Ready for AI

Key facts

• None

  The AI plan for Wales does not yet have a budget of its own, and no spending has yet been reported from the Welsh Government’s £2.1m package to help small businesses adopt AI.

• 2p

  Of every £1 of public money awarded for AI work in Wales since April 2022, about 2p has gone to helping businesses adopt AI.

• 99.3%

  Small and medium-sized businesses make up 99.3% of businesses in Wales, and they employ 62.3% of the people who work in Welsh businesses.

• 24.8% against 34.9%

  In 2024, 24.8% of Welsh companies with ten or more employees that used digital production technologies were using AI, the lowest share of the four UK nations, against 34.9% in England.

• About four times

  Northern Ireland spends about four times as much per person as Wales has committed on hands-on help for businesses to adopt AI.

• 6p

  For every £1 of public money Wales spent building superfast broadband, it spent about 6p helping businesses to use it.

• £1.1bn

  Wales could gain about £1.1bn of extra output over the ten years to 2035 if its rate of AI adoption catches up with the UK’s by 2030.

• £4.5m a year

  Matching Northern Ireland’s rate of spending per person would cost Wales about £4.5m a year, about 20% of the core Business Wales budget.

Introduction

1This is an initial report on how Wales is helping its small businesses to adopt artificial intelligence (AI). It brings together what has been published about the AI plan for Wales and the public money committed to AI. It also looks at where that money has gone, how Wales compares with its neighbours and what earlier programmes have taught. It is offered as a first account of the evidence, to inform discussion and further work.

2Wales starts from a good position. It has an AI plan, an Office for AI and two AI Growth Zones.^{1, 2, 3} Business Wales is running free AI events for businesses, delivered across Wales by M-SParc, Bangor University’s science park on Anglesey, and a training subsidy pays three quarters of the cost of AI courses.^{4, 5} Welsh small companies have won millions of pounds from Innovate UK to build AI products of their own.⁶

3The evidence also shows gaps. The AI plan does not yet have a budget, targets or timeline.¹ No spending has yet been reported from the Welsh Government’s £2.1m package to help small businesses adopt AI.⁵ Since April 2022 about 2p in every £1 of public money awarded for AI work in Wales has gone to helping businesses adopt it.⁷

4The timing is relevant. A new government took office in May 2026, and its minister responsible for AI has promised a more coordinated national approach.^{8, 9} Its first draft budget, for 2027-28, is due on 17 November 2026, when the outline and detailed plans will be published together, with the Final Budget due on 16 February 2027.^{10, 11, 12} The findings may be of use to those preparing and scrutinising that budget and the ones that follow. The Senedd’s Economy, Energy and Connectivity Committee has already asked the Cabinet Minister for Enterprise, Connectivity and Energy to set out, alongside the draft budget, the allocations to Business Wales, how the government’s support for businesses assists small and medium-sized businesses, and support to grow key technology sectors such as AI.¹³

5The study is based on a review of policy documents, budget tables, Senedd proceedings, grant and contract records, and evaluations of earlier programmes. Public money for AI in Wales from April 2022 to mid-October 2026 was brought together from published grants, contracts and payments, and an economic model was built of what AI adoption could add to Welsh output. Every figure was checked against its original source, and Appendix A describes the method.

6There are limits to the evidence. No Freedom of Information requests were sent and no businesses were surveyed, and the research did not test AI tools or assess the quality of the support that Business Wales provides. Where this report says that something does not exist, it means that it could not be found in published sources. The estimate of the economic gain from AI is indicative and is not a forecast.

7The report is in seven sections. Sections 1 to 3 describe the plan, the money and how Wales compares with its neighbours. Section 4 looks back at earlier programmes, Section 5 at the risks and Section 6 at what closing the adoption gap could be worth. Section 7 sets out the issues the evidence raises and the further work needed to answer them.

Findings

8The report makes the following findings.

Finding 1 Wales has an AI plan and has begun to deliver it, but the plan has no budget of its own.

• The AI plan for Wales has no budget, targets or timeline, and it says that achieving its vision “will require significant investment and prioritisation”.
• Reviews and a Senedd committee called for clearer support for small companies, and the government accepted the committee’s recommendations at least in principle, recording their financial implications as “None” or “Yet to be identified”.
• No spending has yet been reported from the £2.1m package for small businesses, and the only money that any Welsh budget labels as AI, £2.5m a year, is for the Office for AI and public services.

Finding 2 Little of the public money for AI in Wales has gone to helping the small businesses that make up most of the Welsh economy.

