Taking over software nobody masters any more, without stopping the business.
Your critical system is fifteen to thirty years old, its last maintainer is close to retirement, and nobody fully understands the code that keeps the company running. We map that system, extract its business logic in plain language and produce a costed modernisation plan — in weeks, not years.
Migration happens in stages, old and new running side by side, with equivalence tests on your real data before every cutover: no service interruption, no bet on the unknown.
This is not about mainframes. What actually causes trouble in Swiss SMEs is far more ordinary — which is precisely why nobody wants to touch it any more.
The first step is not writing code, it is understanding. We map what exists, extract the business rules in plain language and have you confirm them — often the first time they have been written down anywhere.
Then comes a costed plan, split into deliverable batches. The cutover happens module by module, old and new in parallel, with equivalence tests on your real data before each go-live. The business never stops, and every step stays reversible.
AI considerably speeds up reading code nobody knows any more, and the first conversion pass. It signs nothing. What ships has been read, tested and measured: we validate every decision, with you if you want to be involved.
The same approach serves the automation of your internal processes — documents, email, reporting — and the technical audit of a software estate during a succession or an acquisition.
No, and it is rarely a good idea. The takeover runs module by module: each delivered batch runs in production alongside the existing system, and rolling back stays possible until the cutover is confirmed.
That happens often. We then work from observed behaviour, the screens, the data and what your staff know. It takes longer; it is not out of reach.
Mapping and the costed plan take a few weeks, not months. Migration then depends on the number of modules: splitting it into batches lets you fund part of it and stop if your priorities change.
It keeps running. Until the equivalence tests are green on your real data, nothing is switched off.
One conversation is usually enough to say whether your case is workable, and roughly at what scale.