Manufacturing and trade · When the numbers disagree
Two systems show two numbers. Both are wrong.
I find out why the numbers disagree, then build a report that holds up when someone checks it. Dynamics 365, production management systems, Microsoft Fabric, Power BI.
ExampleQuestion: how much did we sell in August?
ERP, finance
318 400 €
Too lowMisses the invoices still sitting in the production system. They reach the accounts next month.
Production system
329 750 €
Too highCredit notes added instead of subtracted, and one customer entered twice under two codes.
After reconciliation
324 180 €
The real oneBoth systems joined on a shared key, credit notes subtracted, duplicates merged. Every line traces back to a document.
01 Services
Four jobs I take on
Reports you can trust
A Power BI or Fabric report built straight on your ERP data, with a written definition of exactly what each number means. So nobody has to guess six months later.
Reconciling data across systems
For when the same customer, invoice or order lives in two systems. I find the key that joins them, catch the duplicates and write down which system owns which fact.
An AI assistant that knows your ERP
Not a generic chatbot. An agent loaded with your schema, what the fields mean and your business rules: you ask in plain language and get the answer together with an explanation of which data it was calculated from.
A diagnostic before the project
One week. I go through what you already have and tell you what is worth doing and what is not. Fixed price, quoted on the first call. The findings are yours, even if we go no further together.
Separate service · Sutampa
Still keying in supplier invoices by hand?
Each invoice is read, checked and posted to your accounting software, and a person approves only the exceptions. For companies that receive 100 or more invoices a month, and for accounting firms.
02 How I work
From the first call to something that works, no slide deck stage
I work alone. The person you talk to in the first call is the person who does the work. No salespeople, no handover to somebody else.
01
The call
45 minutes, free. What you are trying to find out, and why the answer you get today doesn't do the job.
02
Access to the data
Read only. I change nothing in your system and install nothing in it.
03
A prototype, not a presentation
A working report on your real data. We agree the deadline after the call. That is where it becomes clear whether this is worth continuing.
04
Handover
The code, the definitions and the documentation stay with you. Nothing ties you to me.
03 Fit
Get in touch if
- you have an ERP and data, but you don't trust it
- you asked three people the same question and got three answers
- an ERP migration is under way and you need to know what is really in the data
- somebody rebuilds the same report by hand in Excel every month
Don't get in touch if
- you need an ERP rolled out from scratch to dozens of users: that takes a team, not one person
- you need on-call support: I work project by project
- the decision is already made and you only need hands to carry it out
04 About me
More than ten years in IT and ERP with manufacturers, trading companies and online shops, from consultant to head of an IT team. In practice that means migrations between systems, reports that have to agree with the accounts, integrations between applications, and lately AI agents loaded with a company's data schema, so that the answer to a business question can be checked.
One rule: if a report cannot work something out, it has to say so, instead of showing a tidy number.
- Receivables reportA customer debt report that matches the accounting report line for line. Until then three people had three versions and argued every month about which one was right.
- Two systemsSales were in both the production system and the accounts, and the numbers differed. Afterwards there was one number, plus a list of which system is the source for which fact.
- AI assistantA question about ERP data meant waiting until someone had time to pull the data. Afterwards the team could ask in plain language and get the answer with an explanation of which data it was calculated from.
05 Contact
Tell me what you're trying to find out
No technical brief needed. A question that nobody answers the same way twice is enough.