Appoly Intelligence
7 min read adoptionstrategytrends

The AI adoption gap: what the 2026 data actually means for your business

UK AI adoption is rising, but most usage is shallow. The gap between intent and action is where the real risk sits. Here's what the verifiable data says and what to do about it.

By Appoly Intelligence

The headline figures for UK AI adoption look encouraging until you read them properly. The British Chambers of Commerce reports that 35% of SMEs now use AI in some form, up from 25% a year earlier. The Office for National Statistics, across a broader sample, puts usage at roughly 23%, up from around 9% two years before. Investment is up. Awareness is up. The government has an AI Opportunities Action Plan.

But dig into what those numbers actually mean and a different picture emerges. One that matters more for leadership teams trying to decide what to do next.

What the numbers really say

Gallup’s late-2025 survey of more than 22,000 workers found that 49% had never used AI at work. Not once. Only 12% were daily users, up from 10% the year before. That is not a revolution. That is a slow creep.

McKinsey’s global survey is more optimistic on the surface: 88% of organisations say they use AI in at least one business function, up from 78% a year earlier. But only around 6% qualify as high performers seeing meaningful financial returns, and most organisations describe their rollout as experimental rather than scaled. The vast majority are piloting, not deploying.

The UK government data is consistent with that gap between use and maturity. DSIT’s research found only 16% of firms with five or more employees using at least one AI technology, and the single most common reason for not adopting was that 71% had not identified a clear use for it. Adoption without a clear purpose isn’t progress — it’s just risk piling up quietly.

The gap is not between businesses that use AI and businesses that do not. It is between businesses that use AI with intent and businesses that use it without structure.

What people are actually doing with it

The most common use case, by a wide margin, is text generation. Emails, reports, proposals, meeting summaries. The boring stuff, done faster.

That is not a criticism. Speed on repetitive communication tasks is a genuine gain. But it is a narrow gain. OpenAI’s analysis of how people use ChatGPT found that the large majority of conversations are guidance, information-seeking, and writing tasks. The dominant behaviour is still “ask a question, get an answer” rather than integrating AI into a workflow.

Microsoft’s Work Trend Index research draws a related distinction between people who treat AI as a command-based tool and people who use it as a thinking partner. The first group is larger. The second group tends to see better results. Same tools, different habits.

Gallup found another pattern worth noting: leaders and senior managers use AI noticeably more often than individual contributors. That makes sense, leadership roles involve more synthesis and drafting. But it also means the people doing the day-to-day work where AI could save the most time are often the ones using it least.

The barriers are not what most people think

Cost is rarely the blocker. The DSIT research found that 71% of businesses had not identified a clear use for AI, and roughly 60% cited a lack of skills and expertise. Budget was a primary barrier for only a small minority.

This is important because it changes the solution. If the problem were cost, the answer would be cheaper tools. The problem is clarity. Businesses do not know where AI would help, or they do not have the internal capability to find out, or they have tried something that did not work and lost momentum.

This matches our diagnostic work. The organisations getting a return are not the ones with the biggest budgets. They are the ones that wrote down what they were trying to achieve, which forces the prioritising, measuring, and iterating that actually produces value.

The lowest-return programmes are not the ones that spent too little. They are the ones that spread modest budgets across too many disconnected pilots. Focus beats scale in the early stages.

Where the real risk sits

There is a temptation to read these numbers as evidence that AI is overhyped and can wait. That would be a mistake.

The businesses that are getting value from AI are not necessarily the ones with the largest budgets. They are the ones that identified a specific problem, ran a focused pilot, measured the result, and expanded from there. The gap between those businesses and everyone else is widening. Not because of technology. Because of execution.

The competitive risk is not that a competitor will suddenly deploy a transformative AI system and leap ahead. It is that they will quietly remove cost and friction from a process you both share, and you will not notice until your margins are thinner and your response times slower.

This is particularly relevant for independent UK businesses, from SMEs through to enterprise. You do not have the resources to experiment broadly. You need to choose the right intervention, implement it properly, and move on. The businesses doing this well are treating AI as operational improvement, not innovation theatre.

Three practical moves for the next quarter

If you are in the majority of UK businesses that have not yet found a clear path, there are three things worth doing in the next twelve weeks.

1. Look at one process properly

Pick the process that costs you the most time or the most errors. Not the one that would be most impressive to automate. The one that hurts every week.

Measure it. How long does it take? How often does it go wrong? What does that cost? Then ask whether AI could make it faster, cheaper, or less error-prone. If the answer is yes, you have a candidate. If the answer is no, you have ruled something out, which is also useful.

The businesses that make progress start with specificity. “We want to use AI” is not a plan. “We spend twelve hours a week reconciling invoices and most errors happen because suppliers use different formats” is a plan.

2. Write the strategy down, even if it is one page

You do not need a forty-slide deck. You need a single document that says: here are the two use cases we will pursue this year, here is what success looks like, here are the guardrails, and here is who owns it.

Writing it down is what matters. Not because the document creates value, but because the act of writing it forces clarity. Without that clarity, budgets drift, pilots multiply, and nothing reaches production.

3. Train for the role, not the tool

The skills gap is real, but the solution is often misapplied. Sending one person on a generic AI course does not change how the business operates. Training needs to be role-specific: what does AI mean for a customer service agent, a finance controller, a project manager?

Tie the training to a live use case and the person who owns it. Capability that is embedded in how people actually work sticks. A standalone course detached from the work rarely does.

What to ignore for now

Three things are worth leaving alone until you have working systems in the basics.

Agentic AI. The vendor pitches for autonomous workflows are loud and getting louder. The reality is that deployment of genuinely agentic systems is still rare, and the failure rate for early implementations is high. Master assisted workflows first. Autonomy comes later.

Custom model training. Building or fine-tuning your own language model is almost never the right answer for an independent business. The costs are high, the expertise is scarce, and the improvement over good prompting with existing models is usually marginal. Start with what is available off the shelf.

Strategy without implementation. There is a thriving market for AI maturity assessments, roadmaps, and framework documents. These are not useless, but they are not a substitute for running one pilot and learning from it. Strategy without implementation is expensive speculation.

The bottom line

The 2026 data tells a clear story. AI adoption is growing but shallow. Most usage is basic. The barriers are clarity and skills, not cost. The businesses seeing returns are the ones with focus, measurement, and a written plan.

The question for leadership teams is not whether to engage with AI. It is whether to engage with it now, with structure, or later, when the gap has grown harder to close.

If you want an outside view on where to start, get in touch. Twenty minutes on the phone is usually enough to know whether a diagnostic is worth doing.

Sources

Related reading

Most UK businesses are adopting AI without a plan. Here's what that costs.

UK AI adoption is climbing fast, but most firms still have no written strategy. The verifiable data shows the gap between intent and structure is where budgets get burned.

Read

AI adoption in 2026: where to start when you don't know where to start

Most UK businesses know AI matters. Few know what to do about it. Here's a practical starting point based on what we're seeing in the field.

Read

Twenty minutes. No deck.

A short call. We'll talk through what you're looking at, give you an honest read on whether AI fits, and tell you roughly what it would cost to find out properly.

Start with a diagnostic
Start with a diagnostic →