The State of UK AI Adoption 2026
The State of UK AI Adoption Survey 2026 is the largest recent study of AI adoption in the UK mid-market. 755 senior leaders in UK organisations with revenue between £50m and £500m answered, across technology, operations and finance, marketing and sales, HR, and the C-suite.
Fieldwork ran from January to March 2026, in partnership with Executive Summary. The findings below are free to read, quote and cite with acknowledgement.
The mid-market has adopted AI, but most of it cannot yet prove the return. Half of leaders can point to measurable ROI, and only 14 per cent have scaled AI across three or more functions with a number to show for it. What separates the companies that get a return is not their tooling or their budget. It is how they govern AI, and how well their people can actually use it.
Headline findings
Adoption is settled; proof is not. 68 per cent of leaders describe themselves as all-in or building momentum on AI, and 65 per cent can approve a new AI tool in days or weeks. But only 49.6 per cent report measurable AI ROI today, and just 14 per cent have scaled AI across three or more functions with a measurable return.
- Governance is the difference. The share of companies reporting measurable ROI rises from 22.2 per cent with no governance to 85.3 per cent with mature, embedded governance, a 63-point gap on much the same tools.
- A policy no one follows can be worse than none. Companies with a documented-but-inconsistent AI policy sometimes report lower measurable ROI than companies with no policy at all.
- Capability is the second lever. Where people are well equipped and the company is governed, measurable ROI reaches 81 per cent. With only one of the two it sits near 40 per cent; with neither, 15.
- Shadow AI is the majority experience. 55 per cent of leaders say unapproved AI use is common or very common, driven mostly by official approval being too slow.
Governance decides who gets a return
Sort every company by how mature its AI governance is, and the share that can show a measurable return climbs almost in a straight line. The companies at the top are not running better tools. They have built the measurement and accountability to show what the tools do.
| Governance level | Measurable ROI | Share of market |
|---|---|---|
| No governance | 22.2% | 5% |
| Informal guidelines | 27.7% | 19% |
| Defined, applied inconsistently | 31.7% | 26% |
| Formal, organisation-wide | 59.0% | 29% |
| Mature and embedded | 85.3% | 21% |
The largest single jump is from a policy that merely exists to one that is actually enforced, where the return roughly doubles. Half the market sits below that step.
The policy trap
A written AI policy that no one follows is not a safe halfway point. In HR, companies whose governance is defined but applied inconsistently report 16 per cent measurable ROI, below the 33 per cent reported by HR functions with no governance at all. In technology and IT, informal guidelines (40 per cent) beat a defined-but-inconsistent framework (33 per cent). The reason is false confidence: a framework on the intranet feels like the job is done, so the money to enforce it never gets spent.
Governance beats enthusiasm
We crossed how enthusiastic a company is about AI with how well it is governed. Governance moved the return; enthusiasm on its own did not.
| Governed | Ungoverned | |
|---|---|---|
| Momentum or all-in | 75% | 39% |
| Not enthusiastic | 54% | 14% |
A company that is all-in but ungoverned (39 per cent) does worse than one that is well governed but unenthusiastic (54 per cent). The biggest single group in the whole market is exactly this one, all-in with no governance, and it also returns the least.
Ownership is the widest structural lever
Who owns AI delivery moves the return more than any other structural choice, a 44-point gap from CEO ownership to no owner.
| Who owns AI delivery | Measurable ROI |
|---|---|
| CEO or C-suite | 62% |
| Central IT | 52% |
| Dedicated data or AI team | 45% |
| Distributed across functions | 41% |
| No single owner | 18% |
The dedicated data or AI team, the model that sounds most serious, comes third: a specialist team set apart from the business tends to build AI that sits at a distance from it.
Governance also brings shadow AI into the open
Shadow AI falls as governance matures, from common in 75 per cent of ungoverned companies to 35 per cent of the most mature. When the approved path works, people stop routing around it. Companies that approve tools in days report 62 per cent measurable ROI and the least shadow AI; those that take months report 38 per cent and far more. When the sanctioned route is quicker than the workaround, people take it. Policing the workaround does not do the same job.
Confidence is not the same as control
One in four of the leaders who say they are very confident their AI use is compliant and enforced are, by their own account of their governance, at the lowest levels of maturity. Their confidence outruns their controls, and it costs them: 51 per cent measurable ROI, against 84 per cent for the leaders whose confidence is backed by real governance. And it starts with agreement: only 24 per cent of companies have full management alignment on AI priorities, where measurable ROI is 79 per cent, against 37 per cent where senior colleagues mostly disagree.
People and capability: the second lever
Governance is what makes AI measurable. Getting people able to use it is a separate problem, and solving only one of the two leaves a company stuck in much the same place.
| People well equipped | People not well equipped | |
|---|---|---|
| Governed | 81% | 42% |
| Ungoverned | 40% | 15% |
Leaders who use AI themselves run better companies
The strongest people signal in the survey is the most personal: how much the leader uses AI themselves. Leaders who touch AI only for the odd admin task run companies at 31 per cent measurable ROI; that lifts to 50 per cent once the leader uses AI for judgement and decisions, and to 71 per cent where AI is built into their working day. It adds to governance rather than duplicating it: a hands-on leader in a governed company reaches 80 per cent, against 62 for a hands-off one. The most senior leaders are the most hands-on, at 43 per cent of the C-suite against 20 per cent of directors.
