5 working days

Law and standards in view

  • No investment adviceA business decision is calculated. It is not a recommendation to buy a particular property.
  • TaxDepreciation and legal form are clarified by the tax adviser.
  • FinancingLoan and collateral are decided by the bank.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launched in the example

The calculation takes nothing away from bank and tax adviser. It makes sure the meeting with them is short.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

An office building for €6.8 million because prices have fallen. The review advises against it.

Assumed: owner family of a wholesale business with 90 employees.

The situation in figures

  • According to the vdp index, office property in Germany was 3.9% cheaper in the third quarter of 2026 than a year earlier, residential property 1.0% dearer14
  • Rents in new office leases rose by 2.5% over the same period14
  • Construction prices for office buildings were 5.7% higher in August 2026 than a year earlier11

Advised against in the example, with counterproposal

The request in the example

“Offices are cheap right now. We will buy an office building in the city centre and let it.”

Example with model values

  • Riskhigh

    Market

    • Office prices keep falling in the index. A low price is not yet a floor.
    • In the example two of six floors in the building are vacant.
  • Riskhigh

    Finance

    • In the model the rent only covers the loan at full occupancy.
    • There is no reserve for renovating the building services.
  • Riskraised

    Knowledge

    • The family has not managed a commercial property so far.

Do not start like this

The counterproposal

Same goal: place company assets in a property that carries itself.

  • Check the own company site first: extension instead of acquisition.
  • If acquiring, then a property with one tenant the family knows.
  • Every property goes through the same calculation with three scenarios.
  • Planned office building, company budget
    €6.8 million
  • Extension at the own site, company budget
    €2.1 million

Assumed result in the example

stays in the company in the example
€4.7 million
for vacancy, interest and renovation per property
3 scenarios
to the decision paper (planning value)
5 working days

Law and standards in view

  • No investment adviceWe calculate a venture of the company. We recommend no financial investment and broker no properties.
  • Tax and lawAcquisition, legal form and leases are clarified by tax adviser and law firm.
  • FinancingFinancing is decided by the bank.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

Cheap is not a reason. The review asks whether the property carries itself in the bad scenario too.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

120 micro apartments in the metropolis: in the model no scenario comes back within 36 months.

Assumed: project developer with twelve employees and a plot in prospect.

The situation in figures

  • Condominiums in Germany's seven largest cities were 0.4% cheaper in the second quarter of 2026 than a year earlier12
  • The effective interest rate on new housing loans to households stood at 4.09% in September 2026, up from 3.74% a year earlier15
  • In 2025 it took an average of 27 months from permit to completion of a new residential building10

Advised against in the example

The request in the example

“Small flats in the big city always sell. We will build 120 micro apartments and sell them one by one.”

Example with model values

Cumulative cash balance of the project € thousand

Advised against
  • Zero line: the money invested is back
  • Favourable −2,100
  • Expected −6,100
  • Unfavourable −10,400

Why we advise against it

  • Prices: in the metropolises condominium prices are easing slightly.
  • Buyers: in the model far fewer buyers than planned can afford the instalment at 4% interest.
  • Time: selling 120 units one by one takes longer in the model than the bridge financing runs.

What happens instead

  • Review a smaller project with 40 flats of mixed size.
  • Calculate a sale to one long term owner instead of unit sales as a second variant.
  • Secure the plot only once one variant carries.

Assumed result in the example

scenarios reach the zero line within 36 months in the model
0 of 3
shortfall in the middle scenario of the model after 36 months
€6.1 million
are calculated instead
2 variants

Law and standards in view

  • No investment adviceThe example calculates a construction project. It is not a statement about a market or a property.
  • § 34c GewO, MaBVThe developer's duties when selling are clarified by the company's law firm.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example

A no from Validate costs €4,900 net, planning value five working days. By then the plot has not been bought yet.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

40 buildings, ten years, one roadmap. First the seven where it is urgent.

Assumed: property management looking after 40 apartment buildings for an owner family.

The situation in figures

  • Just under 60% of Germany's roughly 19.8 million residential buildings were built before the first thermal insulation ordinance of 197716
  • 15% of oil and gas heating systems are older than 30 years16
  • Prices for maintaining residential buildings were 6.0% higher in August 2026 than a year earlier11

Launched in the example

The request in the example

“We know we have to renovate. We just do not know where to start.”

Example with model values

  1. Day 1 to 30Stock

    Year built, heating, consumption and condition of 40 buildings in one table

  2. Day 31 to 60Sequence

    Which buildings first: failure risk, consumption, subsidy option

  3. Day 61 to 90Roadmap

    Ten years, construction phases, budget per year

Buildings covered by the renovation roadmap: plan and assumed run

Buildings recorded
40Plan 40
Heating systems over 25 years
12Plan open
Buildings in the first phase
7Plan 6
Budget per year
€1.1 millionPlan €1.2 million
Years in the roadmap
10Plan 10
  1. Day 30Consumption data is missing for nine buildings in the example. It is collected afterwards.
  2. Day 60Seven buildings move to the front in the example: old heating and high consumption.
  3. Day 90The roadmap is in place in the example. The energy consultant reviews the first measures.

What the review shows

  • In the example the building data sits in folders, spreadsheets and people's heads.
  • So far renovation happens when something fails.

What gets built

  • One stock table from statements, maintenance logs and inspections.
  • Sequence by failure risk, consumption and subsidy option.
  • Roadmap over ten years with a budget per year.

Assumed result in the example

in one table instead of folders
40 buildings
in the first phase, in the example
7 buildings
roadmap with a budget per year
10 years

Law and standards in view

  • Energy lawWhich duties apply to which building is checked by the energy consultant.
  • SubsidiesWe name subsidy programmes as an option. Only the funding bodies give a commitment.
  • Tenancy lawPassing on costs and announcing measures is clarified by the law firm.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The roadmap does not replace an energy consultant. It tells them which buildings to start with.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Four sites, one view. The delay becomes visible before it gets expensive.

Assumed: residential developer with four running projects and three project managers.

The situation in figures

  • At the end of 2025 there were 760,700 dwellings in Germany approved but not yet completed10
  • In 2025 it took an average of 27 months from permit to completion of a new residential building10
  • 104 insolvencies per 10,000 companies in construction in 202522

Built in the example

The request in the example

“Every project runs in a different spreadsheet. We learn from the buyer that a trade is four weeks behind.”

Example with model values

The project manager's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    3 hrs free
  • Sun
    0 hrs free

13 hours a week returned in the example

  • Status

    Dates, costs and open items of all projects on one page, updated every morning.

  • Minutes

    The site meeting is minuted. Tasks go to the trades with a deadline. The project manager approves.

  • Warning

    If one trade slips, the plan shows which dates after it tip.

What the review shows

  • In the example every project manager writes minutes and reports for six hours a week.
  • Delay is noticed three weeks too late on average.

Assumed result in the example

a week returned per project manager, assumed
13 hrs
target until a delay is visible, three weeks before
3 days
approves every set of minutes and every reminder
Human

Law and standards in view

  • Contract deadlinesReminders and notices of obstruction are sent by a human. The wording is checked by the law firm where needed.
  • MinutesEvery set of minutes is approved by the project manager.
  • BuyersInformation to buyers is given by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

Construction time is interest time. What is measured is how early a delay becomes visible.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Enquiry on Sunday evening, preliminary estimate on Monday at ten. The master drives out when it is a fit.

Assumed: heating contractor with 14 employees and 30 heat pump enquiries a month.

The situation in figures

  • 49,983 plumbing and heating businesses were registered in Germany on 30 June 202626
  • 194,500 heat pumps were sold in Germany in the first half of 2026, up 40%38
  • Around 4 million heating systems in Germany are older than 30 years38

Built in the example

The request in the example

“Everyone wants a heat pump and asks three firms at once. I drive to site visits in the evening, and in the end one in four buys.”

Example with model values

  1. Analysis
  2. Questionnaire
  3. Pilot
  4. Operation
  1. Sun 20:14Customer

    We have a house from 1994 with oil heating. Roughly what would a heat pump cost us?

  2. Sun 20:14AI assistant

    Good evening, this is the firm's AI assistant. For a first assessment I need eight details and four photos: boiler room, the boiler's rating plate, meter cabinet and the spot for the outdoor unit.

  3. Sun 20:31Customer

    Photos sent. 160 square metres, 2,400 litres of oil a year, radiators throughout.

  4. Sun 20:32AI assistant

    Thank you. The photo of the meter cabinet is blurred, please send it again. The master will come back to you tomorrow morning with a preliminary estimate.

  5. Mon 07:05Draft

    Preliminary estimate created: air to water heat pump, price range, three open points (bathroom radiator, meter cabinet, distance to the neighbour).

  6. Mon 09:40Approval

    The master raises the range because the meter cabinet has to be replaced, and approves.

  7. Mon 09:45Sent

    Preliminary estimate to the customer, marked as non-binding, with two dates for the site visit.

What the review shows

  • In the example nine days pass from the enquiry to the first figure. In between lies a site visit of 90 minutes including travel.
  • Of 30 enquiries a month seven become jobs in the example. For twelve the site visit alone shows that the house or the budget does not fit.
  • That is an assumed 18 hours of the master's time a month for visits with no prospect.

What gets built

  • Questionnaire by messenger or on the website: year of construction, living space, consumption over the last three years, radiators or underfloor heating, space for the outdoor unit.
  • The customer sends photos of the boiler room, rating plate, meter cabinet and installation spot. The AI asks again when something is missing or blurred.
  • From this comes a preliminary estimate with a price range, open points and a note on whether a site visit is worthwhile. The master checks, changes and approves.
  • Heat load, system design and the binding quote follow the site visit as before.

Assumed result in the example

target from enquiry to preliminary estimate, nine days before in the example
1 day
of the master's time a month on visits with no prospect, meant to fall away in the example
18 hrs
approves every figure before it leaves the firm
Master

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Price and quoteThe preliminary estimate is marked as a non-binding range. The binding quote is written by the master after the site visit.
  • Design and subsidiesHeat load, design and noise distance are calculated by the specialist firm. The estimate makes no commitment on subsidies, the firm clarifies the conditions with the customer.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The preliminary estimate does not replace a site visit. It shows which site visit is worth making.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Saturday, 21:47: heating cold, baby in the house. The fitter on call sees the case sorted and decides.

Assumed: plumbing and heating firm with 22 employees and 1,100 maintenance customers.

The situation in figures

  • The plumbing, heating and air conditioning trade in Germany had around 48,000 businesses and around 390,000 employees in 202530
  • A calculated 10,941 skilled workers were missing in plumbing, heating and air conditioning in Germany in 202527
  • 48% of skilled trades businesses in Germany offer online appointment booking19

Built in the example

The request in the example

“On a winter weekend the duty phone rings thirty times. A handful are real emergencies, the rest could wait until Monday. My fitter still has to pick up every time.”

Example with model values

  1. Analysis
  2. Rules
  3. Pilot
  4. Operation
  1. Sat 21:47Customer

    The heating has been off for an hour, the display says fault. We have a baby in the house.

  2. Sat 21:47AI assistant

    This is the AI assistant of the emergency service. Is there a smell of gas or is water leaking? If not: please send a photo of the display, and how warm is it in the flat?

  3. Sat 21:50Customer

    No gas, no water. Photo attached, it is 18 degrees in the living room.

  4. Sat 21:51Triage

    Level 2 under the master's rule: heating failure in frost with an infant. Maintenance customer, gas boiler from 2016, last service in October. The case goes to the fitter on call.

  5. Sat 21:58On call

    The fitter calls back and has the system reset once, without success. He sets off, arrival around 22:45.

