Summary

Official ONS statistics indicate that UK businesses are not deepening their use of artificial intelligence. Around 9 per cent of firms had adopted AI in 2023, far below cloud computing and specialist software, with adoption projected to rise towards 22 per cent. The most common uses are free or low-cost tools for text generation, data processing and images, and firms overwhelmingly deploy AI for efficiency savings rather than new products. Larger, better-managed firms adopt fastest, while adopters and non-adopters show similar productivity, so AI's growth payoff remains unproven. For Andy Burnham's productivity-focused government, turning broad but shallow adoption into deep integration is the central challenge.

Five Key Takeaways

  1. AI adoption in the UK is widening but remains shallow, with most firms relying on free or low-cost tools.
  2. Businesses overwhelmingly use AI for efficiency savings rather than to develop new products or services.
  3. Large, well-managed firms are adopting far faster than smaller ones, risking a wider productivity divide.
  4. AI adopters are not yet outperforming non-adopters on productivity, leaving the growth payoff unproven.
  5. For Burnham's growth agenda, converting broad AI use into deep integration is the real economic test.

What Happened

Fresh scrutiny of how British businesses use artificial intelligence points to a country with a breadth problem and a depth problem at the same time. Official statistics, and the analyses built on them, describe a familiar pattern: many firms have tried AI, most of those that use it lean on cheap or free tools, and comparatively few are rebuilding their products or business models around it. As Andy Burnham takes office promising to raise growth in every part of the United Kingdom, the gap between dabbling and deploying has become a live economic question.

The Office for National Statistics found that around 9 per cent of UK firms had adopted at least one form of artificial intelligence in 2023, with adoption projected to climb towards 22 per cent the following year. Even the higher figure sits well below the take-up of older digital technologies: roughly 69 per cent of firms used cloud computing and about 61 per cent used specialist software. AI, in other words, remains the exception rather than the rule in the average British workplace.

The kinds of AI in use point to shallow rather than transformative adoption. The most commonly reported applications are text generation using large language models, data processing using machine learning and the creation of visual content, tasks that widely available and often free consumer tools now perform. More capital-intensive forms, such as AI-enabled robotics, remain rare, reported by only around 3 per cent of firms.

Crucially, the reasons businesses give for adopting AI tilt heavily towards efficiency rather than invention. Among firms planning to take it up, close to 69 per cent intended to use AI to upgrade existing processes and methods, and around 42 per cent to automate tasks currently done by staff. Only about 24 per cent said they planned to use it to broaden the products or services they offer. The technology is being used to do the same things more cheaply, not to make new things.

The picture is one of enthusiasm running ahead of embedding. Britain does not lack curiosity about AI; workers experiment with chatbots, marketing teams generate copy and analysts lean on machine learning to sort data. What is missing is the step from casual, often free-of-charge use towards the deeper integration that changes how a business fundamentally operates. That step is where economists expect the real productivity dividend to lie.

Why It Matters

The distinction between using AI to save money and using it to create something new matters more than it sounds. Efficiency gains can lift a single firm's margins, but the larger prizes for a national economy arrive when a technology spawns new products, services and markets. Britain's productivity has grown weakly since the financial crisis, and successive governments have looked to digital technology to break the stagnation. If AI is absorbed mainly as a cost-cutting tool, its contribution to overall growth may prove more modest than the surrounding enthusiasm implies.

So far the data offer little comfort to the optimists. Analyses of the firm-level figures find that AI adopters and non-adopters show broadly similar productivity distributions, meaning users are not yet pulling clearly ahead of the pack. That may reflect how early it is, since the productivity benefits of general-purpose technologies have historically arrived with a lag, once firms reorganise around them. But it is also consistent with a world in which shallow, bolt-on use of free tools simply does not move the needle.

There is a distributional dimension too. Larger companies are adopting AI far faster than smaller ones, with take-up among firms of 250 or more employees running at roughly 44 per cent against around 26 per cent for the smallest businesses. That gap risks compounding existing divides between well-resourced incumbents and the long tail of smaller firms, an uneven pattern that speaks directly to the new government's promise to spread growth beyond a handful of prosperous hotspots.

The reliance on free tools carries its own trap. Consumer-grade chatbots are easy to try and cost nothing, which lowers the barrier to a first experiment. But they are rarely integrated with a firm's own data, workflows or governance, so many pilots stall before they deliver measurable value. Depth, not breadth, is where the economic return tends to sit, and it is precisely depth that the statistics suggest is missing.

The stakes extend beyond any single balance sheet. Rival economies are chasing the same technology, and a country whose firms treat AI as a marginal cost-cutting aid risks falling behind those that use it to reinvent industries. For Britain, still searching for a reliable engine of growth, the difference between shallow and deep adoption could help decide whether AI becomes a genuine productivity story or another wave of investment that flatters activity without transforming it.

Detailed Analysis

Behind the headline adoption rate lies a more textured story about which firms are moving, what is holding others back, and why free tools have become the default entry point. Three threads help explain why British AI use has stayed broad but shallow.

Why free tools dominate

The most widely reported uses of AI in British firms cluster around tasks that free or low-cost software already handles well. Writing and summarising text with large language models, sifting data with machine learning and generating images are all available through consumer services that require no bespoke engineering and little upfront spending. For a cautious business, that is an attractive way to test the water without committing a budget.