• Small and medium-sized businesses make up 99.3% of businesses in Wales and employ about 715,000 people, 62.3% of those working in Welsh businesses.
• Of £30.1m of public money awarded for AI work in Wales since April 2022, 2.3% went to helping businesses adopt AI, while 82.5% went to research and development and 13.4% to public bodies’ own use.
• The AI Growth Zones bring billions of pounds of announced private investment, and the money for local adoption, up to £5m per zone, is awaiting a published scheme.
• Larger Welsh companies use AI less than companies in any other UK nation, and the smallest are measured only by samples too small to be reliable.

Finding 3 Wales’s neighbours spend more per person on helping businesses adopt AI, and Wales wins less than its share of UK funding.

• Northern Ireland spends about £1.43 per person a year on hands-on AI help for businesses, roughly four times the £0.35 that Wales has committed on the most favourable reading.
• Scotland, Ireland and the Basque Country fund programmes that go further, and Wales won 2.0% of Innovate UK’s AI competition money against 3.4% of UK businesses.

Finding 4 Earlier Welsh programmes reached many small companies but rarely showed lasting gains.

• Opportunity Wales reached nearly 12,000 small companies through local partners, and Superfast Business Wales supported 6,537 companies, 91% of them microbusinesses.
• For every £1 spent building superfast broadband, Wales spent about 6p helping businesses to use it, and earlier programmes stopped when their funding stopped.

Finding 5 The risks of AI fall unevenly, and hardest on rural areas, Welsh speakers and the smallest companies.

• About 45,000 jobs could be displaced by 2035 on the current path, and places with weaker connectivity, Welsh speakers and the smallest companies are most at risk of being left behind.
• Most of the modelled displacement happens on the current path, whatever support is offered.

Finding 6 If Wales catches up with the UK’s rate of AI adoption by 2030, it could gain about £1.1bn of extra output over ten years.

• The gain rises to about £326m a year by 2035, small and medium-sized companies would receive about 37% of the whole gain from AI, and this report shows every step of the calculation.

Issues for consideration

9The findings raise five issues for the Welsh Government, Business Wales, councils and regional bodies, and the UK Government. They are set out as questions for discussion rather than recommendations, with the evidence that bears on each.

Issue 1 Whether the AI plan should have a dedicated, multi-year budget for small business AI adoption.

• The 2027-28 draft budget, due on 17 November 2026, and the supplementary budgets that follow are the next opportunities to decide this.
• For scale, matching Northern Ireland’s rate of spending per person would cost about £4.5m a year, or about £13.5m over three years.

Issue 2 How support could reach microbusinesses and sole traders, who make up most Welsh businesses.

• The Business Support Review recommended a clear pathway for small companies to adopt AI, and no part of the current offer is designed around the smallest companies.
• Ireland and the Basque Country offer routes open to very small companies and some self-employed people.

Issue 3 How far support could be led from Wales, by advisers, colleges, universities and companies based in Wales, in Welsh and English.

• Past programmes delivered through local partners reached the most companies, and the Welsh Language Commissioner advises testing AI tools in Welsh before they are used.

Issue 4 Whether council business grant rules should allow AI software, subscriptions and training.

• Local Growth Fund business grants in at least seven council areas, in north and south Wales, exclude or restrict these costs, although most AI tools are sold as subscriptions.
• A change would need no new money, but would need to guard against paying for subscriptions that companies would have bought anyway.

Issue 5 How the AI Growth Zone money for local adoption will reach local small companies.

• No scheme, delivery body or amount reserved for small companies had been published by late October 2026, and the UK Budget of 28 October 2026 did not mention it, so businesses cannot yet see what will be available to them.

Section 01

Wales has an AI plan and has begun to deliver it, but the plan has no budget of its own

The plan and its aim

10The AI plan for Wales, “AI Cymru”, was published by the previous Labour government on 18 November 2025.¹ Its aim for business is to “support SMEs, including social enterprises, and key industries to adopt AI and compete globally”.¹

11The plan states what it needs. It says that “some of the immediate actions within the plan are funded” but that “achievement of the vision will require significant investment and prioritisation”.¹ It sets out no budget, quantified targets, timeline or monitoring framework, and it does not mention microbusinesses, sole traders or the self-employed.¹

12The advice behind the plan reached similar conclusions. The government said that it fully accepted the findings of its expert review of AI.¹⁴ The only published account of that review, in the Business Support Review of November 2025, found business AI support “fragmented and not always tailored to AI”, and the Business Support Review recommended “a clear pathway including dedicated support, for SMEs to adopt AI”.¹⁴

13The Senedd has asked for something similar. Its only committee report on AI and the economy, published in May 2025, recommended a review of business AI support and an analysis of the groups most at risk of losing work to AI.¹⁵ The government accepted all eight recommendations at least in principle, but recorded the financial implications of each as “None” or “Yet to be identified”.¹⁶ In late October 2026 the current Senedd’s Economy, Energy and Connectivity Committee had no inquiry into AI under way.¹⁷ The plan, the reviews and the Senedd describe similar needs, but none of them attaches money to meeting them.

