The business system NewData builds for itself.

Safe like ordinary software. Access based on roles and rights, as few rights per person as possible, encryption in transit and at rest, a virus scan on every uploaded file and an immutable audit trail. Sensitive data is only unlocked when there is a legal reason for it.

No lock-in. Every external provider, including the AI provider, sits behind its own internal layer. The core of the system depends nowhere on the response format of one party, and for every connection it is recorded how it can be left behind.

No burning tokens. AI is only used where it earns its place. Where ordinary programming logic is enough, no language model is called, and the system shows up front whether an AI run adds anything to what is already there.

Europe first. Data belongs within Europe. If a choice deviates from that, it happens only as a deliberate decision of the owner, with a dated substantiation in the decision register.

Deviations are logged. Every substantial decision is in a decision register with date and reason, and every deviation from the principles above is recorded there instead of silently tolerated.

The software follows NewData's process, not the other way round: that is the whole point of custom software, and the reason this system exists. See also custom AI solutions for how the same approach works at clients, and the client stories for what was built earlier.

Every change is traceable: who changed something, what the old and the new value was and why. Undoing is possible, but only with a real reason attached, and the undo itself is recorded too.

AI makes proposals in the system, no silent changes. Every proposal carries evidence and sources, and a far-reaching change requires a double confirmation from a person.

The project is honest about its own status: for every part it is recorded whether it is production-ready, a working prototype or a sketch. What is not finished is not called finished.