How we use AI to build your software

The short version

AI coding agents write most of the code. A human decides what gets built, reviews every line before it ships, and is accountable for the result. That’s how we deliver in weeks what traditionally takes months.

What the AI does

AI coding agents write the scaffolding, the boilerplate, the repetitive work, and the first draft of tests. This is the part of software development that is time-consuming without being difficult, and it is what used to make projects take months.

What a human does

Everything that determines whether the software actually works for your business:

• Architecture and technical decisions — how the system is structured and what it’s built on
• Reviewing every line before it ships — AI-generated code is read, not skimmed
• Line-by-line review of the risky parts — anything touching login, permissions, or payments gets manual scrutiny, because that’s where automated output is most confidently wrong
• Security review
• Testing against the real requirements, not just whether the code runs
• Deciding what ships
• Talking to you — you deal with whoever is building your software, not an account manager

What we don’t let the AI decide

• Your architecture or technology stack
• Your user experience
• What's secure enough to ship
• What “done” means
• Anything we'd have to explain to you afterward

What this means for you

Speed. Weeks instead of months, because the slow mechanical part is automated.

Cost. You’re paying for judgment and review rather than for someone typing boilerplate.

Quality. Automated tests on the paths that matter, plus human review. Faster does not mean less tested — if anything, AI makes comprehensive test coverage cheaper, so there tends to be more of it.

Transparency. You always know how your software was built. It’s in your contract.

Your data

• Business and enterprise AI tools configured so your material is not used to train models. We keep a register of every tool we use, its data-retention terms, and when we last verified them. Ask for it any time and we’ll send it.
• No secrets in AI tools. Passwords, API keys, and credentials are never pasted into an AI system.
• No real customer data for testing. We use synthetic or anonymised data.
• If you have restrictions, tell us before we start. Some businesses, particularly in regulated industries, have policies against third-party AI processing. That’s a fine constraint to work within — but we need to know at the start, not at delivery.
• If personal data is involved, we put a data processing agreement in place first.

Two things we’ll tell you that others might not

We’d rather you hear these from us than from your lawyer after signing.

1. Copyright in AI-generated code is legally unsettled

The U.S. Copyright Office’s position is that material generated purely by AI isn’t protected by copyright, because copyright requires human authorship. The law here is still developing.

What this means practically: we assign you every right we hold in your software, and you get complete freedom to use, modify, sell, and build on it. What we won’t do is promise that an exclusive copyright exists in the AI-generated portions, because that may not be legally true — and a promise we can’t back is worth nothing to you.

2. AI tools can reproduce existing code

AI coding assistants learn from public code, and they can occasionally reproduce parts of it — including code under licences that would impose obligations on your business if it ended up in your product.

What we do about it:

• Run automated licence and provenance scans before delivery
• Provide a software bill of materials listing every component and its licence
• Configure our tools to suppress suggestions matching public code where that setting exists
• Flag anything that needs your attention before delivery, not in a footnote after

What we won’t claim: that this is a perfect guarantee. No scan catches everything, and this risk exists in traditional development too — developers have always copied from Stack Overflow. AI makes it more likely and more invisible, which is why we scan rather than assume.

Questions clients ask

“If AI writes the code, why am I paying you?”

Because the code is the easy part. Knowing what to build, structuring it so it still works in two years, catching what the AI got confidently wrong, and being accountable when something breaks — that’s the work. The AI made the typing faster; it didn’t make the judgment unnecessary.

“Is AI-built software lower quality?”

It’s as good as the review. Unreviewed AI code is genuinely bad, and there are studios shipping exactly that. Reviewed AI code, with tests, is comparable to hand-written code and usually better tested, because writing tests got cheap. That’s why our review process is in the contract rather than just on the website.

“What if I don’t want AI used on my project?”

Tell us. We can discuss traditional development, though the timeline and cost would be closer to conventional agency rates — that difference is exactly what AI-assisted delivery buys you.

“Who owns the software?”

You do, on full payment. See the copyright note above for the one nuance.

“What if you disappear?”

You get the source code, the repository, documentation, and a deployment you control. Nothing is locked to us. That’s true from the first delivery, not just at the end.