The first steps, and the money behind them

14On 12 January 2026 the Welsh Government announced £2.1m to help small and medium-sized businesses, entrepreneurs and microbusinesses adopt AI.⁵ It comprised £600,000 for a Business Wales AI awareness and adoption programme, £500,000 for AI support in tourism and events, and £1m for subsidised AI training through the Flexible Skills Programme.⁵

“[This] will help us ensure SMEs across all sectors in Wales are prepared to optimise the potential of emerging AI technologies in responsible, inclusive and ambitious ways.”

15Delivery has begun. M-SParc is delivering the Business Wales programme across Wales, with advisers in post until 31 March 2027.⁴ In early October 2026 the Business Wales listing showed 22 free AI events for the following three months, including two-day accelerators open to sole traders. At the end of October it showed 16 free AI sessions for November and December, nine of them in person.¹⁸ AI workshops for tourism businesses run by MWT Cymru filled quickly, and further workshops funded by the Welsh Government continue to 24 November 2026.^{19, 20} The training subsidy has been extended to the end of 2026-27.^{21, 22} These are the first practical steps under the plan.

16Published detail on these first steps is still limited. No budget line, contract value, targets, take-up or spending has been published for the Business Wales programme, and at the end of October 2026 the Business Wales page on AI still said that Business Wales “is developing an AI awareness and adoption support programme”, although delivery had begun.^{4, 23} The announcement of the £500,000 for tourism and events describes an AI Pollination Project with the Hartree Centre Cardiff Hub and in-person workshops for up to 1,000 tourism businesses, but no source says how much of that money has been spent or how many businesses have taken part.^{5, 19, 20} The only reported result is the minister’s statement, quoted by Business News Wales, that businesses taking part in the tourism workshops saw on average a 64% increase in website traffic. A Visit Wales page gives a 63% increase in organic website traffic, and neither source says how many businesses the figure covers or how the increase was measured.²⁰ No take-up figures for the training subsidy had been given to the Senedd by late October 2026, and the minister gave none when he referred to the AI skills pilot in plenary on 21 October.^{22, 24} Only the £1m for training was ever set aside in a budget, and only for 2025-26.^{25, 7} More than nine months after the package was announced, no spending from it has been reported.

The table draws on the Welsh Government’s announcement of 12 January 2026, its answers to Senedd written questions, the M-SParc job description and reports of the tourism workshops.^{5, 4, 25, 21, 22, 20}

17The only money that any Welsh Government budget document labels as AI is £2.5m a year in 2025-26 and 2026-27 for the Office for AI and a fund for AI in public services, and no budget line from 2022-23 to 2026-27 is named for AI, including in the second supplementary budget for 2026-27 laid in October 2026.^{2, 26} So far the only AI money in the Welsh budget is for the government’s own work, and there is no budget line for helping businesses adopt AI.

A new government

18Plaid Cymru took office in May 2026.⁸ Adam Price MS, Cabinet Minister for Enterprise, Connectivity and Energy, is responsible both for policy on AI and for Business Wales.²⁷ The Programme for Government 2026 to 2030 commits to a new Digital Strategy to “deliver the conditions to support safe and ethical adoption of AI to support productivity and more efficient public services”, but it does not mention the AI plan, the £2.1m package, microbusinesses or sole traders.^{28, 29} In a written answer to a Senedd question in October 2026, the First Minister said that work was under way to scope the strategy’s themes and priorities, and neither that answer nor a later one from the minister gave a publication date.³⁰ The government’s goal of halving the productivity gap with the UK within ten years does not mention AI,³¹ although Section 6 suggests that AI adoption could make a measurable contribution to Welsh output.

19The new government has signalled an interest in AI. The minister’s evidence paper to the Senedd’s Economy, Energy and Connectivity Committee, dated 30 September 2026, promises a more coordinated national approach to AI.⁹ The record of the committee session that day does not mention AI by name. Data centres were raised in questions on the electricity grid and land use, and no member asked about support for small businesses to adopt AI.³² As a backbench Member in September 2025, Adam Price, repeating evidence from the company Amplyfi, called for an AI innovation fund, although no such fund has yet been announced.³³ In a talk at Harvard in October 2026 he said that, for Wales, the race to build the most advanced AI models is “the wrong race to watch”, and he stressed the adoption of AI.^{34, 35} On 13 October 2026 two Members asked the government for statements on AI, and the Trefnydd, the minister responsible for government business, said in Welsh that she would raise their requests so that Members were kept updated. No statement on AI was on the plenary agendas of 20 and 21 October.^{36, 24} On 21 October the minister told the Senedd that a national AI governance framework was being drawn up, and he pointed to personal learning accounts and “our AI skills pilot” as the government’s investment in digital and AI skills, without giving budget or take-up figures.²⁴

“…one of the key areas is that we need access to capital, so we need an AI innovation fund that is focused… on where our potential is, but also where our missions are.”