  6. Sat 21:59Confirmation

    The customer receives the arrival time and the call-out surcharge in writing before the fitter sets off.

  7. Mon 07:00Handover

    The office sees every case from the weekend: in the example six call-outs, 25 appointments for the week, two spare parts ordered.

What the review shows

  • In the example 31 calls come in on a winter weekend. Six need a fitter that night or the same day.
  • The fitter on call spends an assumed 3.5 hours per weekend on the phone, on top of the call-outs.
  • Four in ten callers reach nobody in the example and ring the next firm. Maintenance customers are among them.

What gets built

  • Intake outside business hours by phone and messenger. The AI asks for address, system and symptom and whether water is running or people are at risk.
  • Fixed rules set by the master sort into three levels: immediately, today with a call back from the fitter, next working day with a fixed appointment.
  • A smell of gas is not sorted: the message is to leave the house and call the grid operator's fault line. The fitter is alerted at the same time.
  • The fitter receives the case with photos and system data from the maintenance contract and decides whether to drive out.

Assumed result in the example

calls need a fitter straight away in the example. The others get a fixed appointment
6 of 31
target for on-call phone time per weekend, 3.5 hours before
1 hr
decides on every call-out
Fitter

Law and standards in view

  • AI Act Art. 50The assistant says at the start of every conversation that it is an AI.
  • Gas and dangerWith a smell of gas, water near electrics or suspected flue gas the AI classifies nothing. It refers straight to the fault line or the emergency number.
  • Call-out and priceThe fitter decides on every call-out. The customer learns the surcharge before agreeing.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

Being reachable does not mean a fitter answers every call himself. It means no real emergency waits and no maintenance customer rings into the void.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Solar system, battery, charging point: three packages, seven fixed steps, a 90 day test.

Assumed: electrical contractor with 16 employees and 40 solar installations a year.

The situation in figures

  • 65,851 electrical engineering businesses were registered in Germany on 30 June 202626
  • Around 6.4 million solar installations with 130,675 MW were in operation in Germany at the end of September 202639
  • A calculated 16,270 skilled workers were missing in building electrics in Germany in 202527

Launched in the example

The request in the example

“We cost every system from scratch, and in the end the customer wants a battery and a wallbox on top after all. Can we not sell that as one package?”

Example with model values

  1. Day 1 to 30Build the packages

    Three sizes, parts lists and package price from 40 costed installations

  2. Day 31 to 60Sell

    Package page, roof check by photo, quote within two working days

  3. Day 61 to 90Install and measure

    First installations, costing review per package, decision on expansion

Package orders since the start: plan and assumed run

Package orders
11Plan 10
Hours per quote
1.5Plan 2
Days from installation to commissioning
12Plan 14
  1. Day 30In the example the large package needs a new meter cabinet in every second house. It becomes a fixed extra with its own price.
  2. Day 60Every second package quote becomes an order in the example. Commercial roofs stay with individual planning.
  3. Day 90The costing review of the first six installations is 4% above plan on materials in the example. The package price is adjusted, the offer stays.

What the review shows

  • In the example a quote takes six hours because every system is planned individually. Every third quote becomes an order.
  • Eight in ten of last year's installations fit three sizes in the example. For half of them battery and charging point only came later as an add-on.
  • An assumed 24 days lie between installation and commissioning, mostly because of the meter appointment and missing documents for the grid operator.

What gets built

  • Three packages (small, medium, large) with a fixed scope and package price, battery and charging point included.
  • Seven dated steps: roof check by photo, site visit, quote, grid application, installation, meter appointment, commissioning with handover.
  • The customer sees the status of every step. Documents for the grid operator are drafted from the package data, the master checks and submits them.
  • Whatever does not fit the package (meter cabinet, scaffolding over two storeys, listed building) is named as an extra at the site visit.

Assumed result in the example

package orders in the model run after 90 days, plan 10
11
per quote in the model run, six before
1.5 hrs
from installation to commissioning in the model run, 24 before
12 days

Law and standards in view

  • Electrical work and grid connectionConnection and commissioning are done by the registered electrical firm. Notification to the grid operator and entry in the core energy market data register are fixed steps in the package.
  • Package pricePrivate customers see the total price and what is not included. Prices and extras are approved by the master.
  • Consumer contract and taxCancellation notice, contract template and VAT are clarified by the firm's lawyers and tax adviser.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

A package is a commitment about the process and not just a price. That is why the test first checks whether the firm keeps its seven steps.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Alongside 600 one-off jobs there are to be 60 inspection contracts. In the model that carries in two of three scenarios.

Assumed: electrical contractor with nine employees and 600 one-off jobs a year.

The situation in figures

  • 65,851 electrical engineering businesses were registered in Germany on 30 June 202626
  • A calculated 16,270 skilled workers were missing in building electrics in Germany in 202527
  • From 1 January 2027 companies in Germany with prior-year revenue above €800,000 issue e-invoices between businesses, from 1 January 2028 all others41

Launched in the example

The request in the example

“In summer we drown in work, in January I am on the phone chasing jobs. The property managers only ring when something fails.”

Example with model values

Cumulative cash balance of the inspection contracts € thousand

Carries in the model
  • Zero line: the money invested is back
  • Favourable 66
  • Expected 26
  • Unfavourable −28

What the review shows

  • In the example 85% of revenue comes from one-off jobs. January and February are 40% below the annual average.
  • The customer base holds an assumed 14 property managers and 35 commercial customers with recurring inspections: fixed installations, appliances, emergency lighting, charging points.
  • For every hour of testing there are 40 minutes of report writing and scheduling in the example.

What gets built

  • Three contract tiers with a fixed annual price per property: inspection, inspection with minor repairs, inspection with on-call cover.
  • Deadline calendar per property and installation. The system proposes dates for the quiet months first.
  • Readings come from the test device, the report is drafted. The qualified electrician assesses, signs and decides on defects.
  • Stop rule: below 20 contracts after nine months the build-up ends.

Assumed result in the example

scenarios reach the zero line within 18 months in the model
2 of 3
target after 18 months, threshold 20 after nine months
60 contracts
target share of revenue from contracts, 15% recurring before in the example
25%

Law and standards in view

  • Inspection of electrical installationsTesting and assessment are done by qualified electricians only. The report is signed by whoever took the readings.
  • Inspection intervalsThe intervals are set by the operator on the basis of its risk assessment. The calendar reminds, it does not decide.
  • E-invoicing (§ 14 UStG)Recurring invoices to companies are produced in the structured format. The changeover is accompanied by the firm's tax adviser.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

A maintenance contract does not sell an hour. It sells a date the customer no longer has to think about. Whether enough customers pay for that is what the test shows.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

After the storm 30 roofs are waiting. The drone measures in 20 minutes, the master assesses the damage.

Assumed: roofing firm with 18 employees and 260 repair quotes a year.

The situation in figures

  • 15,108 roofing businesses were registered in Germany on 30 June 202626
  • 10% of skilled trades businesses in Germany use drones19
  • In main construction in Germany order backlogs averaged 11.7 weeks in the third quarter of 202624

Launched in the example

The request in the example

“For every repair quote one of us climbs onto the roof. After a storm that is thirty roofs in a week, and half the quotes come to nothing.”

Example with model values

  1. Day 1 to 30Fly and compare

    Two employees with proof of competence, registration and insurance settled, ten roofs measured twice

  2. Day 31 to 60Quote from images

    Measurements and damage list as a draft, the master approves

  3. Day 61 to 90Review the costing

    Quote against final account for the first 15 jobs

Repair quotes per week with drone survey: plan and assumed run

Deviation from manual measurement
2%Plan 3%
Days to the quote
2Plan 3
Person hours per survey
1Plan 1.5
  1. Day 30Across ten roofs measured twice the area deviates by 2% at most in the example. The drone measures valleys under trees poorly, the tape measure stays there.
  2. Day 60Two of 20 quotes still need a look under the tiles in the example. The reservation stays standard.
  3. Day 90The costing review of 15 jobs is within range in the example. The firm adds gutters and chimneys.

What the review shows

  • In the example a survey by ladder ties up two people for three hours including travel.
  • Of 260 repair quotes a year an assumed 120 become jobs. For the other 140 someone still climbs onto the roof.
  • After a storm the customer waits eleven days for the quote in the example. The insurer asks for photos nobody has sorted.

What gets built

  • An employee with proof of competence flies over the roof in 20 minutes, with the owner's consent.
  • The images yield areas, lengths of ridge, eaves and valley and a list of conspicuous spots as a suggestion.
  • The master looks at every image, removes or adds damage and approves the quote. What the drone cannot see (battens, underlay, moisture in the insulation) is stated as a reservation in the quote.
  • The test covers repairs to pitched roofs of houses. Refurbishments and flat roofs stay with manual measurement.

Assumed result in the example

flight time per roof in the example, three hours for two people before
20 min
to the quote in the model run, eleven after a storm before
2 days
deviation from manual measurement in the model run, plan 3%
2%

Law and standards in view

  • Drone flightRegistration, proof of competence and liability insurance lie with the firm. Flights take place with the owner's consent, the pilot checks flight restrictions before every take-off.
  • GDPR Art. 6Neighbouring properties and people are not analysed. The images serve the job only.
  • Damage and safety at workThe master assesses damage and approves prices. Who climbs onto the roof and when remains governed by the firm's risk assessment.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The drone replaces the ladder and not the roofer. It fetches the measurements, the judgement on the damage stays with the master.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Three rooms, six photos, one floor plan: the quote goes out the same day, and Sunday stays free.

Assumed: painting and decorating firm with eleven employees and 420 quotes a year.

The situation in figures

  • 36,883 painting and varnishing businesses were registered in Germany on 30 June 202626
  • 68% of skilled trades businesses in Germany send quotes digitally, 62% invoices19
  • In the finishing trades in Germany order backlogs averaged 9.2 weeks in the third quarter of 202624

Built in the example

The request in the example

“I drive to every customer, measure with the laser, and write the quote three evenings later. By then half of them already have someone else.”

Example with model values

The master painter's week

  • Mon
    1 hrs free
  • Tue
    1 hrs free
  • Wed
    2 hrs free
  • Thu
    1 hrs free
  • Fri
    1 hrs free
  • Sat
    2 hrs free
  • Sun
    2 hrs free

10 hours a week returned in the example

  • Measurement

    Wall and ceiling areas from photos and floor plan, windows and doors deducted by the firm's rule. Unclear rooms go to a site visit.

  • Quote

    Draft from measurements, work descriptions and price list: substrate, coating system, masking. The master checks and approves.

  • Follow-up

    Three days after sending the firm follows up once. Rejections are noted with the reason.

What the review shows

  • In the example nine quotes are produced a week. Each costs 50 minutes for travel and measuring and 50 minutes at the desk.
  • An assumed five days lie between viewing and quote. 28% of quotes become jobs.
  • Six in ten enquiries in the example concern lived-in rooms with straight walls. Photos and a floor plan are enough for those.

What gets built

  • The customer uploads photos per room and a floor plan or two dimensions. The AI calculates the areas and asks when a wall is missing.
  • Period properties with stucco, water damage, facades and varnishing work go through a site visit as before.
  • The master checks every tenth photo measurement with the laser when the job starts. The deviation appears in the weekly report.

