A team that maintains what it builds

We’ve been right inside a lot of companies. We’ve sat with people who type in invoices one by one and with people who need a figure out in the field, with no computer in front of them. What strikes us most is that these people know very well which part of their work is unnecessary, and they almost never have the time to get rid of it.

That’s where we learned to tell what works from what only looks good in a demo, and what’s worth automating from what isn’t. Sometimes the answer is an AI agent and sometimes a simple automation with no model at all. When doing nothing is the better deal, we tell you before we charge for it.

Before connecting any software, we test the AI on your real cases. If it doesn’t get enough right, the project stops there. If it does, we start with a pilot, which is paid and whose price counts toward the final project.

The split is always the same: AI interprets and code executes. The model understands the request and the code decides which action runs.

No change reaches production without passing the test suite, not even a new version of the model. Before applying it we back up the database, and if the service then stops responding, it rolls back to the previous version on its own. On top of that, every week a batch of test conversations runs through every system we maintain.

The repository and the cloud where the system runs are in your name from day one, and model usage is paid through your own accounts. The monthly fee covers only monitoring and maintenance, has no lock-in, and if you cancel it, what was built keeps working.

We don’t audit or certify that your company complies with the GDPR or the EU AI Act, because that’s your lawyers’ job. We build the system so that compliance can be demonstrated.