20The new government’s first supplementary budget contained no money for AI and was defeated on 14 July 2026, so the 2027-28 draft budget will be its first full budget.^{37, 10} A second supplementary budget, laid on 13 October 2026 after an agreement with Welsh Labour on £60m for additional learning needs, also contains no money for AI and leaves the Business Wales budget line unchanged. It is due to be debated on 3 November 2026.^{37, 26, 38} The first supplementary budget was defeated and the second followed an agreement with another party, which suggests that the 2027-28 budget will also need the support of at least one other party.

21One feature of the Welsh arrangements is clear ownership. Carnegie UK has warned that at UK level there is “a risk that AI becomes everybody’s priority and nobody’s clear responsibility”.³⁹ In Wales, one minister holds both AI policy and Business Wales.²⁷ The government is also designing a new national development agency. The full membership of its expert advisory panel was appointed in September 2026, and the minister expects to update the Senedd early in 2027, after the panel’s interim report. Nothing published so far says whether Business Wales or support for AI adoption will form part of the agency.⁴⁰ In September 2026 the Senedd’s Economy, Energy and Connectivity Committee began an initial inquiry into the agency, which seeks greater clarity on the government’s thinking about matters that include its remit, budget and powers.¹⁷ The plan has an owner and an aim, but not yet a budget.

Section 02

Small businesses make up most of the Welsh economy, but little of the public money for AI has gone to helping them adopt it

Small businesses are the Welsh economy

22Wales is an economy of small companies. Of the 246,500 businesses active in Wales in 2025, 99.3% had fewer than 250 employees, and together they employed 714,900 people, 62.3% of everyone working in Welsh businesses. Microbusinesses, with fewer than ten employees, were 94.6% of all businesses.⁴¹

The figures are for businesses active in Wales in 2025, from the Welsh Government’s size analysis of active businesses.⁴¹ The rows do not add exactly to the published totals because of rounding, and the shares for all small and medium-sized and for large businesses are the published figures.

23The smallest companies matter most outside the cities. Microbusinesses provide 52.6% of employment in Ceredigion and 51.6% on Anglesey, against 24.3% in Cardiff.^{41, 42} Support that does not reach microbusinesses and sole traders would miss most Welsh businesses and much of the rural economy.

Where the money has gone

24Every item of public money for AI in Wales that could be found in published grants, contracts and payments from April 2022 to mid-October 2026 was brought together.^{7, 43, 44} Of £30.1m of grants and contracts that were clearly for AI, £702,123 went to helping businesses adopt it. That is 2.3%, or about 2p in every £1. Research and development took 82.5%, and public bodies’ own use of AI took 13.4%.^{7, 43, 44}

The table covers grants and contracts that are clearly for AI, awarded in Wales from April 2022 to mid-October 2026, and includes the awards added in Innovate UK’s records of 12 October 2026 and a UKRI contract with Redpen AI of Newtown.^{7, 43, 44} AI skills includes AI Workforce Bridge Wales, one of five Welsh projects named in October 2026 as recipients of the UK TechFirst Local Fund. Two others, Tech Talent Cymru and VIBE Amplified, cover AI and cyber security together, so they are left out of the main figures.^{43, 45} If one project that tests AI in a company’s own production is counted as research rather than adoption, the share for helping businesses adopt AI is 2.1%.⁴³

25Research matters, and Welsh small companies have won about £9.67m from Innovate UK to build AI products.^{6, 43} Spending on adoption has been much smaller. Spread across the 246,500 businesses in Wales, the money awarded to help businesses adopt AI comes to about £2.85 per business over four and a half years.^{7, 43, 41}

26The adoption money was also not designed for the smallest companies. It went to Innovate UK projects worth £602,123 and to the Cardiff Capital Region’s AI Enable programme, a £100,000 contract for May 2026 to May 2027.^{6, 43, 46} A Tramshed Tech article describes that programme as focused on “SME leaders with teams of 10-50 employees”,⁴⁷ but the programme page describes it as for “Ambitious SME leaders from any industry in the Cardiff Capital Region”, the contract sets no size limit, and in late October 2026 the page said the programme was “not live yet”.^{46, 48} It serves one region and has no route designed for microbusinesses and sole traders. The UK’s newest adoption competition, BridgeAI Deployment Catalyst, opened on 21 October 2026 with up to £5m for its first phase, and Innovate UK expects up to £52m across three phases, subject to approval. The organisation adopting the AI must be a medium-sized or large business or a public body.⁴⁹

