Today
Every night dozens of handovers run between your systems: the ERP sends prices to the customer portal and orders to the warehouse, customers send delivery schedules by EDI, the bank delivers statements, the carrier gets the shipping notices. A job that runs reports success, even when the file is empty, half written or duplicated. The error only shows when the warehouse works on yesterday's orders, a customer sees zero prices or a customer's line stops.
An example
At Nordhanse the agent reports three deviations at 5:30 am. The container movement file from the terminal has 212 rows instead of the usual 1,400 or so, because the transfer broke off at 2:10 am. Of 412 released orders, 386 reached the warehouse management system; 26 orders of a retail customer are missing because the item weight is empty in the master data. And the rate file for the customer portal holds 14 lanes priced at zero. The dispatcher has the terminal resend the container file before the first trucks arrive at 6 am.
Your data, wherever it lives
What flows between your systems at night sits in transfer folders and on the SFTP server, in interface tables, in the ERP and the warehouse system today, and whether it worked is written in job mails hardly anyone reads. demi reads, understands and links these handovers with customers, orders and items, with your rights. Every morning the AI checks in minutes whether everything arrived complete and plausible, and the owner decides what gets restarted.
Where the data lives today
- Folders & drives
- Databases
- ERP
- E-mails
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: transfer folders and SFTP directories, interface tables in the database, the ERP, the warehouse management system and the mails that jobs and EDI converters send. demi replaces nothing and keeps a list of every expected handover: what, from where to where, by when, how many rows it usually has and which fields must be filled. Each handover is linked to the customers, orders and items it concerns.
- 2 · PermitPermission Layer
IT sees all data flows, departments see those that touch their data: sales the price files, logistics the warehouse orders. Content with prices or personal data is visible only to those who may see it in the source system. By default demi only reads; restarting a job or resending a file is up to the owner, and every alert is logged.
- 3 · ActIntelligence Layer
The agent checks every handover against fixed rules: on time, row count in line with comparable days, mandatory fields filled, no zero prices, no duplicates, totals matching the source system. When something is off, it names the affected customers and orders and the likely cause, such as an export that started before the posting run. demi computes the checks exactly; the language model only writes the alert.
How it runs
- 1
You name the important handovers once; demi adds the rest from transfer folders, logs and job mails.
- 2
Every night demi follows what flows between ERP, warehouse, portal, bank and partners.
- 3
At half past five the agent checks every handover for arrival, completeness and plausibility.
- 4
When something is off, the owner gets an alert with the affected customers and orders and the likely cause.
- 5
The owner restarts the job or corrects the data; demi checks again and closes the alert.
What changes
Interface errors surface before customers or the warehouse notice
Every handover has an owner and a status
Less digging through logs and error mails
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
+108 hfreed per year
Assumptions and calculation
demi spots missing or wrong handovers in the morning; you fix them before any damage is done.
2 h → 1.4 h per case
How we calculate
15 data flow failures × 2 h × 30% × 1.0 (manual) × 12 months = 108 h
Value per year, before demi costs
Freed per year+108 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
+€7,020
gained per year
108 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
Which data flows can demi watch?
Anything that leaves a trace: files in transfer folders and on the SFTP server, EDI messages, exports and imports between ERP, warehouse, portal and bank, database tables and the mails that jobs send. You decide which handovers matter; demi suggests the rest from the logs.
Does this replace our job monitoring or EDI converter?
No, both stay. Monitoring tells you whether a job ran. demi checks whether what arrived is right: row count, mandatory fields, values compared with previous days and with the source system.
How does demi know what is plausible for a handover?
From rules you set, for example that a price is never zero or a delivery date is not in the past. On top comes the comparison with previous days: the agent learns from history how big a file usually is on a Monday. You set the thresholds, and every alert names the rule that fired.
Does demi restart a failed job by itself?
Not by default: demi reads and reports. For single internal handovers you can decide once that the agent triggers the transfer again, for example when a file was only picked up too early. Anything going to customers, suppliers or banks is always released by a person.
