The Assay · 8 July 2026

Where AI Actually Fits in a UK SME

The AI conversation has become very loud very quickly. Hundreds of billions in capital committed to a handful of companies. CFOs at the largest organisations in the world repositioning themselves as the people accountable for AI return on investment. A much-cited MIT finding that 95% of enterprise AI pilots show no measurable return within six months. All of this is real. None of it was written for the owner of a £15m business in the East Midlands.

That gap matters — because the conclusions owners draw from a conversation that wasn’t designed for them tend to be one of two things. Either AI is for bigger businesses, and the relevant moment hasn’t arrived yet. Or they buy a tool, deploy it without a clear problem to solve, and confirm the first conclusion six months later.

Both paths lead to the same place: a business where the owner’s intuition and institutional memory remain the primary operating system.

What’s actually happening

Two things are true simultaneously, and the noise tends to collapse them into one.

The first: the tools available to a UK SME today are genuinely capable, affordable, and require no enterprise infrastructure to deploy. The gap between what a £500m business can do with AI and what a £15m business can do has narrowed considerably — not because the enterprise tools became cheaper, but because a different tier of tooling emerged that requires no IT department, no systems integration, and no six-figure implementation budget.

The second: the pattern of failure that produces that 95% statistic in large organisations is replicated at smaller scale. Less money wasted. Less visible. But the same mechanism underneath. The organisation — or the owner — selects a tool before defining the problem. The tool gets used for tasks it can do, rather than for the tasks that would actually matter. Productivity in the wrong direction is still motion. It’s just not progress.

The owners I speak to who are quietly getting value from AI are doing it the other way round. They started with an operational question — where does this business depend most heavily on one person, one process, or one piece of knowledge that lives nowhere except inside someone’s head? — and worked backwards to find out whether any available tool could reduce that dependency. Some found one. Some found that the dependency was structural and no tool was the right answer. Both outcomes are useful.

Why the misunderstanding persists

The AI conversation is dominated by the people building infrastructure, the people buying it, and the people writing about both. That conversation is not dishonest. It simply doesn’t include the UK SME owner, because the scale isn’t commercially interesting to the vendors or the commentators driving the narrative.

The result is a signal problem. The information that reaches an owner — through newsletters, LinkedIn, conferences — is calibrated for an audience operating at a different altitude. When a CFO at a global technology company describes AI as a capital allocation decision involving billions of dollars of compute, they are describing their reality accurately. That description, received by an owner running a manufacturing business with 40 people, produces one of two responses: irrelevance or anxiety. Neither is the right frame.

There is a quieter set of conversations happening — between owners who have found a specific, contained use for AI in their operation and are not writing Substack newsletters about it. Those conversations don’t reach the feed. They don’t become case studies. They remain competitive advantage, which is exactly where their owners want them.

What changes when you see it clearly

The relevant question for an SME owner is not “what AI tools should I be using?” It is: which part of this operation depends most on me — or on one other specific person — and what would it mean if that dependency could be reduced?

That is an operational question. It has been the right operational question for thirty years. AI is one possible answer to it today, where previously the answer might have been a documented process, a trained deputy, or a software system built for a specific industry. The mechanism is the same. The tooling has improved.

The connection to business robustness is direct. A business that has reduced its key-person dependencies — through AI, through process, through people — is a more robust business. Outsiders — a fractional CFO, a consultant, anyone paid to look with fresh eyes — can see this immediately. Owners often cannot, because they are inside the dependency and it feels like indispensability rather than fragility. The distinction matters more than most owners realise, and it tends to surface at exactly the wrong moment.

The quieter question

The AI conversation will get louder before it gets quieter. The capital is committed. The tools will continue to improve. The commentators will continue to describe what the largest organisations are doing with the most advanced versions of technology that most SMEs will never need.

None of that is a reason to wait.

The owners who will look back on this period as one in which they made genuine operational progress will not be the ones who followed the enterprise narrative. They will be the ones who ignored the altitude of that conversation, asked a much simpler question about their own operation, and used whatever tool — AI or otherwise — happened to be the right answer to it.

If that question is worth asking about your business — you know where to find me.

The Assay is written by Rudi Prenzlin at Scalable Growth, a UK practice that gives owner-managed businesses a fractional CFO and the operational structure to run without every decision going through the owner.

← All notes Begin a conversation