Assumed result in the example

a week returned to the master, assumed
10 hrs
target for the quote after the photos arrive, five days before
Same day
target for quotes won, 28% before in the example
38%

Law and standards in view

  • AI Act Art. 50Where the customer writes with an AI while uploading, the customer is told.
  • Quote or estimateWhat the customer receives is stated clearly on the document. Prices are checked and sent by the master.
  • GDPR Art. 6, 28Photos from homes serve the quote only and are deleted after a fixed period. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The job often goes to whoever sends a clean quote first. Measuring from photos does not make the master redundant, it lets him finish earlier.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€60,000 is budgeted for an own 3D bathroom planner. The review proposes twelve finished bathrooms with a package price.

Assumed: tiling firm with eight employees and 45 bathrooms a year.

The situation in figures

  • 47,174 tile, slab and mosaic laying businesses were registered in Germany on 30 June 202626
  • 12% of skilled trades businesses in Germany use 3D technologies19
  • 48% of skilled trades businesses in Germany offer online appointment booking19

Advised against in the example, with counterproposal

The request in the example

“The big players all have a 3D bathroom planner. I want one on my site, with my tiles, then people will come to me.”

Example with model values

  • Riskhigh

    Effort

    • A planner needs every tile, every fitting and every piece of furniture as a 3D model with a price. The range changes every year.
    • In the model upkeep costs €15,000 a year on top of the build.
  • Riskhigh

    Demand

    • In the example the 45 bathrooms a year come from recommendation and the local area. 3D bathroom planners are available free from manufacturers and retailers.
    • A self-planned bathroom is not yet a site measurement: the planner does not see wall framing, falls or waterproofing.
  • Riskraised

    Capital

    • In the example €60,000 equals the earnings from seven bathrooms, tied up before the first enquiry.

Do not build it like this

The counterproposal

Same goal: whoever is looking for a new bathroom sees it beforehand and ends up with the firm.

  • Twelve completed bathrooms as templates: photos, floor plan, materials list and package price for the stated room size.
  • The customer picks a template, one of two tiles and books the measuring appointment online.
  • Test over eight weeks with 15 booked measuring appointments as the threshold.
  • Own 3D planner, company budget
    €60,000
  • Test of the bathroom templates, company budget
    €6,000

Assumed result in the example

stays in the budget in the example until the templates bring measuring appointments
€54,000
booked measuring appointments in eight weeks as the threshold, assumed
15
as templates with a price instead of a planner for every conceivable bathroom
12 bathrooms

Law and standards in view

  • Package pricePrivate customers see the total price and the room size it applies to. The binding quote is written by the master after the site measurement.
  • Other tradesWater and electrics in the bathroom are connected by registered specialist firms. Every template names who carries out what.
  • PhotosOnly bathrooms whose owners have agreed are shown.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

The customer does not want to operate a tool. The customer wants to see what a bathroom like this costs and when someone comes to measure.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€250,000 for an own timber construction software to sell: in the model no scenario reaches the zero line.

Assumed: carpentry firm with 24 employees, its own timber cutting line and 35 timber buildings a year.

The situation in figures

  • 17,304 carpentry businesses were registered in Germany on 30 June 202626
  • 72% of skilled trades businesses in Germany say they have too much to do to deal with digitalisation. 4% use AI19
  • 69% of skilled trades businesses in Germany name high capital costs as an obstacle19

Advised against in the example

The request in the example

“I have built our costing and cutting lists in spreadsheets over years, nobody does it the way we do. We turn that into software and sell it to other carpentry firms.”

Example with model values

Cumulative cash balance of the own timber construction software € thousand

Advised against
  • Zero line: the money invested is back
  • Favourable −290
  • Expected −590
  • Unfavourable −770

Why we advise against it

  • Market: the target group is small and already served. For cutting and costing there is established industry software with links to the machines.
  • Cost: in the model the build costs €250,000, then €120,000 a year for upkeep, support and sales. At an assumed €150 a month per firm it would take around 70 paying firms just to cover the running costs.
  • Edge: the strength lies in the firm's empirical values and not in the program. Those are exactly the values it does not want to sell to competitors.

What happens instead

  • Decision: no software product. The €250,000 stays in the firm.
  • The empirical values per component move from the spreadsheets into a maintained database so that they do not depend on one person.
  • Handover to the existing industry software through its interface. In the example the budget for this is below a tenth.
  • Costing review after every build: planned and actual figures per trade flow back into the values.

Assumed result in the example

scenarios reach the zero line within 36 months in the model
0 of 3
shortfall in the middle scenario of the model after 36 months
€590,000
would have to pay in the model just to cover the running costs
70 firms

Law and standards in view

  • Structural stabilityStructural design and verification stay with the structural engineer. A costing does not replace them.
  • Software for third partiesLicence, liability and support would be a business of their own beside the carpentry firm. A law firm clarifies that before any start.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example

Knowledge that makes the firm strong belongs in the firm. Becoming a software house is a second profession and not a by-product.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Tuesday, 10:20: the architect changes the wall. By 16:00 the checked variation is with the client.

Assumed: building contractor for masonry and concrete work with 19 employees and five sites at a time.

The situation in figures

  • 35,297 bricklaying and concreting businesses were registered in Germany on 30 June 202626
  • 54% of skilled trades businesses in Germany name proof and documentation duties as the biggest burden (survey Q1 2023)43
  • In 2025 there were 104 insolvencies per 10,000 companies in construction in Germany, 69 on average across all sectors22

Built in the example

The request in the example

“On site the client says make the wall thicker. We do it, and three months later nobody wants to remember. I write the site diary on Sundays from memory.”

Example with model values

  1. Analysis
  2. Templates
  3. Pilot site
  4. All sites
  1. Tue 10:20Foreman

    Voice note with three photos: the architect was here. The internal wall on axis C is to be 24 instead of 17.5 centimetres thick, for sound insulation.

  2. Tue 10:21AI assistant

    Noted for the site diary. This deviates from item 3.12 in the bill of quantities. Is the instruction available in writing?

  3. Tue 10:23Foreman

    Verbal only.

  4. Tue 10:24Draft

    Draft variation created: 38 m² of masonry at 24 instead of 17.5 cm, extra quantity of blocks and mortar, three photos, verbal instruction by the architect on Tuesday at 10:15.

  5. Tue 16:05Approval

    The site manager enters the prices, adds one day of construction time and sends the notice to the client before the wall is built.

  6. Tue 17:30Site diary

    Daily entry from the foreman's evening note: weather, six men, crane, delivery of blocks, the architect's instruction with a reference to the variation.

  7. Wed 08:10Client

    Change on axis C confirmed. Please carry it out that way.

What the review shows

  • In the example the site manager writes up the site diary for five sites at the weekend, around four hours a week.
  • Of 60 changed or additional works a year an assumed 22 are never submitted as a variation. Value in the example: €48,000.
  • The evidence exists but is scattered: photos on three phones, agreements in the messenger, delivery notes in the site hut.

What gets built

  • In the evening the foreman speaks into the phone for two minutes and sends photos. This becomes the daily entry: weather, crew, plant, work done, deliveries, obstructions.
  • When the AI spots an instruction or a deviation from the bill of quantities, it drafts a variation: item, quantity, photo, who instructed what and when.
  • The site manager checks, enters prices and decides whether and when the notice goes out. Nothing leaves the firm without his approval.
  • Daywork goes to the client's site management for signature as a daywork sheet the same day.

Assumed result in the example

a week returned to the site manager, assumed
4 hrs
from instruction to notice in the example, often weeks before
6 hrs
variations a year are left lying so far in the example. Target: none
22

Law and standards in view

  • Building contractWhich notice is needed when depends on the contract. The firm agrees the templates with its lawyers, every notice is approved by the site manager.
  • Prices and construction timeThe AI suggests quantities. Prices, deadlines and the decision on the variation stay with the site manager.
  • GDPR Art. 6, employeesVoice notes and photos document the site and are not used to monitor staff. The workforce is informed.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

A variation often fails not on the entitlement but on the evidence. Whoever records the same day what was instructed has something in hand later.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

640 systems, each with its own deadline: the calendar lays out the routes, and the master no longer plans on Sunday.

Assumed: refrigeration contractor with twelve employees and 640 systems looked after for 210 customers.

The situation in figures

  • 3,576 refrigeration engineering businesses were registered in Germany on 30 June 202626
  • 54% of skilled trades businesses in Germany name proof and documentation duties as the biggest burden (survey Q1 2023)43
  • 17% of skilled trades businesses in Germany use predictive maintenance, 20% tracking systems19

Built in the example

The request in the example

“I have 640 systems in my head and in a spreadsheet. Which one is due for a leak check and when, I mostly know. But mostly is not enough when the butcher's cooling is down.”

Example with model values

The refrigeration master's week

  • Mon
    1 hrs free
  • Tue
    2 hrs free
  • Wed
    2 hrs free
  • Thu
    2 hrs free
  • Fri
    1 hrs free
  • Sat
    2 hrs free
  • Sun
    3 hrs free

13 hours a week returned in the example

  • Deadlines

    Every system is in the calendar with refrigerant, charge and inspection interval. Six weeks before the deadline a proposed date goes to the operator.

  • Routes

    Inspections and services due in one area fall on one day. The master slots in breakdowns himself.

  • Reports

    The engineer dictates findings and quantities at the system. The report is drafted, he checks and signs on site.

What the review shows

  • In the example the master plans deadlines and routes on Sunday, around three hours. He writes up reports on four evenings, six hours in total.
  • With 640 systems an assumed 900 inspections and services fall due a year. In the example 70 of them run past the deadline before anyone notices.
  • Scheduling calls and duplicate trips to the same town cost a further four hours a week in the example.

What gets built

  • Transfer of the 640 systems from the spreadsheet and maintenance contracts. The master confirms refrigerant and charge for every system.
  • Quantities topped up and recovered are tracked per system. The system flags conspicuous top-ups to the master.
  • The operator receives the report on the day of the inspection and an overview of its systems with the next deadline.

Assumed result in the example

a week returned to the master, assumed
13 hrs
missed deadlines as the target, 70 a year before in the example
0
target for the report reaching the operator, up to two weeks before
Same day

Law and standards in view

  • F-gas RegulationThe intervals depend on refrigerant and charge. The master sets them per system, the system only reminds.
  • Inspection and reportLeak checks are carried out by certified personnel. Findings and signature stay with the engineer, the duty to have the check done stays with the operator.
  • Record keepingReports and refrigerant quantities are filed in order per system. The firm sets the retention periods with its lawyers.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

A deadline in the master's head is a risk for the firm and for the customer. In the calendar it is a route.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Photograph the alcove, enter three dimensions, price range by the next evening. Only then does the master drive out to measure.

Assumed: joinery with nine employees and around 60 fitted wardrobes a year.

The situation in figures

  • 34,104 joinery businesses were registered in Germany on 30 June 202626
  • 68% of skilled trades businesses in Germany send quotes digitally, 48% offer online appointment booking19
  • Order backlogs in the German finishing trades averaged 9.2 weeks in the third quarter of 202624

Launched in the example

The request in the example

“For every enquiry I drive out and measure first. With two in three I never hear another word after the quote.”

Example with model values

  1. Day 1 to 30One product

    Fitted wardrobes only: three dimensions, photos of the alcove, fittings

  2. Day 31 to 60Price range

    Range from the firm's own costing, approved by the master for each enquiry

  3. Day 61 to 90Expand or stop

    Second product group only if the range matches the later quote

Complete enquiries for fitted wardrobes per week: plan and assumed run

Complete enquiries per week
5Plan 4
Site surveys that end in an order
57%Plan 50%
Time to the price range
20 hrsPlan 24 hrs
  1. Day 30For sloping ceilings three dimensions are not enough in the example. The knee wall height is added as a fourth.
  2. Day 60In three of 22 enquiries the quote falls outside the range in the example. The costing for special fronts is adjusted.
  3. Day 90The range holds in the example. Hall units and shelving are reviewed as a second group, kitchens stay out.

