Today
The same screw exists under several numbers, the same supplier under several names. Wrong master data leads to wrong orders, wrong stock and wrong reports.
An example
At Talwerk Antriebstechnik demi finds articles 10-2231, 10-2231A and 77-0045: the same DIN 912 M8 × 30 socket head screw three times, with stock in two plants. The agent proposes 10-2231 as the surviving record and shows which BOMs point to the other numbers.
Your data, wherever it lives
Part masters sit today in the ERP and PLM, often per plant and with old numbers from earlier systems, customers and suppliers also in the CRM, routings in the MES, plus part lists in spreadsheets. demi reads, understands and links them into one record per real thing, with your rights. The AI finds duplicates across the whole item master in hours instead of months of manual work, and the master-data owner decides which record leads.
Where the data lives today
- ERP
- PLM
- CRM
- MES
- Legacy systems
- Office files
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Your data sits in the ERP, PLM, CRM and MES. demi replaces nothing, normalizes units, names and attributes and joins everything into one record per real thing. demi writes back only after approval.
- 2 · PermitPermission Layer
The master-data owner approves every merge. Each change is recorded and can be undone.
- 3 · ActIntelligence Layer
The agent finds duplicates with fixed rules and similarity, explains attribute by attribute why records belong together and proposes the surviving record. Every proposal shows which system each value comes from.
How it runs
- 1
Unify master data from your own data
- 2
Review duplicate proposals with their evidence
- 3
Choose the surviving record and approve
- 4
Write the cleanup back after approval
What changes
Clean, unambiguous master data
Correct stock and purchasing reports
A reliable base for every further use case
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
+160 hfreed per year
Assumptions and calculation
demi finds duplicate items, suppliers and customers and proposes the merge; you approve.
10 min → 6 min per case
How we calculate
200 duplicates × 10 min × 40% × 1.0 (manual) × 12 months = 9,600 min = 160 h
Value per year, before demi costs
Freed per year+160 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
+€10,400
gained per year
160 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 sure is a proposed duplicate?
Every proposal shows its evidence: which attributes match, which differ and where the values come from. Clear cases with the same standard, size and material come first. Doubtful cases are marked and wait for a decision.
Can a merge be undone?
Yes. demi stores the state before every merge. Undoing it restores the old records and references, and the undo is recorded too.
Does it work across different ERP systems?
Yes. That is where most duplicates come from, for example after an acquisition or across several plants. How the data from the individual systems gets to demi, we work out in a call. After that, demi compares across all of them.
What happens to orders on the old number?
Before approval, demi shows where the old number is still in use. You decide whether it is blocked, remapped or kept as a reference. Nothing is switched over without approval.
