Building startups in the age of AI

First Thursday: building startups in the age of AI (and learning how not to lose your mind doing it)

This month’s First Thursday at the Innovation Centre focused on what it’s actually like to build something right now, with AI in the mix, expectations shifting, and productivity tools changing what we think a “startup” even is.

Pete Keevill opened the session by admitting what a lot of people still quietly feel - that a lot of AI output still feels beige. But things have moved quickly. He talked about conversations with founders building near-finished products in six weeks, solo. Which raises a real question. If you don’t need a team of ten, or investment just to get started, what does that mean for how we support early-stage ventures?

We heard from Tim Robinson and Aaron Fordonnell, both working on products where AI isn’t just a tool, it’s central. Both talked about the strange shift that happens when most of the usual startup friction disappears. Tim described it as a recursive rabbit hole. Prompting, iterating, researching, repeating. Aaron pointed out that the problem isn’t being blocked by constraints, it’s the opposite. When you can do ten things in a day, the real risk is burnout. The tools accelerate everything, including your own exhaustion.

Steve Edwards brought a useful framework for understanding how people are using AI. A lot of startups are still sitting at stage one - speeding up things they were already doing. That’s fine, but the real change kicks in when AI stops being a timesaver and starts becoming a strategic decision. At that point, the work itself changes. That’s where most of us are heading, whether we mean to or not.

We stayed in the practical. We heard examples of AI being used to create product images, write code, generate customer responses. It was useful to see how many people are experimenting, but also clear that this isn’t just about efficiency. These tools shape how decisions get made, especially in the early days of a business. Critical thinking still matters. It might matter more than it did before. You still have to know what you’re trying to build, and you still have to decide when it’s good enough.

The room had questions about learning, authenticity, and what happens when junior roles disappear. Several people talked about feeling overwhelmed - not by the technology, but by the volume of work it creates. Faster iterations, more output, more decisions, but fewer natural breaks and fewer people to share the thinking with.

One idea that got traction was the possibility of an AI Advisor in Residence. Not a tool tutor or a prompt coach, but someone who could challenge assumptions and help founders think clearly about their choices. Especially useful when the speed of execution makes it easy to skip the thinking part.

We’ll pick that up again at the next session, which is the Christmas edition, and in the afternoon. Less tech, more reflection, and a mince pie or two.

Finally, a thought to leave you with. If time, money and technical skill are no longer the main constraints, what is? And how do we make sure we’re still building things that matter, instead of just building fast?