The barriers firms cite explain why many go no further. Asked what stops them adopting AI, close to 39 per cent pointed to the difficulty of identifying a worthwhile business application, around 21 per cent cited cost and about 16 per cent a lack of in-house expertise. These are not problems a free chatbot solves. Identifying where AI genuinely adds value, and building the skills to embed it, requires the kind of investment that a no-cost tool implicitly discourages.

The result is a kind of ceiling. Free tools deliver quick, visible wins on routine tasks, which is often enough to satisfy an initial business case. But because they sit outside a company's core systems, they seldom become the foundation for redesigning how the firm actually works. Adoption widens across the workforce without ever deepening into the operating model.

The efficiency reflex

When British firms do commit to AI, they overwhelmingly frame it as an efficiency exercise. Around 36 per cent of adopters said they used it to improve business operations, and the stated intentions of prospective adopters point the same way, with process improvement and task automation far outranking the expansion of products and services.

This efficiency reflex is understandable. Cost savings are easier to measure, quicker to realise and less risky than betting on an untested product line. A finance director can sign off automating invoice processing far more readily than a speculative venture into AI-enabled services. But an economy that uses a powerful new technology mainly to trim existing costs captures only a fraction of its potential.

The contrast with earlier general-purpose technologies is instructive. Electricity and the computer eventually reshaped what firms produced, not merely how cheaply they produced it, but only after businesses reorganised around them. The ONS picture suggests British firms are, for now, stuck at the earlier stage, using AI to speed up yesterday's tasks rather than to imagine tomorrow's.

None of this means the efficiency use is worthless. Automating routine administration frees staff for higher-value work and can, over time, create the headroom and confidence to attempt something more ambitious. The concern is that many firms treat the first efficiency win as the destination rather than a staging post, banking the saving and going no further.

Management, size and the productivity puzzle

The firms that adopt advanced technology are not, at this stage, the most productive so much as the best managed and the largest. Structured management practices, such as tracking performance against targets and promoting on merit, are among the strongest predictors of adoption. Among the best-managed firms, some 88 per cent had taken up at least one advanced technology, against around 51 per cent of the worst-managed.

Size compounds the pattern. Bigger firms have the cash, data and specialist staff to move from experimentation to integration, which is why their adoption rates run so far ahead of smaller companies. Technology adopters overall have been associated with around 19 per cent higher turnover per worker than non-adopters, though that advantage is concentrated among well-managed businesses rather than spread evenly across the economy.

Yet the AI-specific productivity signal remains faint. Because adopters and non-adopters currently show similar productivity, it is too early to claim that AI is delivering broad economic gains. The optimistic reading is that the payoff is still building as firms learn; the pessimistic reading is that shallow adoption of free tools will never generate much. Which interpretation proves right will shape how far AI contributes to the growth the new government has promised.

What Happens Next

Policy is already nudging in the direction of wider access. Government-backed initiatives have made free AI training and tools available to smaller firms, on the logic that lowering the barrier to entry will lift the national adoption rate. The harder task, on the evidence, is not getting more businesses to try AI but helping those that have to move from casual use towards genuine integration with their data, systems and staff.

For Burnham, whose growth mission leans heavily on raising productivity outside London and the South East, that distinction is central. A devolution agenda that spreads investment and skills could, in principle, help smaller and regional firms deepen their use of technology rather than stall at the free-trial stage. But such gains take years to show up, and they compete for money at a moment when the Iran-war energy shock and tight fiscal rules leave little room for open-handed spending.

The risk is a widening divide. If large, well-managed firms keep pulling ahead while smaller businesses cycle through free tools without embedding them, the productivity gap between the two could grow rather than close. Targeted support for skills, data infrastructure and practical use-case identification, rather than tools alone, is where analysts increasingly say the effort should go.

There are reasons for cautious optimism. Adoption is a moving target, and the tools themselves are improving quickly, lowering the cost of the deeper integration that has so far eluded most firms. If better products make it easier to connect AI to a company's own data and processes, the depth problem could ease even without a dramatic policy intervention. Equally, it could harden if only the largest firms can afford to keep pace.

The next official statistics will be the test. The projected rise towards 22 per cent adoption suggests breadth is still increasing, but the more telling number will be whether the share of firms using AI to build new products, rather than merely to cut costs, starts to climb. Until it does, Britain's AI story will remain one of enthusiasm outrunning depth.

Conclusion

Britain is not short of curiosity about artificial intelligence. Firms across the economy have tried it, workers use it daily, and adoption is still spreading. What the official data expose is a shortage of depth: a tendency to reach for free tools and quick efficiency savings while leaving the harder, more valuable work of building new products and rebuilt processes largely undone.

That matters because the economic prize from AI lies less in doing old tasks a little cheaper than in creating things that did not exist before. On current evidence, adopters are not yet outpacing their rivals on productivity, and the firms going deepest are the largest and best managed, not necessarily the most numerous.

For a government that has staked its credibility on faster, more evenly spread growth, converting breadth into depth is the real challenge. Whether Britain's businesses can turn a broad but shallow embrace of AI into measurable productivity gains will be one quiet but telling measure of the new administration's economic ambitions.