questions we actually get

Every engagement gets a fixed-scope proposal with a fixed price — no hourly meter running while you wonder what the invoice will say. A thirty-minute call is enough for us to give you an honest range, and the proposal is written so you can take it straight to your budget owner.

AI-assisted development compresses the build phase dramatically: a working prototype typically exists within days, not months. Total time to production depends on scope, how many systems we integrate with, and the hardening pass — which we never skip. You see real builds weekly, so there’s no long silence followed by a reveal.

You do, outright. Everything we build lands in a repository your team controls, with documentation and tests. No licensing tail, no lock-in, no "platform fee" for your own software.

Claude and other frontier models, chosen per task. We build with commercial API access under terms that don’t use your data to train models. Where a smaller local model does the job better — cost, latency, data locality — we’ll say so and use that instead.

Access is scoped to the minimum that works, granted late, logged, and revoked at handover. Where possible we develop against sandboxes or synthetic data and only touch production data for the steps that genuinely need it.

Every build is designed to be handed over — documentation, tests, and a runbook your team can operate without us. If you’d rather we stay involved, ongoing support is scoped separately, on the same fixed-scope terms as everything else.

Then we tell you in the first call, for free. Some processes need a script, a form, or a policy change — not a model. Selling you AI where AI doesn’t belong would violate the one asset this business actually runs on.

Same hands, both jobs. verrou is the security consulting firm; rouage is its development practice. The practical consequence: every rouage build goes through the same security scrutiny verrou applies to client production systems — it’s also a standalone service.