Data centres and the AI Growth Zones

27The two Welsh AI Growth Zones come with billions of pounds of announced private investment in data centres, and with the promise, in the UK Government’s words, of more than 8,000 jobs.^{3, 50, 51}

“In North Wales, that means thousands of new roles and £5 million to boost skills and help local businesses adopt AI.”

28A data centre will not, on its own, show a local company how to use AI. That is the purpose of the local adoption money, up to £5m for each Welsh zone “for business adoption and skills”, although its stated purposes also include research and start-ups.^{3, 50, 52} No scheme, delivery body or award had been published by late October 2026, and the UK Budget of 28 October 2026 did not mention it.^{52, 53, 54} The functions of the UK department that announced the money have since been redistributed, and in October 2026 the UK Government confirmed that the Department for Business, Innovation, Science and Trade now leads on AI Growth Zones and on economy-wide AI adoption.^{55, 56} The £10m for the two Welsh zones is most of the £11.1m announced for helping businesses in Wales adopt AI, and it does not yet have a published route to small companies.⁷ The same Budget introduced a Data Centre Covenant, under which strategic data centres seeking enhanced government support should agree a Local Benefit Plan and create jobs, apprenticeships and training places, although it does not mention the AI Growth Zones.⁵⁴

The current offer and the smallest companies

29Microbusinesses are named in the January 2026 package, and the Business Wales accelerators are open to sole traders.^{5, 18} However, no part of the package has its own design, eligibility or targets for these companies. The training subsidy’s 25% employer contribution fits sole traders poorly, and Personal Learning Accounts, which are open to the self-employed, offer no dedicated AI option.^{57, 58}

30Several council business grants funded by the Local Growth Fund exclude what a microbusiness would buy to adopt AI. Bridgend’s grant excludes “licence fees and subscriptions”, Rhondda Cynon Taf’s excludes “training, software, subscriptions”, and Pembrokeshire and Carmarthenshire allow specialist software only where it is linked to capital spending.^{59, 60, 61, 62} In Denbighshire and Flintshire, the grant run by Cadwyn Clwyd treats digital adoption as capital spending and excludes “ongoing subscriptions, hosting, maintenance or support costs” from software for managing stock, and Blaenau Gwent allows essential software licences only in exceptional circumstances, where they are linked to funded equipment.^{63, 64} Most AI tools are sold as subscriptions, so these rules affect AI adoption directly. The Pembrokeshire grants are now on hold because they were oversubscribed, and Carmarthenshire’s has closed to new expressions of interest, so the rules will matter most when the next rounds are designed.^{61, 62}

Welsh companies already lag, and the smallest are measured only by small samples

31Larger Welsh companies already use AI less than companies elsewhere in the UK. In 2024, 24.8% of Welsh companies with ten or more employees that used digital production technologies were using AI, against 34.9% in England, 26.2% in Scotland and 25.1% in Northern Ireland.⁶⁵

The figures cover companies with ten or more employees that used at least one listed digital production technology, and they come from the UK Innovation Survey 2025.⁶⁵

32Use of AI has risen quickly since 2024, and by September 2026, 38% of UK businesses with ten or more employees used it, so these levels are out of date, but the gap between Wales and England is the more lasting finding.^{66, 67} That gap lies among the companies that innovate. Welsh companies that innovate used AI far less than their English counterparts (26.3% against 41.8%), while companies that do not innovate were close to level (23.5% against 24.4%).⁶⁵ Welsh companies are less likely to put AI to serious use, not only less likely to try it, which suggests that awareness on its own may not close the gap.

33For the smallest companies the only Welsh measures rest on small samples. About half of Welsh private sector businesses are outside the register from which most official business surveys are drawn.⁶⁸ A UK Government survey of all businesses, sole traders included, carried out in late 2025, put AI use in Wales at 34.6% against 34.9% for the UK, but with a margin so wide (19% to 50%) that it can neither confirm nor rule out a gap.⁶⁹ The Longitudinal Small Business Survey, which asked no question about AI in 2024, did so in 2025, for one part of its sample.^{70, 71} Of small and medium-sized employers that used any technology, 28% in Wales used AI, the lowest of the four nations, against 35% in England and 34% across the UK, although the Welsh sample was 150 businesses and Scotland’s figure was only one point higher.⁷¹ Among businesses with no employees that used technology, the Welsh figure was 32% against 42% for the UK, on only 32 Welsh responses.⁷² Private surveys do not fill the gap. NatWest’s September 2026 figures for Wales rest on 59 companies that already used AI, so they cannot measure adoption, although the same survey found that a company’s headcount explains far more of its AI use than its region.⁷³ A nation whose economy is built on small companies cannot yet see clearly how those companies are using AI.