What the review shows

  • In the example 14 enquiries arrive a month. Each comes with a site visit of a good two hours including the drive.
  • Five of the 14 become orders. Nine visits, around 20 hours a month, end without one.
  • Fitted wardrobes make up 45% of enquiries in the example and can be pre-costed from width, height, depth and fittings.

What gets built

  • Enquiry page for fitted wardrobes: three dimensions, photos of the alcove, sloping ceiling yes or no, finish, interior.
  • The AI checks whether dimensions and photos fit together and asks for whatever is missing.
  • Price range from the joinery's own costing. The master sees every range and approves it.
  • Site survey only once the customer says yes to the range. The master then writes the quote.

Assumed result in the example

of site surveys end in an order in the model run, 36% before
57%
to the price range in the model run, planning value 24
20 hrs
at the start: fitted wardrobes, everything else runs as before
1 product

Law and standards in view

  • Price indication rulesThe range states prices including VAT and is labelled as a non-binding estimate. Only the quote after the survey is binding.
  • AI Act Art. 50Where the AI asks follow-up questions, the customer is told that an AI is writing.
  • GDPR Art. 6, 28Photos from the home serve this enquiry only and are deleted after a fixed period. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

The master still drives to every survey. Just no longer to the customers who drop out at the price range.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The balcony is welded in two days. The folder with certificates and declaration of performance costs three evenings.

Assumed: metalwork business with eleven employees, stairs, railings and balconies to EN 1090, around 180 jobs a year.

The situation in figures

  • 24,580 metalwork businesses were registered in Germany on 30 June 202626
  • Owners of small metalworking businesses in the Dresden chamber district in Germany spend 596 hours a year on average on 102 bureaucratic duties42
  • 54% of skilled trades businesses in Germany name proof and documentation duties as the biggest burden (survey Q1 2023)43

Built in the example

The request in the example

“Welding we can do. But when the checking engineer asks for the certificate for the beam, I spend the evening searching three folders and two inboxes.”

Example with model values

The owner's week

  • Mon
    2 hrs free
  • Tue
    1 hrs free
  • Wed
    2 hrs free
  • Thu
    1 hrs free
  • Fri
    2 hrs free
  • Sat
    2 hrs free
  • Sun
    1 hrs free

11 hours a week returned in the example

  • Material certificates

    Inspection certificates enter the job with the delivery note, the heat number by photo at goods-in. If a certificate is missing, it is on the list the same day.

  • Welding documents

    Welding procedure specifications and valid welder qualifications are assigned to the job. The system sends a reminder 60 days before a qualification expires. The welding coordinator checks and signs off.

  • Declaration of performance

    Draft from job data and the firm's template. Content, signature and CE marking remain with the owner.

What the review shows

  • In the example nine hours a week go on searching for, copying and chasing documents, mostly in the evening and on Saturday.
  • In every fourth job a material certificate is missing at acceptance. The final invoice then waits up to twelve days.
  • Certificates sit in emails, in the folder and with the steel merchant. Nobody sees what is missing before the customer asks.

Assumed result in the example

a week returned to the owner, assumed
11 hrs
missing certificates on acceptance day as the target, every fourth job before
0
target from acceptance to final invoice, up to twelve before
2 days

Law and standards in view

  • EN 1090, construction productsThe firm is responsible for the declaration of performance and CE marking. The system collects evidence and prefills drafts.
  • Welding coordinationWhether procedure, weld and test are right is judged by the welding coordinator. The AI assesses no weld.
  • Record keepingWhich document is kept for how long is settled by the firm with its certification body.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The duties do not shrink. But the evidence is created where the material arrives and the weld is laid, not at the desk in the evening.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

10:24 am: photo of the brake disc, reading, extra cost. At 10:31 am the customer approves, and the lift keeps working.

Assumed: independent car workshop with seven employees, four lifts and around 2,400 jobs a year.

The situation in figures

  • 60,070 motor vehicle workshops were registered in Germany on 30 June 202626
  • The German motor trade had revenue of about €210.8 billion in 2025, €36.6 billion of it in car servicing31
  • 48% of skilled trades businesses in Germany offer online appointment booking19

Built in the example

The request in the example

“When more turns up on the lift, I cannot reach the customer. Then the car sits up there until the afternoon, and the next one waits in the yard.”

Example with model values

  1. Analysis
  2. Prices and labour times
  3. Pilot
  4. Operation
  1. Mon 07:05Customer

    The front brake is squealing, and the roadworthiness test is due in March. When can I come in?

  2. Mon 07:05AI assistant

    This is the workshop's AI assistant. Wednesday 7:30 am is free. For the job I still need the registration number and mileage.

  3. Mon 07:09Estimate

    Job created. Check and replace front brake pads: €189 including VAT, from the workshop's labour times and parts prices.

  4. Wed 10:20Master

    Front discs below minimum thickness, photo and reading are on the job. The discs have to be done too, otherwise we do not hand the brake back like this.

  5. Wed 10:24Additional work

    To the customer: photo of the disc, reading 19.8 mm against a minimum of 20.0 mm, additional cost €214 including VAT, new collection time 3:30 pm.

  6. Wed 10:31Customer

    Approved. Both discs, please.

  7. Wed 15:10Approval

    The master checks the draft invoice with both approvals and releases the vehicle after the test drive.

What the review shows

  • In the example extra work turns up on the lift in every third job. Reaching the customer takes 2.5 hours on average.
  • For that long the lift stands idle or the car is moved off. In the model six lift hours a week are lost.
  • Every fifth addition is agreed on the phone and disputed at collection. In the example none of them can be evidenced.

What gets built

  • Appointment booking with vehicle data and the concern in free text. The AI assistant asks for whatever the job is missing.
  • Estimate from the workshop's labour times and parts prices. The mechanic makes the finding, the master sets the scope.
  • Extra work goes to the customer's phone with photo, reading and price. Nothing is extended without approval.
  • Every approval sits on the job with time and amount and appears on the invoice.

Assumed result in the example

to approval in the example case, 2.5 hours on average before
7 min
of lift time a week returned, assumed
6 hrs
of additions with documented approval as the target
100%

Law and standards in view

  • Estimate and additional workExtra work starts only after the customer's documented approval. Prices are shown including VAT.
  • Road safetyWhat is safety relevant is judged by the master. If the customer declines, the warning goes on the job and the invoice.
  • AI Act Art. 50, GDPR Art. 6, 28The assistant identifies itself as AI in every conversation. Keeper and vehicle data serve the job only.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI makes no diagnosis and sells nothing. It gets the master's finding to the customer while the car is still on the lift.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Eleven days from reception to the insurer's approval. The test with one type of claim aims for five.

Assumed: body and paint shop with 16 employees and around 900 accident repairs a year, two thirds of them through insurers.

The situation in figures

  • The German motor trade had revenue of about €210.8 billion in 2025, €36.6 billion of it in car servicing31
  • Order backlogs in the German motor trades averaged 3.0 weeks in the third quarter of 2026, capacity utilisation was 75%24
  • 62% of skilled trades businesses in Germany send invoices digitally, 20% use tracking systems19

Launched in the example

The request in the example

“Two people in the office spend all day on the phone chasing approvals. And the customer still rings three times asking where the car is.”

Example with model values

  1. Day 1 to 30One type of claim

    Own-damage claims without an assessor only, guided photos at one reception desk

  2. Day 31 to 60Send it complete

    File check before sending, status page for the customer

  3. Day 61 to 90Expand or stop

    Third-party claims only if insurer queries are below plan

Claims in the digital file per week: plan and assumed run

Days from reception to approval
4Plan 5
Cases with an insurer query
15%Plan 20%
Customer calls per case
1.1Plan 1.5
  1. Day 30In every fifth case the photo of the vehicle identification number is missing in the example. The photo guide now asks for it first.
  2. Day 60Customers open the status page four times per case in the example. Calls to the office go down.
  3. Day 90Queries are below plan in the example. Third-party claims are added as the second type.

What the review shows

  • In the example eleven days pass from reception to repair approval. For seven of them the file waits for missing photos, signatures or replies.
  • In four out of ten cases the insurer asks for more photos or line items because the first submission is incomplete.
  • Per claim the customer calls 3.2 times on average to ask how things stand.

What gets built

  • Guided photos at reception: the application asks for the views the costing needs and flags blurred or missing images at once.
  • Claim file with a completeness check before sending: photos, costing from the firm's estimating system, order, signatures.
  • Status page for the customer: reception, approval, parts, paint, collection. Every step with a date.
  • The costing is produced and owned by the master body builder. The office negotiates with the insurer.

Assumed result in the example

from reception to approval in the model run, eleven before
4 days
of cases with an insurer query in the model run, 40% before
15%
customer calls per case in the model run, 3.2 before
1.1

Law and standards in view

  • Legal questions on the claimNeither the firm nor the AI advises on liability, shares of fault or entitlements. The customer settles that with the insurer or a lawyer.
  • Repair methodRepair or replace is the master's decision. The AI checks whether the file is complete, not the damage.
  • GDPR Art. 6, 28Photos, registration numbers and claim data go only to the parties to the case. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Approval comes from the insurer, and no system changes that. But a file that is complete first time does not wait on itself.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€85,000 is budgeted for a nationwide glass shop. The review says: the jobs are within 20 kilometres of the workshop.

Assumed: glazier with eight employees, two vans and around 1,100 jobs a year.

The situation in figures

  • 3,535 glazing businesses were registered in Germany on 30 June 202626
  • 27% of skilled trades businesses in Germany offer online payment, 48% online appointment booking19
  • 69% of skilled trades businesses in Germany name the high cost of investing as an obstacle to digitalisation19

Advised against in the example, with counterproposal

The request in the example

“Everyone orders cut-to-size glass on the web these days. Cutting we can do, so I want my own shop and I will deliver nationwide.”

Example with model values

  • Riskhigh

    Shipping and breakage

    • Glass travels on racks or in crates by haulier. In the model shipping a shower screen costs more than cutting it.
    • In the model every twelfth consignment arrives broken or with edge damage. The glazier supplies the replacement first.
  • Riskhigh

    Competition

    • Large cutting plants sell online with a configurator and a price per square metre. There every cent decides.
    • A new shop only becomes visible with advertising. In the model a first-time buyer costs €38.
  • Riskraised

    Dimensions and complaints

    • The customer measures. If the pane does not fit, the argument starts over who pays for the mistake.
    • A shop cannot check which glass may be installed in which place.

Do not start like this

The counterproposal

Same goal: more predictable orders through the web.

  • Repair glazing within 20 kilometres: report the damage with photo and dimensions, reply with an appointment the same day.
  • Framework agreements with property managers: fixed prices for common panes, report per property, invoice per job.
  • Test over eight weeks with five property managers. Threshold: 30 jobs from their properties.
  • Planned glass shop, company budget
    €85,000
  • Local test, company budget
    €9,000

Assumed result in the example

stays in the budget in the example until the test shows results
€76,000
jobs from five property managers as the threshold, assumed
30
radius for the counterproposal
20 km

Law and standards in view

  • Distance selling and made to measureWhich rules on withdrawal and warranty apply to glass cut to the customer's dimensions is for the firm's solicitors to settle before any shop is built.
  • Glass type and locationWhether safety glass is needed is judged by the master glazier on site. The AI sorts reports and suggests appointments.
  • GDPR Art. 6, 28Tenant data from property managers' reports serve the job only. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

Online is won by whoever can ship cheaply. A glazier wins where someone still has to come today, measure and install.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The family sees the stone with name and lettering before it is cut. The application to the cemetery comes from the same design.

