Today
New service technicians, dispatchers or production staff need months before they work confidently. Experienced colleagues answer the same questions again and again. How well someone is onboarded depends on who happens to have time.
An example
A new service technician starts at Talwerk Antriebstechnik. The agent builds him a 6-week plan for the G-400 and G-600 gearbox series, with 30 closed service cases to work through. It answers his question about the oil change interval from the maintenance instruction, revision D; a question about a custom conversion with no source goes to his mentor.
Your data, wherever it lives
What new employees need to know is scattered today: role and start date in the HR system, work instructions and manuals in the DMS and in folders, checklists on the intranet, solved cases in tickets and chats. demi reads, understands and links it to role, product and machine, with your rights. The AI puts the learning path together in minutes and answers questions with a source, and the mentor decides what goes into the plan.
Where the data lives today
- HR system
- DMS & archive
- Folders & drives
- Wiki & intranet
- Tickets
- Chats
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: the DMS, HR system, chat and file share. demi replaces nothing and links role profiles, work instructions, manuals, product structures and closed cases. So it is clear which knowledge a role needs and where it is kept.
- 2 · PermitPermission Layer
New employees see the material for their role, and personal data stays with HR. The mentor reviews and approves the learning path and sees progress, not the personnel file, and every step is recorded.
- 3 · ActIntelligence Layer
The agent builds a learning path per role and turns real cases into exercises. It answers questions only from the released instruction or the matching case, with its revision, and when it finds no source it passes the question to the mentor.
How it runs
- 1
Take over role and start date from the HR system
- 2
Build a learning path from instructions, manuals and real cases
- 3
The mentor reviews and adjusts the plan
- 4
New employees ask in the chat at any time and get answers with sources
- 5
Progress and open questions go to the mentor
What changes
Confident in the job sooner
Less load on experts
Consistent onboarding
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+154 hfreed per year
Assumptions and calculation
demi builds the plan per role and answers questions from real cases; you mentor the newcomers.
16 h → 12.8 h per case
How we calculate
4 new hires × 16 h × 20% × 1.0 (manual) × 12 months = 154 h
Value per year, before demi costs
Freed per year+154 h
Assumption: labor cost of $48.62 per hour in US manufacturing in June 2026 (BLS), plus workplace, IT and overhead.
= Value per year, before demi costs
+$10,010
gained per year
154 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
Can demi create plans per role?
Yes. The agent takes the role profile, for example service technician or dispatcher, and finds the instructions, products and typical cases that go with it. The result is a week-by-week plan that the mentor adjusts and approves.
Does onboarding use our real cases?
Yes. Closed cases from service, production or dispatch become exercises. Customer data the exercise does not need is hidden. New people learn from what actually happens in your company.
Can mentors follow progress?
Yes. They see which sections are done, which exercises are solved and which questions are open. It is a tool for onboarding, not for performance monitoring. You decide who sees what.
What if an instruction is out of date?
The agent only answers from released versions and names the revision. If there is no source or two documents disagree, the question goes to the mentor. Frequent open questions show where instructions are missing.
