Today
The old host holds customers, items, orders and recipes from three decades, and often part of the daily business too. Fields are called KDNR, ARTNR or LFDAT, reports run through green terminal screens, and only two colleagues know where everything is. Every question from sales or controlling lands on their desks, and one of them retires next year.
An example
Fibrana introduced a new ERP in 2021; orders from 1994 to 2020 remain on the host. Sales asks which customers bought tarpaulin fabric in PES dtex 1100 with PU coating over the last ten years. demi connects both systems through an old item number conversion table, names 23 customers with quantities and last prices, each figure with file and field as its source, and shows 4 customers who have not ordered since the switch. The head of sales calls all four that same week.
Your data, wherever it lives
Customers, items, orders and recipes from three decades sit in the files of the old host, along with reports as spreadsheets and PDFs on the file share and the new transactions in today's ERP. demi reads, understands and links them, with your rights. The AI answers in minutes questions for which a colleague would otherwise write a query at the terminal, and you decide what follows from the answer.
Where the data lives today
- Legacy systems
- Databases
- ERP
- Office files
- PDFs
- Folders & drives
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 host system, today's ERP and reports on the file share. demi replaces nothing, reads the host files with field descriptions, keys and record types and links them with customers, items and orders in the ERP. A dictionary records what each field means and how it is converted.
- 2 · PermitPermission Layer
Everyone sees only what they may see in the host or the ERP today; prices and terms stay with sales and controlling. demi only reads and never writes back to the host. A specialist approves new dictionary entries, and every approval is recorded.
- 3 · ActIntelligence Layer
The agent proposes what each field means from its name, content, screens and program sources, for example LFDAT as the delivery date, and a specialist confirms it. It turns questions into queries across host and ERP; the figures come from the data, each with file and field as its source, and the language model only words the answer.
How it runs
- 1
demi reads the host files with field names, keys and content without touching daily operations.
- 2
The agent proposes a meaning for each field, for example KDNR as customer number and LFDAT as delivery date.
- 3
The colleagues who know the host confirm or correct the dictionary.
- 4
demi links host and ERP through customer, item and order numbers, even where number ranges have changed.
- 5
You ask in plain language and get the answer with a source for every figure; what follows is your call.
What changes
Questions to the old system without a terminal or a programmer
What the host experts know lives in the dictionary, not only in their heads
One view across old and new data, every figure with its source
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
+144 hfreed per year
Assumptions and calculation
demi answers questions across host and ERP with a source per figure, so nobody has to write a query.
45 min → 27 min per case
How we calculate
40 data requests × 45 min × 40% × 1.0 (manual) × 12 months = 8,640 min = 144 h
Value per year, before demi costs
Freed per year+144 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
+€9,360
gained per year
144 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
Does the host have to be changed for this?
No. demi only reads, through an export or a database connection, whichever works for you. Nothing changes in the host, its programs or its screens. We agree with you on how the data gets to demi.
How does demi know what KDNR or LFDAT means?
The agent proposes a meaning and shows its evidence: field descriptions, typical content, the screen where the field appears and programs that use it. LFDAT, for example, always lies after the order date and changes when a date is pushed back. A specialist confirms or corrects each entry before it counts.
How do old and new numbers fit together?
Customer or item numbers often changed in the switch. demi uses conversion tables where they exist and otherwise proposes matches by name, address or item features. The agent shows uncertain pairs for review instead of linking them silently.
Can demi handle packed numbers and old date formats?
Yes. demi converts packed decimals, old character sets and date fields with a century flag as it reads them, and the dictionary records how each field is converted. Quantities, amounts and dates then match what the terminal shows.