Assumed: stonemason's business with six employees and around 140 memorials a year in twelve cemeteries of the region.

The situation in figures

  • 4,494 stonemasonry and stone sculpting businesses were registered in Germany on 30 June 202626
  • 12% of skilled trades businesses in Germany use 3D technologies19
  • 35% of skilled trades businesses in Germany offer online consultation19

Launched in the example

The request in the example

“No family can picture the stone from a pencil sketch. And when the cemetery sends the application back, I am the one who has to break that to them as well.”

Example with model values

  1. Day 1 to 30Three cemeteries

    Upright headstones, design rules of three cemeteries on file

  2. Day 31 to 60Application from the design

    Drawing and application come from the preview, the master checks

  3. Day 61 to 90Expand or stop

    More cemeteries only if fewer applications come back than planned

Designs with preview image, cumulative: plan and assumed run

Meetings until the decision
2Plan 2
Applications sent back
6%Plan 10%
Weeks until the family decides
2.5Plan 3
  1. Day 30Polished stone looks lighter on screen than in the hand in the example. A hand sample now lies on the table with every preview.
  2. Day 60One set of regulations changes during the test in the example. The master now reviews the rules on file once a quarter.
  3. Day 90Of 34 applications two come back in the example. Flat grave slabs and four more cemeteries follow.

What the review shows

  • In the example a family needs three meetings before shape, stone and lettering are settled. Five weeks lie between the first and the third.
  • Every fourth application comes back from the cemetery: size, material or finish do not match the rules for that section.
  • Each return delays the setting of the stone by three to four weeks in the example. For the family that is the hardest news.

What gets built

  • Preview at the table: shape, stone, lettering and ornament in true proportion, in a photo of the grave if wished. The stonemason leads the conversation, not an assistant.
  • The cemeteries' design rules are on file for each section. The design shows at once what is not permitted there.
  • Application with drawing, dimensions, material and fixing drafted from the chosen stone. The master checks and submits.
  • Start with upright headstones in three cemeteries whose rules the firm has entered and checked itself.

Assumed result in the example

applications come back in the model run, every fourth before
2 of 34
until the family decides in the model run, five before
2.5 weeks
at the start instead of all twelve
3 cemeteries

Law and standards in view

  • Cemetery regulationsThe cemetery administration decides on the application. The rules on file are a preliminary check for which the master is responsible.
  • StabilityFoundation and fixing are dimensioned by the master stonemason under the technical rules. The preview calculates none of it.
  • Preview and dataEvery view is marked as a computer-generated draft. The family's details serve this order only (GDPR Art. 6).

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

In this trade speed is not the aim. The aim is that the family decides once and the cemetery agrees once.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

€140,000 for a bespoke 3D room planner with AI. The review says: the customer wants to touch the fabric, not rotate it.

Assumed: interior fitter with ten employees, one showroom and around 450 jobs a year.

The situation in figures

  • 19,916 interior fitting businesses were registered in Germany on 30 June 202626
  • 12% of skilled trades businesses in Germany use 3D technologies19
  • 69% of skilled trades businesses in Germany name the high cost of investing as an obstacle to digitalisation19

Advised against in the example, with counterproposal

The request in the example

“I want our own 3D planner with AI. The customer uploads the living room and straight away sees our curtains and floors in it.”

Example with model values

  • Riskhigh

    Build and upkeep

    • A 3D planner is a software product with a team of its own. In the model it costs a further €30,000 a year after the build.
    • Every supplier collection has to be entered with colour, drape and texture. In the example that is 3,000 fabrics a year.
  • Riskraised

    Benefit to the customer

    • The colour and feel of a fabric are not true on screen. In the model complaints about colour increase.
    • Room planners can be rented. One of your own brings no advantage the customer would notice.
  • Riskraised

    Usage

    • In the example 450 jobs come in a year. Even if every second customer used the planner, the volume would not carry the cost.

Do not build it like this

The counterproposal

Same goal: the customer decides at home and sooner.

  • Upload a photo of the room: the AI suggests five fabrics or floor coverings from the firm's own collection, the adviser selects and adds.
  • Sample pack sent home, target two working days. The customer sees and feels the samples in their own light.
  • Measuring appointment bookable online. Test over eight weeks, threshold: 40 sample packs and 15 measuring appointments.
  • Planned 3D planner, company budget
    €140,000
  • Test with sample pack, company budget
    €12,000

Assumed result in the example

stays in the budget in the example until the test shows results
€128,000
sample packs in eight weeks as the threshold, assumed
40
target from room photo to sample pack
2 working days

Law and standards in view

  • AI Act Art. 50Suggestions from the AI are marked as such. The adviser is responsible for the selection.
  • GDPR Art. 6, 28Photos from the home serve the consultation only and are deleted after a fixed period. Providers are bound by contract.
  • Measuring and quoteThe firm takes the measurements on site. The quote is written and sent by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example, with counterproposal

Software of your own pays off when it can do something that cannot be rented. Here the sample pack can do more than the planner.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Sick call at 5:10 am, cover arranged by 5:16 am. At six the surgery is unlocked by someone who knows the site.

Assumed: contract cleaning firm with 140 employees, many part-time, and 95 sites.

The situation in figures

  • 89,025 contract cleaning businesses were registered in Germany on 30 June 202626
  • Contract cleaning in Germany counted 1,005,213 employment relationships in 202435
  • 75% of skilled trades businesses in Germany report a shortage of skilled workers19

Built in the example

The request in the example

“The sick call comes at five, the surgery has to be clean by six. By then I have rung eight people, and seven are still asleep.”

Example with model values

  1. Analysis
  2. Site data
  3. Pilot
  4. Rollout
  1. Mon 05:10Cleaner

    I am ill today. I cannot do the surgery on the market square at six.

  2. Mon 05:10AI assistant

    This is the scheduling team's AI assistant. Get well soon. The absence is recorded, the site manager is arranging cover.

  3. Mon 05:11Suggestion

    Two people can cover: both inducted at the site, rest period observed, less than 20 minutes away. One still has six contract hours free this week.

  4. Mon 05:16Site manager

    The second colleague takes it, she has confirmed.

  5. Mon 05:17Assignment

    Shift reassigned. The site folder with the specification and the note on key handover goes to the cover, the hours go into the time record.

  6. Mon 07:42Quality

    Photo check: eleven of twelve points evidenced. The photo of the washroom is missing.

  7. Mon 07:50Site manager

    Photo supplied and checked. The surgery receives the daily record.

What the review shows

  • In the example 14 shifts drop out at short notice each week. Each costs the site manager around 35 minutes on the phone, mostly before six.
  • In every fifth case the cover does not know the site. Keys, alarm system and specification are new to them.
  • In the example checks happen on site once a month. Complaints arrive nine days after the defect on average.

What gets built

  • Absence by messenger or phone call. The AI assistant records it and never asks for the diagnosis.
  • Cover suggestion by induction, contract hours, rest period and travel. Who is deployed is decided by the site manager.
  • Photo check per site against the specification: surfaces, no people. The site manager sees missing points the same morning.
  • Pilot in 20 sites of one district, then rollout district by district.

Assumed result in the example

to arranged cover in the example case, around 35 before
6 min
less time on the phone a week for site managers, assumed
7 hrs
in the pilot instead of all 95
20 sites

Law and standards in view

  • Working timeSuggestions respect rest periods and contract hours. A human orders the assignment, working time is recorded.
  • GDPR Art. 9, employee dataThe absence is recorded, never the diagnosis. Photos show surfaces and are not used to assess individual employees.
  • Co-determinationWhere there is a works council, the system is agreed with it before the start. The framework is settled by the firm's solicitors.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The AI does not replace the site manager but the eight calls before six. Who knows the site and may still work is on a list, and a human decides.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Every seventh door stays shut. Residents rebook the visit themselves, the route recalculates, the deadlines stay with the master.

Assumed: authorised district chimney sweep with two journeymen and around 2,300 properties in the district.

The situation in figures

  • 7,854 chimney sweep businesses were registered in Germany on 30 June 202626
  • There were around 19.5 million fossil heating systems in Germany in 2025, 4.7 million of them oil37
  • 48% of skilled trades businesses in Germany offer online appointment booking19

Built in the example

The request in the example

“I put slips through letterboxes, and one in seven does not open the door. Then I drive out a second time, and in the evening I am still sitting over the sweeping register.”

Example with model values

The master's week

  • Mon
    2 hrs free
  • Tue
    2 hrs free
  • Wed
    1 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    2 hrs free
  • Sun
    1 hrs free

12 hours a week returned in the example

  • Notice

    Appointment by letter, email or text message, with the notice period the sweeping and inspection ordinance requires. Residents rebook themselves, but only into slots the master has released.

  • Route

    Sequence by street, confirmations and the deadlines in the fireplace notices. The proposal arrives the evening before, the master changes it and decides.

  • Records

    Readings and work are captured on site, certificates are produced as drafts. Defects are judged by the chimney sweep, entries in the sweeping register are released by him personally.

What the review shows

  • In the example the door stays shut at every seventh visit. That is around 25 second trips a month.
  • Rebookings come by phone, 40 minutes a day on average, often while working on the roof.
  • Certificates and the sweeping register get done in the evening and on Saturday, five hours a week in the example.

Assumed result in the example

a week returned to the master, assumed
12 hrs
visits to a closed door as the target, every seventh before
1 in 20
deadlines moved without the master's decision
0

Law and standards in view

  • Chimney Sweep Trade ActFireplace inspection, fireplace notice and sweeping register are sovereign duties of the authorised district chimney sweep. The system prepares, he decides and signs.
  • DeadlinesResidents can rebook only within the deadline set in the notice. Anything beyond that is seen and decided by the master.
  • GDPR Art. 6, 28Data from the sweeping register serve the statutory duties and not advertising. Providers are bound by contract.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The appointment can move, the deadline cannot. So residents choose, but only from what the district chimney sweep has released.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

A three-week wait in May, four people standing in the yard in January. Maintenance contracts are meant to even out the year.

Assumed: landscape gardening business with 18 employees, 70% of revenue from one-off jobs.

The situation in figures

  • Landscape gardening in Germany had revenue of about €11.11 billion in 2025 (not a skilled trade within the meaning of the German Crafts Code)36
  • 131,746 people worked in landscape gardening in Germany in 202536
  • Bureaucracy costs an SME in Germany 32 hours a month on average7

Launched in the example

The request in the example

“In May I do not know whom to put off first, and in January four people are standing in the yard. I want maintenance contracts that carry us through the year.”

Example with model values

Cumulative cash balance of the maintenance contracts € thousand

Carries in the model
  • Zero line: the money invested is back
  • Favourable 80
  • Expected 44
  • Unfavourable −10

What the review shows

  • In the example monthly revenue swings between €60,000 in January and €210,000 in May.
  • Of 320 private customers from the last three years, 110 ring at least twice a year for pruning and upkeep. Each time a new quote is written.
  • In the model a maintenance round pays from 14 gardens a day. Below that the travel eats the margin.

What gets built

  • Three maintenance packages with fixed services per season and a monthly price. The proposal comes from garden area and photos, the operations manager approves every quote.
  • Offered first to the 110 regular customers and to property managers along two existing rounds.
  • Maintenance report with photos after every visit. The foreman proposes extra work, the customer approves it.
  • Stop rule: below 40 contracts after six months the trial ends.