The companies that train their own people pull ahead
Companies that train their own people do well; those that lean on outside consultants come last of every capability approach, at 14 per cent measurable ROI. A separate study, the 2026 AI Enablement Services Buyer Survey from 10x Humans and AI Enablement Insider (100 senior buyers), found a training-led approach returned 73 per cent measurable ROI against 14 per cent for a consultant-only model, and that every buyer agreed tools alone are not enough without enablement. A quarter of leaders named skills gaps as their single biggest source of friction with AI, second only to security and compliance.
The adoption map: how wide, how deep
Leaders rated the AI maturity of five parts of their business, which separates two things usually blurred together: how many functions use AI, and how many have genuinely scaled it. Cross them and four kinds of company appear.
| No function scaled | At least one scaled | |
|---|---|---|
| Two or fewer functions using AI | Stalled, 21% ROI | Siloed, 49% ROI |
| Three or more functions using AI | Scattered, 57% ROI | Scaled, 62% (72% at two or more) |
Most companies are Stalled. Spreading AI across more functions pays better than going deep in only one, but the highest returns belong to companies that are broad and deep together. The most advanced form of depth is agentic AI, where measurable ROI climbs from 29 per cent for basic retrieval and single-task agents, to 56 per cent for agents that run a full workflow, to 83 per cent for agents that work across departments.
The picture by function
The pattern is the same in every function; the starting points are not. Technology and IT set the ceiling.
| Function | Measurable ROI | ROI at mature governance |
|---|---|---|
| AI infrastructure (technology, IT, data) | 58% | 87% |
| Efficiency (operations, finance, supply chain) | 46% | 96% |
| Strategy (C-suite) | 52% | 85% |
| Growth and customer (marketing, sales, CX) | 45% | 76% |
| Workforce (HR, people, L&D) | 46% | 82% |
Technology and IT are the most mature, with the highest ROI, the highest rate of embedded governance and the lowest shadow AI. Operations and finance are the most disciplined buyers and post the survey’s highest ceiling, 96 per cent measurable ROI among the most mature. Marketing, sales and customer experience are the most active and the least governed: the work lands where customers feel it, with 64 per cent reporting better conversion, but the return is hardest to prove. HR has the broadest deployment and the sharpest version of the policy trap.
Who is winning, and why
Beyond governance and capability, four traits travel with a stronger return. Each is a relationship in the data rather than a proven cause.
- Ownership structure. Private-equity-backed firms report 59 per cent measurable ROI, ahead of listed companies (48 per cent), independent private firms (44 per cent) and VC-backed ones (44 per cent).
- Size. Measurable ROI runs from 58 per cent in the £250m to £500m band down to 43 per cent in the £50m to £100m band.
- Growth. Among companies growing 20 per cent a year or more, 82 per cent report measurable AI ROI; among those in decline, none do.
- Sector. Technology and software (62 per cent) and financial services (59 per cent) lead; construction (31 per cent), transport (36 per cent) and retail (39 per cent) trail. Governance maturity follows the same order.
Failure is now normal
Deployment at this pace comes with breakage. 79 per cent of leaders have had at least one AI failure, and only one in five reports a clean record. The most common are a data, security or regulatory breach (39 per cent), a model or tool failure that broke a business process (37 per cent), and AI output used without a proper check that then caused errors (37 per cent). What separates companies here is not whether they fail but what they do next: the ones that treat a failure as something to fix and log keep moving, while stopping is what costs the rest.
About the research
The State of UK AI Adoption Survey 2026 gathered 755 responses from UK mid-market business leaders, fielded January to March 2026, in partnership with Executive Summary. More than three-quarters of respondents hold board-level authority over AI investment. Respondents are individual senior leaders in organisations with revenue between £50m and £500m, so figures describe the share of leaders in a given group. Executive Summary, which designed and ran the survey, is the research team behind Adobe’s annual Digital Trends Survey.
Measurable ROI means a leader reported that AI is delivering measurable business value or ROI today. ‘At scale’ means AI scaled across three or more functions with a measurable return, which describes 14 per cent of the sample. Where this page refers to how organisations buy AI enablement, it draws on a separate study, the 2026 AI Enablement Services Buyer Survey from 10x Humans and AI Enablement Insider (n=100), which is labelled wherever it appears and is not combined with the aibl data.
Frequently asked questions
How many UK companies get measurable ROI from AI?
Just under half. 49.6 per cent of UK mid-market leaders in the survey report measurable AI ROI today, and only 14 per cent have scaled AI across three or more functions with a measurable return.
What is the biggest predictor of AI ROI?
Governance maturity. The share of companies reporting measurable ROI rises from 22.2 per cent with no governance to 85.3 per cent with mature, embedded governance, on much the same tools. Capability is the second lever: governance and well-equipped people together reach 81 per cent.
How common is shadow AI in the UK mid-market?
55 per cent of leaders say unapproved or personal AI use is common or very common, driven mostly by official approval being too slow. It falls as governance matures, from 75 per cent in ungoverned companies to 35 per cent in the most mature.
Does training or hiring consultants get better AI results?
Training. Companies that build capability by training their own people outperform those that rely on outside consultants, which return the least of any approach at 14 per cent measurable ROI in the survey.
Source: aibl State of UK AI Adoption Survey 2026, 755 UK mid-market leaders, in partnership with Executive Summary. Free to quote and cite with attribution to aibl.