Artificial intelligence use in UK businesses has almost tripled in under three years. Office for National Statistics figures published in July 2026 show that around 35 per cent of UK businesses with ten or more employees were using AI by June 2026, up from roughly 12 per cent in late 2023. The same data tells a second, quieter story. Only 10 per cent describe their use as extensive, and just 15 per cent say more than half their workforce uses AI daily. In other words, many more businesses now use AI, but most use it in one or two narrow ways rather than across the business.
Larger organisations see the same pattern from the other end. A BCG survey of 152 chief executives at companies turning over 500 million dollars or more, also published in July 2026, found that 89 per cent were seeing cost or revenue benefits from AI in specific areas, while 74 per cent were struggling to scale that impact across the business. Their conclusion was blunt. The core problem is not technology. It is execution.
That gap is why a group of UK companies, government departments and AI firms spent the early part of 2026 building something new. It is called AI for Growth, and Lumi is one of its partners.
What AI for Growth is
AI for Growth is a voluntary cross-sector alliance set up to help UK businesses adopt AI in a way that is practical, secure and affordable. It was launched by Founders Forum Group and Accenture. Core members joined in February 2026, and the alliance launched publicly at the Reinvention X conference in April 2026. It is supported by the Department for Science, Innovation and Technology, with backing from Kanishka Narayan MP, the minister responsible for AI.
The alliance exists because most AI support on the market was built for large enterprises with their own technical teams. For the businesses that make up the bulk of the UK economy, that support is often too technical, too expensive, or simply the wrong shape.
Membership is contribution-led. Partners commit time, expertise, networks or funding rather than paying a fee and receiving a logo. Alongside Lumi, the partner list includes Accenture, BT, Multiverse, ElevenLabs, Synthesia, Quantexa, Faculty, Upskill Universe, Funding Circle, Ollo, CyberStaff, Fifty One Degrees, A-LIGN and smartR AI.
Three priorities, and why reskilling leads
The alliance organises its work around three national priorities: reskilling, security and infrastructure. The security strand has produced an AI Security Healthcheck with guidance written for specific sectors. The infrastructure strand has produced a procurement guide covering compute, energy, data residency and the risk of being locked into a single supplier. Both answer questions a business only asks once it has already started adopting AI.
Reskilling comes first for a simple reason. The other two are decisions. Reskilling is a change in how people work, and that takes longer than any procurement cycle.
The chief executives in the BCG survey said as much. Fifty-five per cent named redesigning how people work as a barrier to scaling AI, yet only 30 per cent involved HR in how AI decisions were governed, against 82 per cent who involved technology teams. The people question is widely recognised and thinly staffed.
The ONS data makes the same case from the other direction. Among businesses that named a barrier to adoption, a lack of expertise was the one they named most often, ahead of cost. Around 62 per cent of those businesses said they were investing in training their existing staff, yet only 11 per cent could say that more than half their workforce had received any AI training at all. The intent is there. The coverage is not.
The strand supports the UK target of helping 10 million workers build AI skills, and has already produced a free peer community for business owners, a Reskilling Navigator built with ElevenLabs, and an AI Strategy Generator. The wider ambition is to reach more than 100,000 UK small and medium-sized businesses, with members aiming for a 25 per cent improvement in productivity.
The market has already reacted
Demand at the entry level is moving faster than most hiring plans. In the United States, the National Association of Colleges and Employers found in its spring 2026 survey that more than a third of entry-level jobs now require AI skills, up from around 11 per cent in autumn 2025. That is close to a tripling in six months.
Supply has moved just as fast. Revelio Labs, a workforce data firm quoted by Bloomberg, found that AI made up about 2 per cent of the professional certificate market in 2022, before ChatGPT launched. By 2026, short AI programmes, often sold as boot camps, accounted for close to a third of it. Policymakers in Washington are encouraging schools and companies to run more of them, some employers are paying for staff to attend, and universities are leaning on short credentials to shore up falling revenue.
This is US data, so treat it as a leading indicator rather than a description of the UK. The direction is the point. Where there is anxiety and a clear hiring signal, supply arrives quickly, and volume is far easier to produce than quality.
The part that is genuinely hard
Delivering AI training is easier than proving it has been effective.
Most programmes measure what is easy to count: how many people attended, how many finished, and how they felt about it afterwards. None of that tells a board whether the way work gets done has changed. It is why so many organisations can describe their AI training in detail and say almost nothing about what it produced.
This is the problem Lumi was built around, and it shapes how a sprint works. Teams do not practise on an invented exercise. They take a live business problem and work it through a series of facilitated sessions using approved AI tools, so what gets assessed is real work rather than a classroom exercise. Their work is assessed while it is in progress, against a defined skills framework, so the record is built from what people produced rather than what they said about themselves afterwards. Each person leaves with a Skills Passport holding that evidence.
That produces a different kind of answer to the question of what an organisation got for its money.
Lumi is the first platform accredited by TechSkills, a techUK company, to build human and AI collaboration skills.
What members of AI for Growth do
In a crowded market, it is fair to ask who is worth listening to. One useful test is who is prepared to do the unpaid work.
Each AI for Growth priority has a working group, co-chaired by member organisations and coordinated by programme director Alessandra Corti. Decisions are made collectively. As the alliance puts it, the organisations that show up and contribute are the ones that shape what gets built. Nobody is buying visibility. The companies involved are giving away thinking they would normally charge for, so the tools and guidance reach businesses that would otherwise go without.
Selling AI upskilling for business is one thing. Helping to set the national standard for it, in a group that includes government, the frontier AI companies and one of the largest consultancies in the world, is a different kind of commitment.
Three questions worth asking any upskilling partner
What will I have at the end that I can show a board? A completion certificate is not evidence. A solved problem, a costed plan and a record of demonstrated skill are.
Whose problem gets solved while the programme runs? If the answer is nobody’s, the programme is a cost. If teams are working on live business issues, the learning and the return arrive together.
Who checks the standard? Independent accreditation and open participation in bodies such as AI for Growth tell you a supplier is willing to be measured by someone other than itself.
The takeaway
AI adoption in the UK is not held back by a shortage of tools, or now by a shortage of courses. It is held back by a shortage of people who are confident applying AI to real work, and by a shortage of evidence that anything improved. AI for Growth exists to close both gaps, and it has put reskilling at the top of the list for exactly that reason.
Lumi joined it because that is the same problem Lumi spends every day on. If you are weighing up how to build AI capability across your own organisation, the useful place to start is not the tool list. It is the question of what you want to be able to prove in six months, and who is willing to be measured on it with you.