Assumed result in the example

scenarios reach the zero line within 18 months in the model
2 of 3
as the threshold after six months, assumed
40 contracts
regular customers with a recurring need, assumed
110

Law and standards in view

  • Fixed-term contractsTerm, notice and price adjustment in the contract terms are reviewed by the firm's solicitors, above all for private customers.
  • Price indication rulesMonthly prices for private customers are shown including VAT. Every quote is approved by a human.
  • Plant protection and tree inspectionPlant protection products are applied only by certified staff. Whether a tree is safe is judged by a qualified person and not by the AI.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

A maintenance contract does not sell more work but the same work in a plannable way. Whether it carries is decided by the density of the round, not the number of people interested.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

At half past one at night the bake list is written by feel. The test: 90 days of bake plan in three of nine branches.

Assumed: bakery with its own bakehouse, 9 branches and 85 employees.

The situation in figures

  • Bakery trade in Germany at the end of 2025: 8,659 businesses, around 35,000 branches, 232,000 staff, €18.14 billion revenue32
  • 81% of skilled trades businesses in Germany name energy costs as a challenge, 59% material costs19

Launched in the example

The request in the example

“At half past one at night I decide how much we bake. In the evening the driver collects every seventh roll again.”

Example with model values

  1. Day 1 to 30Shadow plan

    The plan runs alongside, baking stays as it is. Every evening it is compared with sales

  2. Day 31 to 60Three branches bake to plan

    20 main items, three similar branches run without the plan as a comparison

  3. Day 61 to 90Top-up baking and decision

    Prompt for the shop oven at 2 pm, then expand or stop

Unsold returns in the three test branches, as a percentage of goods delivered: plan and assumed run

Returns in the test branches
9.2%Plan 9%
Days a week with a main item sold out before 5 pm
2Plan 2
Suggestions accepted unchanged
64%Plan 70%
  1. Day 30In the example the shadow plan is closer to sales than the bake list on 19 of 26 baking days. The test begins.
  2. Day 60Returns fall in the example, but rolls run out too early on Saturdays. Saturday gets a supplement.
  3. Day 90In the example returns are 9.2% against a planning value of 9%. The comparison branches stand at 13.8%. Three more branches follow.

What the review shows

  • In the example 14% of goods delivered come back in the evening, 19% on rainy days.
  • The bake list is written at night from last year's figure and the bakehouse manager's experience. Weather and holidays are worked out in his head.
  • Returns vary between branches from 8 to 21%. The figures sit in the till system and are never compared.

What gets built

  • Bake plan per branch and item for the next day from till data, day of the week, weather forecast, school holidays and market days.
  • Prompt at 2 pm for the shop oven: what is still needed before closing.
  • Bakehouse manager and branch manager can change every line. Each change feeds into the next suggestion.
  • Three similar branches without the plan as a comparison, so that good weather is not counted as success.

Assumed result in the example

returns in the test branches in the model run, 14% before
9.2%
branches in the test, three more as a comparison
3 of 9
of shadow plan before any branch bakes differently
30 days

Law and standards in view

  • Employee dataWhat is analysed is item, branch and time of day. There are no analyses per sales assistant.
  • Baking quantityThe plan is a suggestion. The quantity is set by the bakehouse manager.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

A plan that runs alongside the bake list for a month has to prove itself first. Only then does it change what goes into the oven at night.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Thursday, 19:12: confirmation, 53 guests, one with coeliac disease. On Friday at 7:45 the quote goes out.

Assumed: butcher's shop with party catering, 11 employees and around 400 catering orders a year.

The situation in figures

  • 13,926 butchery businesses were registered in Germany on 30 June 202626
  • In the food trades in Germany order backlogs average 3.1 weeks (third quarter of 2026)24
  • 68% of skilled trades businesses in Germany send quotes digitally, 62% invoices19

Built in the example

The request in the example

“On Fridays the phone rings at the counter, and I jot down 60 people and suckling pig on a till receipt. Then on Saturday someone asks about the allergens.”

Example with model values

  1. Analysis
  2. Recipes
  3. Test weeks
  4. Operation
  1. Thu 19:12Customer

    Good evening, we are celebrating a confirmation on the Sunday three weeks from now. 45 adults, 8 children, hot buffet. One guest has coeliac disease, two do not eat meat.

  2. Thu 19:12AI assistant

    This is the butcher's AI assistant. That Sunday is still free in the calendar. Three questions: delivery or collection, what time, and should warming containers be included?

  3. Thu 19:15Customer

    Delivery at 11:30, with warming containers.

  4. Thu 19:16Draft

    Buffet proposal for 53 people: quantities by the business's portion rates, allergen list from the stored recipes, two dishes without meat, dishes without ingredients containing gluten marked.

  5. Fri 07:40Master

    Checks the allergen list against the recipes, takes out the breaded schnitzel, raises the crackling roast to 9 kg and approves quote and price.

  6. Fri 07:45AI assistant

    The quote with allergen list is on its way. Please note: flour is used in the kitchen, so the business cannot rule out traces of gluten. The master will ring about this today.

  7. Fri before the eventProduction

    Two days before the event: production sheet and shopping list created, delivery on the Sunday round at 11:30.

What the review shows

  • In the example 6 in 10 enquiries arrive after closing time or in the middle of counter trade on Friday and Saturday.
  • A quote takes 35 minutes: working out quantities, looking up allergens in the folder, typing it up.
  • In the example every fifth enquiry sits for more than two days.

What gets built

  • An AI assistant takes enquiries on the website and by messenger: occasion, number of guests, intolerances, delivery.
  • Quantity calculator using the business's own portion rates, for example 180 g of roast per adult.
  • Allergen list per buffet from the stored recipes. If a recipe changes, the list changes.
  • The master approves every quote. In cases of coeliac disease and severe allergies he rings the customer himself.

Assumed result in the example

target for checking a quote, 35 minutes of paperwork before
8 min
target until the quote, up to five before in the example
1 day
a week returned, assuming ten enquiries
4 hrs

Law and standards in view

  • Food labelling (allergens)The allergen list is drawn from the business's recipes. It is checked and approved by the master.
  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Price indicationPrices come from the business's price list, as final prices including VAT. The quote is approved by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The system does the sums and the typing. The master puts his name to the allergen list, which is why he approves it.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Saturday, 10 am, two chairs empty. The waiting list fills the gap before the owner reaches for the phone.

Assumed: hair salon with 7 employees and around 190 appointments a week.

The situation in figures

  • End of 2025 in Germany: 80,118 hair salons, around 222,000 people working in them, just under €7.7 billion revenue33
  • 48% of skilled trades businesses in Germany offer online appointment booking, 27% online payment19

Built in the example

The request in the example

“I am at the chair and the phone rings. And on Saturday two do not turn up, whose slots I could have filled all week.”

Example with model values

The owner's week

  • Mon
    2 hrs free
  • Tue
    1 hrs free
  • Wed
    1 hrs free
  • Thu
    1 hrs free
  • Fri
    1 hrs free
  • Sat
    1 hrs free
  • Sun
    1 hrs free

8 hours a week returned in the example

  • Booking

    Customers book for themselves around the clock, with the right duration for each service. Colour and first visits go to the salon with a query.

  • Reminder and cancellation

    Reminder 48 hours ahead, cancellation with one tap. The salon's no-show rule is stated in the confirmation.

  • Waiting list

    When a slot opens up, the offer goes down the waiting list in order. Whoever accepts has it.

What the review shows

  • In the example 40 calls a day, 25 of them about appointments. Each one interrupts a customer in the chair.
  • 9 of 190 appointments a week are missed without cancellation. At an assumed €55 per appointment that is around €500 a week.
  • The waiting list is a note by the till. By the time anyone rings round, the gap has passed.

Assumed result in the example

a week returned to the owner, assumed
8 hrs
target for no-shows a week, 9 before
4 of 190
of appointments booked online, target after three months
60%

Law and standards in view

  • GDPR Art. 6, 28Name, number and appointment are used only for the booking. The calendar provider is bound by contract.
  • Messages to customers (Unfair Competition Act § 7)The reminder belongs to the booked appointment. Offers from the waiting list go only to people who have signed up for it.
  • No-show ruleWhether and how a missed appointment is charged is decided by the owner. She settles the wording with her guild or law firm.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

Almost every second skilled trades business already takes bookings online. The difference lies in what happens to the cancelled appointment.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Glasses from the specialist, lenses from the internet. A subscription with check-ups carries in two of three scenarios in the model.

Assumed: optician with two shops, 9 employees and 6,500 customers on file.

The situation in figures

  • Optical trade in Germany: around 10,700 outlets, more than €7 billion revenue. The ten largest companies account for 55% of revenue34
  • 8,796 optician businesses were registered in Germany on 30 June 202626

Launched in the example

The request in the example

“They buy their glasses from me and their lenses online. And after they collect them, I do not see many of them for four years.”

Example with model values

Cumulative cash balance of the lens subscription € thousand

Carries in the model
  • Zero line: the money invested is back
  • Favourable 38
  • Expected 18
  • Unfavourable −25

What the review shows

  • Of 6,500 customers on file, 900 wear contact lenses in the example. 310 still buy them in the shop.
  • For 38% of customers the last eye test in the example was more than three years ago.
  • Per pack, online retail is 15 to 20% below the shop in the model. It does not offer fitting or check-ups.

What gets built

  • Lens subscription for a monthly amount: lenses, care products, one check-up a year, home delivery or collection.
  • Reminder to have eyesight checked, at the interval the master optician sets for each customer. Only with consent, with an opt-out in every message.
  • Start with the 310 lens customers in the shop, then the glasses customers who have given consent.
  • Stop rule: below 150 subscriptions after six months the trial ends.

Assumed result in the example

scenarios reach the zero line within 18 months in the model
2 of 3
as the threshold after six months, assumed
150 subscriptions
lens customers in the shop as the starting group, assumed
310

Law and standards in view

  • GDPR Art. 9Eyesight values are health data. They stay in the business's own trade software, and reminders contain no values.
  • Medical devices and fittingContact lenses are medical devices. Fitting, check-ups and the advice to see an eye doctor remain with the master optician.
  • Subscription contract with consumersTerm, cancellation and withdrawal are settled by the business's law firm before the start.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

A specialist shop does not win on price per pack. It wins on the fitting and on the appointment it remembers in good time.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The scan arrives at 17:48 without a tooth shade. The query goes out while the practice is still staffed.

Assumed: dental laboratory with 16 employees and 45 dental practices as customers.

The situation in figures

  • 6,610 dental technology businesses were registered in Germany on 30 June 202626
  • In the health trades in Germany order backlogs average 3.3 weeks (third quarter of 2026)24
  • 12% of skilled trades businesses in Germany use 3D technologies19

Built in the example

The request in the example

“The scan comes in the evening with no shade and no date. In the morning we chase the practice by phone, and they are in the middle of treatment.”

Example with model values

  1. Analysis
  2. Integration
  3. Pilot practices
  4. Rollout
  5. Operation
  1. Tue 17:48Practice

    Intraoral scan uploaded. Crown on 36, zirconia, try-in next week.

  2. Tue 17:48AI assistant

    This is the laboratory's AI assistant. The order is recorded. Tooth shade, try-in date and the scan of the opposing jaw are missing.

  3. Tue 17:52Practice

    Shade A3, try-in Tuesday 9 am. Opposing jaw is uploaded now.

  4. Tue 17:53Order

    Order created under an order number and planned backwards: design Wednesday, milling Thursday, delivery Monday on the courier round.

  5. Wed 07:30Dental technician

    Checks the scan: the preparation margin cannot be made out reliably on the distal side. Query to the practice with a marked image before design starts.

