Skip to content
AI in practice 6 May 2026 6 min read

AI made it easy to start. It hasn't made it easier to ship.

Andrew Cowle

Andrew Cowle

Founder

Prototyping has been democratised. The hard parts of software — sequencing, architecture, data, scale, organisational fit — haven't moved.

Anyone with a credit card and an afternoon can stand up something that looks like a working product. The bar to “I built a prototype” has collapsed in two years. We’ve never had a better moment for someone with an idea to test the idea.

But shipping isn’t prototyping. Shipping is the part where real users find the edge cases the demo never tried, where the second-week feature request collides with the architecture choice from week one, where the cost of being wrong about the data model becomes visible in a way it never was in the prototype. None of that has gotten easier. If anything, the gap between “prototype” and “production” has widened — because the prototype is now so easy that more people get there without the experience that makes the rest survivable.

The hard parts haven’t moved

Sequencing — what to build, in what order — was the hardest part of software ten years ago and it still is. AI doesn’t have an opinion about whether you should build the integrations layer first or the user-facing flow. It’ll happily generate either, beautifully, in an afternoon. The question of which one you should generate first is a call that depends on your business, your users, and your runway. AI doesn’t make that call. Someone who’s been in this position before makes that call.

Architecture is the same story. The decisions that are expensive to undo — the data model, the boundaries between services, the choice of where state lives — are made on the first page of code. AI is a fluent collaborator on the second page. The first page is still where seniority earns its keep.

And then there’s everything around the software. Onboarding. Operations. Compliance. The conversation with your CFO about why the AWS bill went up. Real software lives inside a real organisation, and the organisation doesn’t get prototyped.

Where AI actually moves the needle

This isn’t a “don’t use AI” piece. We use it heavily — for scaffolding, tests, documentation, migration grunt-work, the parts of building software that benefit from a fluent collaborator who never gets tired. A senior engineer with modern AI tooling delivers more in a day than one without it. That difference is real, and it’s the reason the team you need from a firm like ours is smaller than the team you’d hire yourself.

The mistake is thinking that compounds upward. It doesn’t. AI accelerates the things AI accelerates; it doesn’t replace the things it doesn’t replace. The teams that will eat the AI-tooled solo-founder’s lunch on anything beyond the prototype are AI-accelerated teams who have done this before.

What this means in practice

If you’re at the prototype-and-deciding-what-next stage, the question isn’t “how do I get more AI in here?” It’s “do I have the senior engineering experience on the decisions that are expensive to undo?” That question gets answered the same way it always did: by getting people in the room who’ve shipped this kind of thing.

We’re those people. The discovery call is where we work out whether we’re the right ones for your specific build.

Andrew Cowle

Andrew Cowle

Founder

Working on something similar? Let's compare notes.

Liverpool, UK. Available across the UK and remote.

What happens next...

Someone on the team usually replies within a working day.

  1. 01

    We read every message

    Usually within a working day. The first reply comes from a person on the team, not an autoresponder.

  2. 02

    If there's a fit, we set up a 30-minute call

    We listen, ask questions, and try to work out whether we're the right people for what you're building.

  3. 03

    If we can help, we send a written proposal within a week of the call

    An honest indicative range, a shape for the engagement, and the people we'd put on it.