Use Cases
Four situations we meet again and again. They are deliberately generic and anonymised — they are here to show how this kind of project is shaped, not to impress with numbers.
If one of these sounds familiar, a conversation about it costs nothing.
01 An inventory system the vendor no longer supports
Situation
A system grown over fifteen years runs the warehouse, purchasing and dispatch. The vendor is ending support. Moving to a standard product would mean bending well-established processes to fit the software rather than the other way round — and giving up the special-case logic that actually makes a difference competitively.
Approach
Not all at once. First the business logic is recovered from the old system and locked down with automated tests — that is what later proves the new system does the same thing. Then it is replaced module by module, with both systems running side by side for a while.
Why this is feasible now
Reconstructing undocumented legacy logic used to be the most expensive part, and often the reason the project never started. That is exactly the part AI-assisted development has made significantly cheaper.
02 The spreadsheet nobody wants to touch
Situation
Quotes are calculated in a workbook of accumulated formulas and macros. Two people in sales maintain it. At some point both are on holiday, and one day both retire. Mistakes only surface once a quote is already with the customer.
Approach
The calculation logic moves into a small, testable application — with user roles, a traceable change history, and an interface that follows the way people already work. The workbook stays as a reference until both routes produce identical results.
Why this is feasible now
For a tool used by ten people, building something custom rarely added up. At today’s build costs it does.
03 A process that never justified the investment
Situation
Incoming supplier confirmations are checked by hand and typed into the system. A good hour a day, spread across several people. Too little to justify a large automation programme, too much to keep ignoring — and too specific for an off-the-shelf product.
Approach
One narrow, clearly bounded step: read the document, map the fields, flag the discrepancies, hand the doubtful cases to a person. Deliberately not a hundred per cent — the last few per cent of edge cases almost always cost more than they return.
Why this is feasible now
This size of task is precisely where the line has moved. Work that was too small to automate no longer is.
04 Knowledge sitting in the archive
Situation
Technical documentation, test reports and project correspondence spanning twenty years. The information exists, but search only works if you already know the right keyword and the right folder. In practice everyone asks the two colleagues who have been there longest.
Approach
The archive is made searchable with citations, so every answer leads back to the source document. It runs on your own infrastructure or in a European data centre — usually the deciding question where design files and contracts are involved.
Why this is feasible now
Making sense of unstructured archives was a research topic for a long time. It is now a craft — with the usual questions about data protection, rights and operating model, which we settle up front.
Does this sound like your situation?
Then let’s spend half an hour on it. You will get an honest assessment of whether the project is worth doing, and in what shape.