You've been experimenting with AI development tools for a while now. What surprised you most?

Without a doubt: the speed.

You can go from a blank canvas to a working application in just a few hours. Even with very little context, AI starts making sensible assumptions and proposes features and flows you hadn't even considered.

What also stood out is how accessible it has become. Experimenting with AI integrations or external functionality no longer requires a complex setup, making it incredibly easy to validate ideas. For prototypes, proof-of-concepts or side projects, what used to take days can now be explored in an afternoon.

It's not always perfect, but it's remarkably good at turning an idea into a convincing first version, fast.

So... is AI replacing developers?

I don't think so. The moment an application becomes more complex, you start noticing the limitations.

Sometimes you'll ask the AI to fix a bug and it'll confidently tell you everything is solved, while nothing has actually changed. It can get stuck repeating the same solution without understanding the underlying problem.

More importantly, you're working inside a bit of a black box.

As a developer, you can inspect the generated code, but you still don't really know why certain architectural decisions were made or whether everything meets the standards you'd expect for production software without thorough inspection.

So for business-critical applications, that's simply not enough.

Is complexity the biggest difference?

Absolutely.

Simple applications are becoming much easier to build with AI, and I think that's a fantastic evolution.

Years ago we had website builders that allowed anyone to create a basic website. AI is now pushing that boundary much further. You can already build relatively simple web applications surprisingly well.

But once you're building software that becomes the heart of a company, complex business logic, integrations, permissions, security, scalability, name it, that's a completely different challenge.

Those are the kinds of products we build at Hatch, and they require far more than simply generating code.

You also mentioned responsibility. Why is that so important?

Because as a software studio, we're not just delivering code.

We're taking responsibility for the quality, reliability and security of what we build.

If an AI generates part of an application and something goes wrong later, the customer won't blame the AI. They'll hold us accountable, and rightly so.

That's why we can't simply say: "The AI built it, so it must be fine."

If AI helps us generate software, we still need to review it, validate it, audit the security, understand the architecture and make sure it's something we're comfortable putting into production.

That's a responsibility you can't outsource.

Where do you think AI fits into the future of software development?

I actually think it's going to strengthen the early stages of product development enormously.

At Hatch we already spend a lot of time validating ideas before writing production code. AI allows us to explore many more concepts in a fraction of the time.

Instead of spending days building one prototype, we can now create several alternatives within hours.

That leads to better conversations with clients because ideas become tangible much earlier in the process.

Designers can also showcase interactions, animations and user flows much sooner, making collaboration between design and development even stronger.

Are there any risks?

Definitely.

We're already seeing clients arrive with AI-generated prototypes. That's actually great, because it helps turn ideas into something concrete.

The risk is assuming that a prototype is already a product.

Often, the underlying questions haven't been answered yet. How should the data model evolve? Will the architecture still work in three years? Is the application secure? Will it scale? Can future features be added without rebuilding everything?

Those aren't questions AI can answer on its own.

That's exactly where experienced product designers and developers continue to make the biggest difference.

What advice would you give companies that are excited to start building with AI?

Use it. Experiment with it. Validate ideas. Build prototypes. Learn what it can do.

But if you're building software that your business will depend on for years, don't confuse speed with quality.

AI is an incredible accelerator, but long-term software still requires thoughtful architecture, technical expertise and human judgement.

The technology is evolving incredibly fast, and we're excited about everything it enables. But we're also careful not to mistake a fast first version for a sustainable product.

So bottomline; AI is incredible at building a first version. But building software isn't just about writing code, it's about taking responsibility for everything that comes after.