We use AI to move faster. We do not use it as an excuse to stop reading the diff. Clients can tell the difference, even when they cannot name it.
AI-assisted development is on our services list because we actually use it: scaffolding, tests, first drafts of copy, the boring glue between two APIs we have written before. It is a multiplier for a senior engineer. It is a liability in the hands of a team that cannot tell when the model is confidently wrong.
What we let it touch
Boilerplate. Migrations we will still review. Test cases we will still run. Alt text and structured-data drafts that a human edits. Internal scripts. The rule is simple: if the output ships to production, a person who understands the product reads it. Our founder still treats generated code like a junior PR. Merge it when it is right. Rewrite it when it is almost right.
What we do not outsource to a model
Tenancy boundaries. Money. Auth. Anything with a lawyer attached. Architecture. The sentence that describes the product to a customer. AI is welcome in the room. It does not get the last word on the things that wake you up.
Faster, not faker
The point of the tool is more hours on the hard problems: the provisioning edge case, the payout that only happens when the booking completes, the admin empty state. If all you get from AI is a generic landing page and a CRUD app that looks like every other CRUD app, you have not hired a studio. You have rented a slot machine.
If you want a team that will use the new tools and still own the result, we are that house.