New research published this week has put a number on something a lot of SME owners suspect but rarely say out loud: most UK businesses think they’re doing better on AI than they actually are. The study found that while a striking 38% of businesses say they’ve moved agentic AI projects beyond the pilot stage, only 12% are making significant company-wide progress and seeing real return on investment from it. In other words, plenty of businesses have an AI pilot running somewhere — far fewer have anything that’s actually changed how the business operates or what it earns.
This isn’t a story about AI failing to deliver. It’s a story about the gap between “we tried a tool” and “we changed a process,” and that gap is where most of the wasted budget and misplaced confidence sits. For SME owners weighing whether to invest further, the finding is genuinely useful: it means the businesses pulling ahead aren’t the ones with the flashiest pilot, they’re the ones who took a narrower slice of the business and actually finished the job.
Why the gap exists
The pattern behind these numbers is consistent across most AI adoption research this year: businesses start a pilot, get an encouraging early result on a narrow task, and then struggle to extend that success beyond the person or team who ran it. A chatbot that helps one person draft emails faster is a pilot. The same capability wired into your actual customer response workflow, used consistently by everyone on the team, with someone accountable for how it performs, is the company-wide progress the research says is rare. The difference isn’t the technology — it’s the unglamorous work of integration, training, and ownership that most pilots never get funded to do.
What separates the 12% from the rest
Businesses seeing real ROI tend to share a few habits: they pick one process with a measurable outcome — invoice chasing, first-line customer queries, report drafting — rather than trying to “do AI” everywhere at once. They assign a named person to own the rollout rather than letting it live as a side project. And they measure something concrete before and after, so “is this working” has an actual answer instead of a gut feeling.
If your business has an AI pilot that’s stalled somewhere between “interesting demo” and “part of how we work,” that’s usually a scoping and ownership problem rather than a tooling one. ApplyAI works with UK SMEs specifically on that gap — taking a pilot that’s proven itself on a small scale and building the plan to actually embed it, rather than leaving it to quietly fade out. For businesses that need something built rather than just configured, BuildApps does the custom implementation work that turns a promising pilot into a tool your whole team actually uses.
The takeaway
If you’re not sure whether your business’s AI efforts fall into the 38% with a pilot or the 12% with real results, the honest test is simple: pick one AI-assisted process and ask whether it’s used by everyone who should be using it, owned by someone specific, and measured against what it looked like before. If the answer to any of those is no, that’s the next thing to fix — not a brand-new tool.