“If AI becomes something only larger firms feel able to adopt with confidence, we risk widening the gap between businesses with the time, money and legal support to experiment, and those already stretched by day-to-day pressures.”

Section 03

Wales’s neighbours spend more per person, and Wales wins less than its share of UK funding

34Other small nations and regions have gone further than Wales in helping small companies adopt AI. Their schemes differ, but they share three features that the current Welsh offer does not yet have. They pay for expert help or for putting AI to work, not only for awareness. They are open to the smallest companies. They also measure what happens.

35Northern Ireland has a £16.3m AI Collaboration Centre that gives eligible businesses fully funded, hands-on help from AI specialists, working with Ulster University and Queen’s University Belfast. Demand has exceeded places, and for that reason applications for its autumn cohort have now closed. The centre publishes an annual census of AI capability, and its 2026 census found that around 45% of Northern Ireland businesses were still not using AI.^{75, 76, 77} Invest Northern Ireland’s share of the cost is £13.8m over five years, about £2.76m a year, which is about £1.43 per person each year.^{75, 78}

36On the most favourable reading for Wales, which assumes that the £1.1m announced for the Business Wales programme and tourism support is spent in a single year, Wales has committed about £0.35 per person, and Northern Ireland spends roughly four times as much. The comparison is indicative, because the schemes differ in period, scope and design.

The figures are calculated from programme budgets and official population estimates, with euro amounts converted at about 85p to the euro.^{5, 79, 75, 80, 78} The Basque scheme is a large grant scheme for industry and is shown for scale only. The call’s published rules set €11m, and the Basque Government and its business development agency now give €14m.^{80, 81} The Welsh figure leaves out the £1m training subsidy, which pays for courses rather than direct help to companies.

37Scotland is in the second funded year of a national programme to help small and medium-sized companies adopt AI, with £1m in 2025-26 and £1.8m in 2026-27, delivered with partners that include the rural enterprise agencies.^{79, 82} The Scottish Government says the first year helped more than 1,000 businesses, and its AI strategy reports more than 160 companies receiving hands-on assistance, mentoring or exploration funding, and says that dedicated rural activity reached businesses in remote areas.^{79, 83} The programme runs to 31 March 2027, and in the south of Scotland it includes AI adoption grants of £1,000 to £5,000.^{84, 85} Wales has committed about as much per person, but Scotland has a published budget for a programme in its second year, while the Welsh money does not yet have a budget line or a stated period.⁷⁹

38In Ireland, companies with between one and 50 paid employees, including self-employed people who employ staff, can receive consultancy and then a voucher worth 50% of costs up to €5,000, which can pay for off-the-shelf AI software, and the country has a national target of 75% take-up of AI, cloud computing and data analytics by 2030.^{86, 87} Take-up was slow at first. In May 2025 Digital Business Ireland reported a Parliamentary answer showing only 13 applications and six approvals since the voucher’s launch in September 2024.⁸⁸ Ireland’s Budget for 2027, announced on 6 October 2026, set aside €150m over three years from the National Training Fund to train people in AI skills.^{89, 90} In the Basque Country, companies can apply until 2 November 2026 for grants of 60% of the cost of AI projects up to €100,000, and live online training in digital skills is offered to microbusinesses and the self-employed.^{80, 81, 91} Wales does not yet have such a grant.

39Wales also wins less than its share of UK funding for AI in business. Wales has 3.4% of UK private sector businesses, and 3.8% of the businesses registered for VAT or PAYE,^{68, 92} but it received 2.0% of the money in Innovate UK competitions dedicated to AI and 2.4% of the money from Innovate UK’s BridgeAI programme.^{6, 43} Helping more Welsh companies to bid could raise that share. In June 2026 the UK Government announced £100m to expand BridgeAI, as part of £200m to help businesses adopt AI, so the share that Wales wins matters more.⁹³ The UK Government is also starting to fund local delivery in England. Its Budget of 28 October 2026 set aside money to pilot local partnerships for AI adoption, starting with three combined authorities in Yorkshire that will work across councils to help small businesses grow. The Budget text gives £30m for the pilots, while its costings table gives £15m over 2026-27 to 2029-30, and it does not say whether Wales will receive any equivalent funding.⁵⁴

40Until this year, the only fully funded technical help for small companies adopting AI in Wales came from the Hartree Centre Cardiff Hub, which had assisted 70 small and medium-sized companies by September 2025 on grants that ran to the summer of 2026.^{94, 95, 2} It shows that Welsh universities can offer this kind of help, but it served one part of Wales. Wales as a whole offers awareness events and a training subsidy, but no grant to put AI to work and no route designed for microbusinesses.

