Why most UK firms admit they overstate their AI maturity, and what leaders can do about it
· 5 min read
Avanade's survey of 350 UK leaders finds 71% admit overstating their AI maturity and 61% see no justified return yet. Why it happens and five honest checks for leaders.
In brief
Many large UK organisations say they are well advanced with artificial intelligence, but a new survey suggests the reality is more modest and that many leaders know it. Research commissioned by Avanade, a company that sells AI and Microsoft services, asked 350 UK decision makers in big firms about their progress. Four in five described themselves as established, advanced or AI first. Fewer than one in five said their team had genuinely redesigned how it works. Seven in ten admitted their organisation presents itself as more mature than it really is, and six in ten said AI had not yet paid back the money spent on it. Most had never worked out that return with their finance team.
A separate survey of technology leaders by Accenture, also published this week, found AI is mostly helping individuals with single tasks, and that adoption and training are the biggest obstacles to doing more.
This article explains why organisations round up their AI story, why counting licences and pilots gives a misleading picture, and what leaders can do instead. It suggests five practical checks, from agreeing with finance how returns will be measured to testing public claims against evidence. It also looks at where leadership training helps, and is clear about the limits of vendor research and the points where specialist finance, data or legal advice is needed.
Most organisations would rather sound ahead on AI than behind. New research suggests many of them are saying so whether or not it is true, and that the people saying it know.
What the survey found
Avanade, the Microsoft services company, published its Build Beyond Limits: AI Reality Report on 3 November 2026. The research was carried out for Avanade by Vanson Bourne in August and September 2026, with 3,600 business decision-makers across 11 markets, all in organisations with more than $500 million in annual revenue. The UK sample was 350. Because Avanade sells AI and transformation services, treat this as vendor research: useful, but written by a company with an interest in the answer.
The UK figures are striking all the same.
- 80% describe their organisation as established, advanced or AI-first.
- Only 17% say their team has fundamentally redesigned its operating model around AI.
- 67% believe they are ahead of their peers.
- 71% admit their organisation portrays itself as more AI-mature than it really is.
- 61% say AI has yet to deliver financial returns that justify the investment.
- 65% have not formally calculated and reviewed the return on AI with finance or senior leadership.
- 42% have a documented AI strategy that is not consistently carried out.
Avanade's chief executive, Chris Howarth, summed it up as "a dangerous gap opening up between AI confidence and AI reality".
A second survey points the same way
Accenture's first CIO Pulse survey, also published on 3 November and also vendor research, asked 300 technology leaders in organisations with more than $1 billion in revenue where they stood. 70% said AI is mainly used for individual productivity, such as drafting and copilots on single desks. 53% named user adoption and workforce training as their single biggest barrier, the highest of seven tested. 54% named governance and responsible AI as their top barrier to scaling, yet only 32% rated their governance as advanced or leading.
Put the two together and a picture forms. AI is widely present. It is mostly helping individuals with tasks. Very few organisations have changed how work flows, and fewer still have checked whether the money is coming back.
Why organisations overstate
Nobody sets out to mislead. The pressure comes from every side. Boards want to hear progress. Clients want reassurance that they are buying from someone modern. Recruits ask about it. Competitors announce things. Each update rounds the story up a little, and nobody is asked to show the working.
The trouble is that the easiest things to count are the least meaningful. Licences bought, staff trained, pilots launched and prompts run all go up quickly. Ian Hughes is blunt about this in What Stays the Same: "Usage is an output. It tells you nothing." What matters is what changed as a result: a better decision, a faster resolution for a customer, a cost that genuinely fell.
Esm sees the same pattern through a service management lens. John Sansbury's The Journey to Service Excellence warns about dashboards that glow green while the people on the receiving end are unhappy, which he calls the Watermelon Effect: "Just because all the readings are positive does not guarantee fully satisfied consumers." An AI programme reported through licences and pilot counts is a watermelon waiting to be cut open.
What it means for leaders
Overstating maturity is not just a reputational risk. It changes decisions. If the board believes the organisation is already advanced, it funds the next shiny project rather than the unglamorous work of redesigning a process, cleaning the data or training managers to lead differently. If staff hear a story that does not match their day, they stop trusting the next one. And if a client is told AI is improving their service, they will eventually look for the evidence.
Practical steps
- Pick one end-to-end piece of work, such as handling a customer complaint or onboarding a new starter, and describe honestly how much of it has actually changed. One true example beats ten claimed ones.
- Agree with finance how the return on AI will be measured, and separate cash saved from time freed. Time freed is only worth something if it goes somewhere useful.
- Ask the people doing the work how they would describe progress, before the next board update rather than after it.
- Replace activity measures with outcome measures in your reporting. Keep usage as context, not as the headline.
- Review public claims, on your website, in bids and in recruitment material, against evidence you could show on request.
Where training fits
Both surveys point to people rather than technology as the sticking point. Accenture's respondents put adoption and training top of their barriers. Avanade's found strategies written down but not carried out. That is a leadership capability gap: knowing what good looks like, asking the right questions of a pilot, and measuring outcomes instead of activity. Training helps when it is built around real work and gives managers the confidence to report honestly, including when the answer is "not yet".
Honest limits
Both studies are vendor research from firms that sell AI services, and both surveyed only large organisations. Self-reported maturity is, by definition, a matter of opinion. Neither survey tells you which use cases work. Calculating the return on AI properly may need finance and data expertise, and claims made to customers or regulators may need legal advice. A clearer story is a start, not a substitute for that work.
The organisations that will be believed in two years are the ones that describe their AI progress accurately today, even when the accurate version is less impressive.
Tags: AI adoption, AI maturity, leadership, measurement, outcomes
Sources
Join the Exchange free
Join free to unlock the complete knowledge library.
Founder membership is free for life
More from Esm
I have ITIL 5 Foundation. Which certificate should I take next to move into service delivery?
4 min read
How do I stop routing every ticket by hand when IT and HR share one queue?
5 min read
Are We Interviewing a Person or ChatGPT? What CMI's Hiring Poll Means for Leaders
5 min read
How should IT prepare for staff using their own AI agents at work?
4 min read
When staff stop raising AI concerns: what Proton's cancelled re-interviews tell leaders
5 min read
How should IT handle old SharePoint data when every department wants to keep everything?
4 min read
Related articles
I have ITIL 5 Foundation. Which certificate should I take next to move into service delivery?
Which ITIL certificate to take after Foundation (Version 5) if you want to move from technical support into service delivery or people management.
How do I stop routing every ticket by hand when IT and HR share one queue?
How to separate requests from incidents, keep categories short and owned, and correct them at closure so a shared IT and HR queue routes itself.
Are We Interviewing a Person or ChatGPT? What CMI's Hiring Poll Means for Leaders
UK hiring managers suspect candidates are using AI in interviews, yet want employers to disclose their own AI use. Here is how leaders can make honest AI use visible on both sides.
How should IT prepare for staff using their own AI agents at work?
Personal AI agents act inside real systems, so plan for what they can change: find what is in use, control consent, name an owner and log their actions.
When staff stop raising AI concerns: what Proton's cancelled re-interviews tell leaders
Proton cancelled AI re-interviews after reports staff felt unable to raise concerns; new research shows the same quiet gap between managers and teams elsewhere.
How should IT handle old SharePoint data when every department wants to keep everything?
Treat old data as a business decision IT runs as a service: one policy owner, agreed retention periods, clear archive tiers and automated rules.