  6. Wed 08:20Practice

    The second scan from yesterday is attached. Please use that one.

  7. Mon 10:00Final inspection

    The master dental technician checks fit, contact points and shade on the model and releases the crown for delivery.

What the review shows

  • In the example 30 orders arrive a day, 9 of them incomplete: shade, date or opposing jaw are missing.
  • Each query costs two phone calls and half a day's delay on average.
  • Orders arrive by four routes: scan portals, email, courier slips, phone. The overview is on a whiteboard.

What gets built

  • One order intake for all four routes, with mandatory details for each type of work.
  • An AI assistant checks completeness and queries the practice in writing straight away. It does not judge the quality of a scan.
  • Backward scheduling from the practice's date, with a warning when time gets tight.
  • Start with six pilot practices, then rollout in groups.

Assumed result in the example

target for incomplete orders a day, 9 before
2 of 30
target until the first query, the next morning before
5 min
pilot practices in the first stage
6

Law and standards in view

  • GDPR Art. 9, 28Patient data is health data. The laboratory processes it only for the practice's order. Who is responsible for what is settled by contract between practice and laboratory. The assistant works with order numbers.
  • Medical devices (custom-made)A dental prosthesis is a custom-made device. Inspection and release remain with the master dental technician.
  • AI Act Art. 50The assistant identifies itself to the practice as AI.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

A query raised within five minutes in the evening does not cost half a day in the morning. What goes into the mouth is still checked by the technician.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

One hour in the workshop, three weeks waiting for the insurer: 11 of 60 cost estimates come back in the example.

Assumed: medical supply shop with orthopaedic workshop, 14 employees and 60 cost estimates a week.

The situation in figures

  • 2,014 orthopaedic technology businesses were registered in Germany on 30 June 202626
  • 54% of skilled trades businesses in Germany name proof and documentation duties as the biggest burden (survey Q1 2023)43
  • Bureaucracy costs an SME in Germany 32 hours a month on average7

Built in the example

The request in the example

“The insole is built in an hour. We wait three weeks for approval because one detail was missing on the prescription.”

Example with model values

The master orthopaedic technician's week

  • Mon
    2 hrs free
  • Tue
    1 hrs free
  • Wed
    1 hrs free
  • Thu
    2 hrs free
  • Fri
    2 hrs free
  • Sat
    1 hrs free
  • Sun
    0 hrs free

9 hours a week returned in the example

  • Documents

    Prescription, aid number, diagnosis, measurement sheet and photos are checked for completeness per insurer before the cost estimate goes out.

  • Status

    Submitted, queried, approved, refused: one list for all insurers. If no answer comes, the system sends a reminder.

  • Queries

    Draft reply to the insurer from the file. The master checks it, adds the justification and sends it.

What the review shows

  • In the example 11 of 60 cost estimates a week come back, mostly because details are missing on the prescription or the justification is missing.
  • A return delays the fitting by 12 days on average. The patient rings twice in that time.
  • The status of approvals sits in three insurer portals, in the inbox and in people's heads.

Assumed result in the example

a week returned to the master, assumed
9 hrs
target for returned estimates a week, 11 before
4 of 60
less waiting per avoided return, in the example
12 days

Law and standards in view

  • GDPR Art. 9, 28Diagnoses and prescriptions are health data. Only the staff involved in the fitting have access, and providers are bound by contract.
  • Fitting decisionWhich aid is suitable is decided by doctor and orthopaedic technician. The system checks completeness, not the fitting.
  • Cost estimatePrices follow the contracts with the insurers. Every cost estimate is approved by a human.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The insurer rarely sends an estimate back because the fitting is wrong. It sends it back because a field is missing. A system can check fields, not the fitting.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Three applications a year, eleven days to reply. The test: reply in 24 hours, trial day, 90 days.

Assumed: plumbing, heating and air conditioning business with 18 employees and two apprenticeship places a year.

The situation in figures

  • A calculated 10,941 skilled workers were missing in plumbing, heating and air conditioning in Germany in 202527
  • 16,500 apprenticeship places in the skilled trades in Germany were still open at the end of October 202628
  • 83% of skilled trades businesses in Germany report a shortage of apprentices19

Launched in the example

The request in the example

“Last year three applications came in. I replied to one after two weeks, and by then he had signed somewhere else.”

Example with model values

  1. Day 1 to 30The short route

    One page for the phone, enquiry in two minutes with no covering letter, reply within 24 hours

  2. Day 31 to 60Trial day

    Fixed trial days with a journeyman as mentor, feedback the same day

  3. Day 61 to 90Placement and contract

    A week's placement in the holidays, a talk with the parents, an offer from the master

Contacts from interested young people, cumulative: plan and assumed run

Contacts in 90 days
41Plan 36
Time to first reply
9 hrsPlan 24 hrs
Trial days
12Plan 10
  1. Day 3010 contacts in the example: 7 through two schools and the sports club, 1 through the job advert. The advertising budget moves to school contacts.
  2. Day 608 trial days in the example, 2 cancellations on the morning. The reminder the evening before now comes from the mentor himself.
  3. Day 90In the example 12 trial days, 5 placements, 2 signed contracts. The process stays in place.

What the review shows

  • In the example the first reply to an application takes 11 days. A covering letter, CV and school reports are required.
  • In the example the job advert costs the business €900 a year and brings three applications.
  • Two of the last three apprentices came through a placement. So far placements are not planned.

What gets built

  • Page for the phone with pictures from the business and an enquiry in two minutes: name, school, preferred date.
  • An AI assistant answers questions about training, pay and working hours and suggests a trial day. It assesses no one.
  • Process from contact to contract with deadlines: reply, trial day, feedback, placement, offer.
  • Who is invited and who is taken on is decided by the master.

Assumed result in the example

in the model run after 90 days, none the year before
2 contracts
to the first reply in the model run, 11 days before
9 hrs
trial days in the model run, planning value 10
12

Law and standards in view

  • General Equal Treatment Act (AGG)Only what is needed for the trial day is asked. There is no automatic pre-selection, and every acceptance and every rejection comes from a human.
  • AI Act Art. 50 and Annex IIIThe assistant identifies itself as AI. Selecting or assessing applicants would be a high-risk use. It does neither.
  • MinorsPlacement and contract only together with the parents. Working hours on the trial day follow youth employment protection law.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

Young people do not apply, they ask. Whoever replies the same day and offers a trial day is in the conversation while others are still waiting for the application folder.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

The master is 61 and the business lives in his head. 90 days, 24 processes, then two weeks without him in the office.

Assumed: joinery with 12 employees, €1.6 million revenue and an owner aged 61 with no successor.

The situation in figures

  • 34,104 joinery businesses were registered in Germany on 30 June 202626
  • Around 545,000 SMEs in Germany seek a successor by the end of 2029, about 109,000 a year29
  • 47% of skilled trades businesses in Germany name succession as a challenge19

Launched in the example

The request in the example

“I am 61. If I am out of action tomorrow, nobody knows how we cost a job or where the regular customers' measurements are. Nobody will buy it like that.”

Example with model values

  1. Day 1 to 30Stocktake

    What depends on the master: 24 processes, 40 regular customers, costing, suppliers

  2. Day 31 to 60Out of his head, onto paper

    The master talks, the AI writes the draft, a journeyman works from it and corrects it

  3. Day 61 to 90Figures and deputising

    Costing review per job, then two weeks of business without the master in the office

Processes written down and tried out: plan and assumed run

Processes written down
22Plan 24
Jobs with a costing review
86%Plan 80%
Days without a query to the master, out of 10
7Plan 10
  1. Day 30In the example 17 of 24 processes depend on the master alone. Costing is at the top of the list.
  2. Day 60In the example 16 processes are written down and tried out by journeymen. Three descriptions were wrong and are corrected.
  3. Day 90Two weeks without the master in the office: 7 of 10 days without a query in the example. The file for the tax adviser and the chamber is ready.

What the review shows

  • In the example costing, regular customers' measurements and supplier terms are in the master's head and in handwritten notebooks.
  • In the example 42% of revenue comes from customers who ring only the master.
  • Nobody knows exactly which type of job makes money: in the example every tenth job gets a costing review.

What gets built

  • Operations manual from conversations: the master talks for 20 minutes, the AI writes up the process, a journeyman tests it on the job.
  • Costing review per job from hours and material, with contribution margin per type of job.
  • Regular customers with measurements, drawings and contacts in one place. A second face from the business is introduced to them.
  • Handover file: processes, key figures and contracts at a glance.

Assumed result in the example

processes written down and tried out in the model run
22 of 24
of jobs with a costing review in the model run, every tenth before
86%
days without a query to the master in the test, planning value 10
7 of 10

Law and standards in view

  • Valuation and purchase contractCompany value, tax and contract are settled by the business's tax adviser and law firm. The project supplies the documents for it.
  • Master craftsman requirement (Crafts Code)Whoever takes over the business needs entry in the register of craftsmen. The chamber of skilled crafts advises on this.
  • GDPR Art. 6Interested buyers see customer and employee data only after a confidentiality agreement and to the extent the law firm releases.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Launched in the example

A successor takes on a business that runs without the seller. Every process that moves from the master's head onto paper is a piece of that.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Acceptance on Thursday at 15:20. At 16:41 the e-invoice is with the property manager.

Assumed: painting contractor with 22 employees and €1.9 million revenue, 70% of it with property managers and commercial customers.

The situation in figures

  • 36,883 painting and varnishing businesses were registered in Germany on 30 June 202626
  • From 1 January 2027 companies in Germany with prior-year revenue above €800,000 issue e-invoices between businesses, from 1 January 2028 all others41
  • 62% of skilled trades businesses in Germany send invoices digitally19

Built in the example

The request in the example

“Acceptance is on Thursday, and I write the invoice three weeks later on a Sunday. And from January it is all supposed to be a data set.”

Example with model values

  1. Analysis
  2. Integration
  3. Trial run
  4. Changeover
  5. Operation
  1. Thu 15:20Foreman

    Acceptance of stairwell, property 14: record signed, two photos, variation for 18 metres of skirting.

  2. Thu 15:21Draft

    Draft invoice from quote, job sheets and variation. The property manager's order number and property number are carried over.

  3. Thu 15:21AI assistant

    Two points to check: there is no written order attached for the variation. Two job sheets lack the customer's signature.

  4. Thu 16:40Master

    Attaches the property manager's message about the variation, checks quantities and prices and approves the invoice.

  5. Thu 16:41E-invoice

    Invoice generated as a structured data set to EN 16931, technically validated, sent and filed unalterably.

  6. Fri 09:05Property manager

    Invoice has been read in and is in the next payment run.

  7. Fri 09:30Incoming

    E-invoice from the paint wholesaler read, assigned to the property and put before the office for approval.

What the review shows

  • In the example 17 days pass between acceptance and invoice. At €160,000 of work a month that is around €90,000 waiting only to be written up.
  • The invoice goes out as a PDF by email. That is digital, but not an e-invoice in the sense of the new duty.
  • In the example incoming e-invoices from suppliers are printed out and typed in.

What gets built

  • Acceptance on the phone: record, photos, variations. The draft invoice is produced from it the same day.
  • Outgoing as an e-invoice to EN 16931 for business customers. Private customers still receive a readable invoice.
  • The AI checks for gaps: variation without an order, missing signature, deviation from the quote.
  • Trial run with three property managers before the deadline.