Section 04

Wales’s earlier programmes offer lessons on what works

41Wales has tried before to help small companies adopt new technology, and its record contains both real successes and clear warnings. Table 6 summarises the main programmes.

The cost per company is calculated from the sources cited.^{96, 97, 98, 99, 100, 101, 102, 103, 104, 105}

What worked

42Welsh programmes have reached small companies in large numbers. Opportunity Wales, which ran from 2001 to 2007, reached nearly 12,000 small companies with free reviews, small matched grants and local delivery partners.^{106, 107} Superfast Business Wales supported 6,537 companies, and 91% of them were microbusinesses.¹⁰⁰ Welsh delivery bodies can find and help the smallest companies.

What did not work

43Wales has paid for access far more than for use. Between 2002 and 2006 the availability of broadband in Wales rose from about 30% to about 99% of the population, while take-up reached only 32%.¹⁰⁸ When Wales later built superfast broadband, it spent £213.7m of public money on the network under Superfast Cymru.⁹⁸ About £13.1m, from an approved budget of £14.9m, was spent on Superfast Business Wales, which helped companies to make use of it.^{100, 101} That is about 6p for every £1. In 2015 the Wales Audit Office criticised Superfast Cymru because no targets had been set for take-up.⁹⁹

44Meeting output targets has not always meant better results. Superfast Business Wales beat every output target, yet the real turnover of the companies it supported grew 11.4% less than that of matched comparison companies over three years.¹⁰⁰ The Technium network produced 86 new high-technology companies against a target of 250.^{96, 97} Help to Grow: Digital, a UK voucher scheme, redeemed only 20 of its 830 vouchers in Wales.¹⁰³ Success has often been judged by activity rather than results.

45Programmes have stopped when their funding stopped. Opportunity Wales stopped helping companies directly in 2007 while it sought new money.¹⁰⁶ Superfast Business Wales ended with its European funding in April 2023, and no funded successor could be found.¹⁰⁰ The barriers named in Wales’s 2001 strategy for information technology, which included a lack of awareness, a lack of skilled staff and a lack of Welsh language tools, are close to those that Welsh companies report for AI today.^{109, 66}

The lessons

46The record offers five lessons.

• Money for access has been far larger than money to help companies use what they can access.
• Help that went beyond awareness, such as diagnosis and hands-on support, is where earlier programmes spent least.
• Local partners have reached the most companies.
• Output targets were met more often than results for companies were shown.
• Programmes have tended to stop when their funding stopped, just as they began to work.

Section 05

The risks of AI fall unevenly across Wales

47AI brings risks as well as gains. They fall on jobs, on places, on the Welsh language and on the smallest companies.

48The economic model built for this report puts about 565,000 Welsh job equivalents, 38.7% of employment, in tasks exposed to AI, slightly below the UK’s 40%. Exposure is not the same as job loss, and the Office for Budget Responsibility expects most exposed occupations to be complemented rather than replaced by AI.^{110, 111} On the current path, the central estimate of jobs displaced by 2035, before people find new work, is about 45,000, rising to about 57,000 if adoption speeds up.¹¹⁰

49Displacement costs more in Wales, where economic inactivity among people aged 16 to 64 was 24.0% against 20.9% for the UK in June to August 2026.¹¹² The Senedd asked in May 2025 for an analysis of the groups most at risk, and the government accepted the request in principle, but no such analysis had been published by late October 2026.^{15, 16} On 21 October 2026, answering a question on AI and entry-level jobs for young people, the minister said that “the evidence is still emerging” and that work to understand how AI is “playing out across different sectors and across different geographies” was ongoing.²⁴

50The places where companies are smallest are often those least ready to benefit. Ceredigion, Powys, Anglesey, Gwynedd, Pembrokeshire and Carmarthenshire have the lowest full fibre broadband coverage in Wales, between 58.7% and 73.0% of premises against 80.8% for Wales as a whole, and four of them have the highest shares of Welsh speakers.^{42, 113} For business premises the figure is lower still, at 57.6% across Wales and 47.6% in rural areas.¹¹³

