Most firms can build a demo, but few can own what comes next

Colin Carmichael has spent the past few years watching UK financial services firms try to work out what to do with generative AI. He is a Client Partner at 4most, a specialist risk, data, and analytics consultancy, that works with banks and insurers in the UK and globally. He is also co-founder of the Business AI Alliance, a not-for-profit giving SMEs a voice on national AI policy.

From engaging with both large and small firms, his read is the same: the easy part is over.

Most organisations Colin works with have already built a proof of concept. The problem is what comes next. “Proof of concepts are great. It gets everyone understanding technology, but moving to an operational state is different.”

A hallucinated response in a demo is something you note and move on from. In production it needs to be “more or less right all the time.” That means prompt engineering, model validation, and decisions that most organisations haven’t had to make before. Who signs off the output? Who’s accountable when something goes wrong with a customer? What’s the remediation?

These questions apply everywhere, regulator or no regulator. Firms that haven’t worked through them before going live are the ones that get stuck, not because the technology fails but because nobody owns what it does.

The usual explanation for why mid-market firms are slower is some version of a skills gap. Colin doesn’t buy it. “I don’t think talent’s the problem. I think it’s the time.”

Owning an AI outcome takes iteration and the capacity to work through failure. Larger firms can carve that out. Smaller ones can’t: their people are already carrying full workloads, often mid-way through other change programmes. Curious individuals end up experimenting in their own time, if at all. What looks like a skills shortage is really an accountability vacuum: nobody has the time to own something that needs sustained attention to get right.

It’s a gap aibl keeps finding. Firms have people who can build with AI and people who understand the business problems, but not someone whose job is to connect the two and own what comes out.

Colin sees it clearly in banking. Neo banks arrived at generative AI with cloud infrastructure and no legacy stack to migrate. They were already set up to own outcomes. Challenger banks and building societies mid-way through digital transformations face a different problem. AI lands on top of a change stack that was already full, and nothing gets the sustained attention it needs.

His answer isn’t headcount, it’s foundations: operating model, technology, data, governance. Without those, adding people just gives you more capacity to experiment.

Underneath the operational problem is a people one. The employees who understand the business problems best are often the least comfortable with the technology. The ones comfortable with the technology don’t know the business well enough to know what’s worth owning. Most firms, in Colin’s experience, have a handful of curious individuals driving experimentation and a majority getting on with their day jobs. Leadership has to bridge the two, or the right people never end up in the same room.

Experimentation was the easy part. The task for 2026 is to filter out the noise, pick two or three ideas with real value, and industrialise them. AI can no longer be delegated to a handful of curious staff while everyone else stays focused on the work already in front of them.