The software you already run

In most cases no, and the order is your decision anyway. Connecting gets you to a first result faster and is the usual entry point. Replacing pays off once maintenance and workarounds cost more than a rebuild, or when a vendor ends support. I build either, and which works out cheaper for you we calculate beforehand.

It can be, and it comes down to two figures. First the fee: what do you pay per year, and how often has that price risen? Second, access to your own data — with many providers an export, an interface or a higher query volume costs extra, and those items grow with your business. Where both apply and you use only a fraction of the feature set, building your own pays off faster than it did a few years ago, because AI-assisted development has noticeably reduced the effort. The side effect often outweighs the saving: your data sits with you, and the process follows your business rather than the other way round. We work it through on your real figures — and if the subscription turns out cheaper, you will hear that from me.

The undertaking is broken into stages, each estimated and quoted on its own. You only ever commission the next one, with a result that is useful by itself. Two items recur: operation, whose cost depends on user count, data volume and where it runs, and maintenance when one of your vendors changes something. Working out where it best runs is part of the advice: on your own hardware, with a German provider, or in the environment you already use. That decision affects both the running cost and who can reach your data — we work the options through together before you commit. I put figures on both once the shape is settled, and both appear in the quote rather than the small print. The initial conversation costs you nothing but time.

I work with test data wherever possible. Where real data is needed, it happens on your systems, with access you grant and can withdraw at any time. I sign a data processing agreement and a non-disclosure agreement before the first access. Your data stays with you, and none of it goes into AI tools.

In stages of a few weeks. Each replaces exactly one area — pricing, for instance — goes live on its own, and can be switched back to the previous state within a day if something does not fit. Undertakings that could otherwise drag on for years are deliberately cut so you could stop after any stage.

Yes, that is the condition I work under. Old and new parts run side by side, switching happens section by section, and every step can be undone. A cut-over date on which everything has to work at once is deliberately absent from this approach.

Yes, and it mainly changes the shape. Then it gets built deliberately simply: few components, little that needs regular maintenance, and nothing that would require you to hire a specialist. What your team needs to handle day to day we settle in advance rather than expecting it afterwards.

Every change automatically runs through the same checks before it reaches the day-to-day: does everything still work that worked before, are known security holes excluded, does the speed hold. That applies whether a person or an AI tool wrote the code — with AI assistance especially, many small changes accumulate, and these checks are why that does not turn into silent repair debt.

Everything produced is yours: the program code sits in your own account, not with me. Along with it come a written description of the procedures, automated tests any successor can use to check that everything still works, and instructions for setting the system up again from scratch. It is built with tools every systems house knows. On larger undertakings I work with your team or bring in established partners.