Today
Product master data lacks weights, materials, dimensions or performance curves, while the values sit in PDFs and old data sheets. Sales, logistics and catalogues all work around the gaps.
An example
For 2,300 gearbox variants at Talwerk Antriebstechnik, demi fills in the missing weight and shaft diameter from the data sheets. 140 values go to product management for review because two sources disagree.
Your data, wherever it lives
The values sit today in data sheets and catalogues as PDFs, in drawings and in scanned legacy documents on drives and in the DMS, just not in the product master in the ERP and PLM. demi reads, understands and links every document to its product's record, with your rights. The AI fills in thousands of attributes in days instead of months, each with the place it was found, and product management decides which value applies.
Where the data lives today
- PDFs
- CAD & drawings
- Scans & paper
- Folders & drives
- DMS & archive
- ERP
- PLM
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
The case needs data from the PLM, ERP, DMS and file share. demi replaces nothing and links data sheets, brochures, drawings and scanned legacy documents to each product's record. demi writes back only after approval.
- 2 · PermitPermission Layer
Product management reviews and approves proposed values before they reach the master data. Every accepted value is recorded with its source.
- 3 · ActIntelligence Layer
The agent extracts values and tables, converts units by fixed rules and proposes each attribute with its source page and a confidence. It flags conflicts between documents instead of picking a value.
How it runs
- 1
Clarify the data sources: file shares, the DMS and PLM
- 2
Detect missing attributes per product
- 3
Propose values with their source page and flag conflicts
- 4
Write reviewed values back to the master data
What changes
More complete master data
Better product search and catalogues
Legacy documents put to use
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
+256 hfreed per year
Assumptions and calculation
demi reads missing attributes from data sheets and drawings, each with its source; you approve them.
8 min → 4.8 min per case
How we calculate
400 master records × 8 min × 40% × 1.0 (manual) × 12 months = 15,360 min = 256 h
Value per year, before demi costs
Freed per year+256 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
+€16,640
gained per year
256 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 are conflicting sources handled?
demi takes no value over automatically. Both sources appear side by side and product management decides. The decision stays on the attribute together with its reason.
Can it extract curves and tables?
Yes. Tables are taken over row by row, curves as value pairs from tables or charts. Values read off a chart are marked, so it is clear they need checking.
Who approves the values?
The role you choose, usually product management. Until approval the values remain proposals and do not appear in the master record.
Does it work with scanned legacy documents?
Yes, scans are read too. With a poor scan the confidence drops, and demi flags the value for review instead of quietly taking it over.
