AI-first setup

We start with the business, not the code: we sit with the people who do the work and map the product, the processes and the flows as they actually run — where work waits, what depends on one person, what’s routine and what genuinely needs a call. That map decides what agents do alone and where a person stays in the loop.

Then we set it up AI-first on your own machines: the agents’ instructions for how the product is built and how each process is meant to run, the tests that prove a change is safe, and the pipeline that takes a passing change to production. Linked to GitHub, so every change is a commit you can see, compare and undo.

Then we run real work through it with you until it feels ordinary.

Your product and your processes, on your own machines, ready to take instructions — with the agents, their instructions, the tests and the deploy pipeline all in your own repositories.

The line between “just do it” and “ask me first”, agreed with you and written into the project. A short guide: how to ask, what runs on its own, what comes back as a question, how to undo anything.

Thirty days of support while you get used to it.

A computer to run the agents on — yours or a team member’s — and an account with an AI provider, which you pay for directly. We tell you what to expect that to cost before we start.

Access to the code, the hosting and the tools your processes run through. If someone else built the product and holds the keys, we help you get them.

Two to four weeks depending on the business: a marketing site and a couple of processes is quick; an app with payments and integrations takes longer because we test more before handing it over.