Assumed result in the example

target from acceptance to invoice, 17 before
1 day
invoiced earlier, in the model at €160,000 of work a month
€85,000
property managers in the trial run before the deadline
3

Law and standards in view

  • § 14 VAT Act (e-invoice)The format follows EN 16931. Which deadlines and exceptions apply to the business is settled by its tax adviser.
  • GoBD (record-keeping principles)The e-invoice is kept in its original form, unalterable and retrievable. The procedure is documented.
  • Invoice and variationQuantities, prices and variations are checked by the master. No invoice goes out without his approval.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

The duty asks for a format. The benefit lies next to it: whoever invoices on the day of acceptance makes the claim 16 days earlier in the example.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

A nationwide tradesperson portal for €250,000: none of the three scenarios reaches the zero line in the model.

Assumed: roofing business with 25 employees whose owner has set aside €250,000 for a second line of business.

The situation in figures

  • 15,108 roofing businesses were registered in Germany on 30 June 202626
  • 1,040,202 skilled trades businesses were registered in Germany on 30 June 202626
  • Order backlogs in Germany average 8.0 weeks across the skilled trades and 11.7 weeks in main construction (third quarter of 2026)24

Advised against in the example

The request in the example

“Every year I pay the portals for enquiries that are worth nothing. I can do that better: a portal by tradespeople for tradespeople, nationwide.”

Example with model values

Cumulative cash balance of the matching portal € thousand

Advised against
  • Zero line: the money invested is back
  • Favourable −640
  • Expected −1,230
  • Unfavourable −1,580

Why we advise against it

  • Two sides at once: without businesses no customers come, without customers no businesses. In the model winning an active business costs €400 and a matched job €60 in advertising.
  • The businesses are busy. Anyone with eight weeks of work in hand will hardly pay for more enquiries.
  • Nationwide visibility needs advertising in every region and every trade. In the middle scenario the €250,000 is used up after nine months.

What happens instead

  • A decision paper with three scenarios and all the assumptions behind them.
  • In the example the €250,000 stays in the business.
  • Instead, the business's own enquiry route with pre-screening: roof type, area, photos, postcode. That way the enquiries that arrive fit the business.
  • Whatever does not fit goes by arrangement to five partner businesses in the region, with no portal in between.

Assumed result in the example

scenarios reach the zero line within 36 months in the model
0 of 3
stays in the business in the example
€250,000
is how long the budget lasts in the middle scenario of the model
9 months

Law and standards in view

  • Platform dutiesA matching portal would fall under the EU rules for online intermediation services and platforms. The model counts the effort for this as cost.
  • Master craftsman requirement (Crafts Code)In the model the portal checks entry in the register of craftsmen for trades that require it. That too costs money per business.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Advised against in the example

The annoyance at poor enquiries is justified. It is a reason to build the business's own enquiry route, and no reason to take on the matching portals.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

In March 30 enquiries arrive every week. The master only drives to a site visit when garden and budget fit.

Assumed: pool builder with nine employees, 35 pools built a year and 210 systems under maintenance.

The situation in figures

  • In the third quarter of 2026 order backlogs across the skilled trades in Germany averaged 8.0 weeks24
  • 68% of skilled trades businesses in Germany send quotes digitally, 35% offer online consultation19
  • 72% of skilled trades businesses in Germany say they have too much to do to deal with digitalisation19

Built in the example

The request in the example

“From March to May everyone wants a pool. Three evenings a week I drive out to people who gulp at the figure, and the serious customers wait four weeks for their quote.”

Example with model values

  1. Analysis
  2. Questionnaire
  3. Spring pilot
  4. Maintenance contracts
  5. Operation
  1. Sat 20:14Customer

    We would like a pool in the garden, about 8 by 4 metres. Roughly what does it cost and when could you build?

  2. Sat 20:14AI assistant

    This is the firm's AI assistant. For a first range I need five details: type of pool, access for the excavator, slope, the heating you want and your budget frame. Please add three photos of the garden.

  3. Sat 20:31Customer

    Prefabricated pool, access 2.40 metres wide, slight slope, heat pump, budget up to €60,000. Photos attached.

  4. Sat 20:32First estimate

    Range from the firm's price list: €52,000 to €68,000 including VAT, plus yearly running costs for electricity, water care and maintenance. Note: the access is tight for the crane. Not a quote.

  5. Mon 07:40Master

    The master looks at the photos and details, rates the enquiry as a fit and approves the range. Site visit on Thursday.

  6. Mon 07:45AI assistant

    Your range and three suggested dates for the site visit. Please check with your building authority whether your pool needs a permit. We prepare the documents for it.

  7. Fri 12:00Report

    Week in the example: 31 enquiries, 12 a fit, 5 site visits. 19 enquiries received an honest answer without a drive.

What the review shows

  • In the example 22 of 100 site visits lead to an order. Six in ten fail on the budget that nobody asked about beforehand.
  • A site visit costs three hours including the drive. In the season that is 14 hours a week for the master.
  • Of 210 systems under maintenance, 80 have a contract. Winterising and start-up of the others are arranged by phone one by one every year.

What gets built

  • Guided enquiry with five details and photos. This produces a price range from the firm's price list including running costs.
  • The master sees every enquiry sorted and approves the range and the site visit. He assesses the technology and the structure on site.
  • Maintenance calendar: winterising, start-up and water checks are proposed for all systems, customers confirm the date themselves.
  • Pilot over one season, measured by site visits per order and days to the quote.

Assumed result in the example

target until the price range, four weeks to the quote before
2 days
fewer hours of driving a week in the season, assumed
9 hrs
maintenance contracts as the target, 80 before
150

Law and standards in view

  • AI Act Art. 50The assistant identifies itself as AI in every conversation.
  • Price indication and quotePrivate customers see prices including VAT. The range is labelled as a first estimate, the quote is written and checked by the master.
  • Building law and technologyWhether a pool needs a permit depends on the federal state and the size and lies with the owner and the building authority. Electrical work and water technology stay with the qualified staff.

The list names topics the company clarifies with its legal advisers. It is not a legal review by us.

  1. Pre-flight check
  2. Validate
  3. Launch
  4. Built in the example

Naming the figure early does not lose customers. It loses the drives to those who would never have become customers.

The stages show the sequence. Scope, price and schedule of such a venture are set out in the offer. Third party costs such as hardware and licences are additional.

Sources

The market figures above the examples come from official statistics and industry studies. Everything else in the examples is invented.

  1. bevh: Interactive retail in Germany, results for 2025, January 2026
  2. bevh: First half of 2026 in online retail, 6 July 2026
  3. HDE und IFH Köln: Online-Monitor 2026, 3 June 2026
  4. ECC Köln im IFH Köln, Intershop und CGI: B2B-Marktmonitor 2026, 5 October 2026
  5. IVH und ZZF: The German pet market 2025, 27 April 2026
  6. KfW Research: KfW Start-up Monitor 2026, 19 May 2026
  7. KfW Research: Bureaucracy in SMEs, KfW SME panel, 25 April 2025
  8. KfW Research: Fokus Volkswirtschaft no. 554: AI in SMEs, 29 July 2026
  9. Statistisches Bundesamt: Building permits for dwellings, January to August 2026, press release no. 370, 16 October 2026
  10. Statistisches Bundesamt: Completed dwellings and construction backlog 2025, press release no. 174, 22 May 2026
  11. Statistisches Bundesamt: Construction prices August 2026, press release no. 360, 9 October 2026
  12. Statistisches Bundesamt: House price index, second quarter of 2026, provisional, press release no. 336, 24 September 2026
  13. Statistisches Bundesamt: Consumer prices October 2026, press release no. 392, 11 November 2026
  14. Verband deutscher Pfandbriefbanken: vdp property price index, third quarter of 2026, 10 November 2026
  15. Deutsche Bundesbank: MFI interest rate statistics, housing loans to households, September 2026, as of 2 November 2026
  16. Deutsche Energie-Agentur: dena building report 2026, September update, September 2026
  17. KfW Research: KfW SME Panel 2026, October 2026
  18. Bitkom: Artificial intelligence in business 2026, companies with 20 or more employees, 14 September 2026
  19. Bitkom: Digitalisation in the skilled trades, 504 businesses, 28 August 2025
  20. Bitkom: Digitalisation of logistics, companies with 20 or more employees, 10 September 2026
  21. Statistisches Bundesamt: ICT survey 2025, reasons against using AI, as of 24 November 2025
  22. Statistisches Bundesamt: Corporate insolvencies 2025, press release no. 085, 13 March 2026
  23. Statistisches Bundesamt: Skilled trades: companies, employees, turnover, as of 23 April 2026
  24. Zentralverband des Deutschen Handwerks: Economic report third quarter 2026, 20,675 businesses, 4 November 2026
  25. Zentralverband des Deutschen Handwerks: Key figures of the skilled trades 2025, as of June 2026
  26. Zentralverband des Deutschen Handwerks: Registered skilled trades businesses by occupation, first half of 2026, 29 September 2026
  27. KOFA, Institut der deutschen Wirtschaft: KOFA Kompakt 7/2026: securing skilled workers in the skilled trades, June 2026
  28. Zentralverband des Deutschen Handwerks: Press release on the training market, 30 October 2026
  29. KfW Research: Succession monitoring of SMEs 2025, 9 January 2026
  30. Zentralverband Sanitär Heizung Klima: Plumbing, heating and air conditioning trade 2025, press release 12/2026, 4 May 2026
  31. Zentralverband Deutsches Kraftfahrzeuggewerbe: Sector data of the motor vehicle trade 2025, 15 June 2026
  32. Zentralverband des Deutschen Bäckerhandwerks: Facts and figures, as of 31 December 2025
  33. Zentralverband des Deutschen Friseurhandwerks: Economic situation of the hairdressing trade, press release 06/2026, 29 June 2026
  34. Zentralverband der Augenoptiker und Optometristen: Sector report 2025/26, 2026
  35. Bundesinnungsverband des Gebäudereiniger-Handwerks: New employment statistics, 9 July 2026
  36. Bundesverband Garten-, Landschafts- und Sportplatzbau: Landscaping sector data 2025, 2026
  37. Bundesverband des Schornsteinfegerhandwerks: Surveys of the chimney sweep trade 2025, 22 May 2026
  38. Bundesverband der Deutschen Heizungsindustrie: Heating market in the first half of 2026, 11 August 2026
  39. Bundesnetzagentur: Statistics on generation capacity of renewable energy sources, September 2026, as of 13 October 2026
  40. Bundesnetzagentur: Charging infrastructure in figures, as of 1 September 2026
  41. Bundesministerium der Finanzen: Questions and answers on e-invoicing, sections 14 and 27(38) of the VAT Act, as of 15 October 2026
  42. Handwerkskammer Dresden: Bureaucratic burden in the metalworking trades, study, April 2025
  43. Zentralverband des Deutschen Handwerks: Special survey on bureaucratic burden in the skilled trades, 10,630 businesses, July 2023
  44. Statistisches Bundesamt: Business register, reporting year 2024, as of 1 December 2025

Retrieved on 11 November 2026. bevh figures are gross values from a consumer survey, HDE figures are net values from the retailers' side. The two series are not comparable. We did not find search volumes and click prices for Germany in a verifiable form. That is why they are not shown here.

Which example is yours?

Which sector you come from makes no difference. Describe your venture in thirteen questions. You receive a report with a recommendation, free of charge and as a rule within 24 hours on working days. Even when the recommendation is no.

Have your venture checked for freeBook a conversationSee prices and timelines

All examples are invented and show how we work. Companies, names, figures and results are model values and not a promise for your venture. Offer for businesses only (§ 14 of the German Civil Code), not for consumers. Prices net plus 19% VAT. Timelines are planning values, the schedule in your offer is binding. We do not provide legal or tax advice.