Today
Customers write requirements into specifications that look fine on paper and cannot be held in production: a tolerance tighter than the part's own thermal expansion, a wall the material will not fill, a cycle time none of your machines can reach. Today an experienced engineer checks this in their head, when there is time. Often it shows only after the order, and then it costs the margin.
An example
Norvelta receives a request for an 8-cavity mould for a POM gear carrier: 150 mm long, tolerance ± 0.02 mm on the length. The agent shows that a difference of just 2 °C in the measuring room uses up almost the whole tolerance, and that the company's own measurement reports for earlier POM parts show a spread of ± 0.06 mm over this length. Its verdict: not feasible as specified, feasible with ± 0.1 mm on the length and ± 0.02 mm only on the two bearing seats, measured at 23 °C. The head of engineering sends the query before any quote goes out.
Your data, wherever it lives
Today the knowledge of what is feasible sits in many places: requests and specifications in e-mails and PDFs, material data and product limits in the PLM, measurement reports and capability figures in the QM system, the MES and spreadsheets, the experience in the heads of two or three people. demi reads and understands all of it and links it per requirement, with your access rights. The AI then checks a request in minutes where someone now spends half a day searching for documents, and your engineer decides.
Where the data lives today
- E-mails
- PDFs
- PLM
- ERP
- MES
- QM system
- Office files
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 mailbox, PLM, ERP, MES, QM system and your spreadsheets. demi replaces nothing and links request and drawing to material data, the limits of your products and the measured capability of your machines from past jobs. So every requirement sits next to what your production has proven it can do.
- 2 · PermitPermission Layer
Process data and capability figures are visible to engineering, production planning and quality. The answer to the customer is released by the head of engineering, and nothing leaves the house before; every release is recorded.
- 3 · ActIntelligence Layer
The agent extracts every measurable requirement from request and drawing and checks it against physics, material limits and your capability figures: thermal expansion, shrinkage, flow length, deflection, cycle time. It says feasible, feasible with conditions or not feasible and shows the calculation, every number with its source. It computes the formulas exactly; the language model only explains the result.
How it runs
- 1
The request arrives by e-mail with a drawing or specification.
- 2
The agent extracts every measurable requirement: dimensions, tolerances, material, quantity, cycle time.
- 3
It checks each requirement against physics, material data and the measured capability of your machines.
- 4
You get feasible, feasible with conditions or not feasible per requirement, with the calculation.
- 5
The head of engineering reviews and releases the query to the customer or the quote.
What changes
Impossible requirements surface before the quote
Customer queries backed by a traceable calculation
Fewer orders that later cost the margin
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
+162 hfreed per year
Assumptions and calculation
demi checks each requirement against physics and your process capability; you decide and reply.
90 min → 63 min per case
How we calculate
30 requests × 90 min × 30% × 1.0 (manual) × 12 months = 9,720 min = 162 h
Value per year, before demi costs
Freed per year+162 h
Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.
= Value per year, before demi costs
+€10,530
gained per year
162 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
How does demi know the limits of my machines?
From your own data: measurement reports, first articles, capability studies and process data from past jobs. The agent uses what your machines have proven to achieve, not the brochure values. If data on a process or material is missing, it says so.
Which physics does the agent calculate?
Whatever matters for the requirement: thermal expansion, shrinkage, deflection, flow length, forces and cycle times. Formulas and material values are fixed and traceable, and every result shows its calculation. An estimate is labelled as an estimate.
What does feasible with conditions mean?
That the requirement works under certain assumptions, such as a fixed measuring temperature, a different material or a tight tolerance only on functional dimensions. The agent names every condition and turns them into a query to the customer. It goes out only after release.
Does this replace my experienced engineer?
No. The agent does the calculation and gathers the evidence; your engineer makes the decision. They no longer search for old measurement reports and material values and only look at the points that stand out.