51In 2021, 17.8% of people in Wales aged three and over could speak Welsh, rising to 64.4% in Gwynedd and 55.8% on Anglesey.¹¹⁴ The Welsh Language Commissioner considers many modern language models able to handle Welsh well, but advises testing them in Welsh before use.¹¹⁵ The Welsh Language Technology Action Plan ran from 2018 to 2024, and the government has since set new priorities rather than a new plan.¹¹⁶ In September 2026 it announced a partnership with the AI company Cosine to improve how Cosine’s model handles Welsh, and it told the Senedd on 1 October 2026 that the partnership involves no commercial agreement and that the government has directed the company to openly available Welsh language resources.^{117, 118} Welsh language technology grants for 2026 to 2027 went to Bangor University, Cardiff University, Data Orchard and the National Library of Wales, but their amounts have not been published.¹¹⁹ No study was found of how Welsh-speaking small businesses use AI tools.

52A paper to the Social Partnership Council said that inequalities in access to AI are “real and widening”, particularly by geography, by language and for groups that include small companies on low incomes.¹²⁰ The Council set up a sub-group on AI and workforce skills in 2025, and in February 2026 its AI Delivery Group was considering how to measure the impact of the training subsidy pilot. No evaluation of the pilot has been published.^{121, 25} AI was not on the agenda of the Council’s meeting of 5 October 2026.¹²² TUC Cymru’s Congress in September 2026 called for productivity gains from AI to be shared with workers through shorter working time with no loss of pay.¹²³

53Most of the displacement in the model happens on the current path, whatever support Wales offers. The risks fall hardest on rural places, on Welsh speakers and on the smallest companies. The analysis of the groups most at risk that the Senedd asked for would show more clearly where help is most needed.

Section 06

If Wales catches up with the UK by 2030, it could gain about £1.1bn of extra output over ten years

54AI may also add to Welsh output. In early October 2026 no credible published estimate of the effect of AI on the Welsh economy could be found, and the only Welsh figure in circulation, from PwC in 2017, predates generative AI.¹²⁴ The Senedd’s updated research briefing of September 2026 no longer repeats that figure and gives only a UK-wide estimate.¹²⁵ An economic model was therefore built for this report. It applies the Office for Budget Responsibility’s method, which works from the tasks that AI can do in each industry,¹¹¹ to Welsh output by industry, and it compares the pace of adoption in Wales with that in the UK.¹¹⁰

55The central estimate is that if Wales’s rate of AI adoption catches up with the UK’s by 2030, Wales gains about £1.1bn of extra output over 2026 to 2035, measured as gross value added in 2024 prices. The gain rises to about £326m a year by 2035.¹¹⁰ Because a figure of this kind is only as good as its assumptions, this report sets out every step of the calculation.

How the figure is calculated

56The calculation has six steps.

• Step 1. The Welsh economy produced £87.4bn of output, measured as gross value added, in 2024. The model assumes that it grows by 1% a year in real terms, in line with the Welsh Government’s assumption for trend productivity growth, so that it reaches about £97.5bn in 2035.¹¹⁰
• Step 2. Once AI is fully adopted, the Office for Budget Responsibility’s central estimate is that it raises the level of UK productivity by 2.3% over a decade, which the model treats as the same rise in output.¹¹¹ Taken as a whole, the mix of Welsh industries and jobs is slightly less exposed to AI than the UK’s, so the Welsh equivalent is about 2.16%, which is 2.3% multiplied by 0.94.¹¹⁰
• Step 3. Adoption takes time. The Office for Budget Responsibility expects the gain to build slowly, adding about 0.2 percentage points a year to productivity growth by the end of its forecast period, around 2030. On the model’s reading of that path, the UK realises about 20% of the gain by 2030 and about 94% by 2035.^{111, 110} The Office’s October 2026 forecast keeps this central assumption, and notes that 30% of all UK businesses, of every size, were using at least one AI technology in September 2026, up from 15% two years earlier.¹²⁶
• Step 4. Wales currently adopts AI at about 84% of the UK rate. This is the mid-point of two measures, 74% from the UK Innovation Survey and 94% from the Office for National Statistics business survey after adjusting for company size.^{65, 66, 110}
• Step 5. If Wales stays at 84% of the UK rate, it captures 84% of the gain each year. If the gap closes in equal steps each year from 2026, Wales reaches the UK rate in 2030 and captures the full gain from then on.
• Step 6. The difference between these two paths in each year is the potential gain. Table 7 shows it year by year.

Source: economic model described in Appendix A.¹¹⁰

57The final year shows how the steps fit together. In 2035 the full gain from AI in Wales would be about £2.1bn, which is 2.16% of £97.5bn. On the UK path, about 94% of it, or about £2.0bn, would be realised by then. A Wales that stayed at 84% of the UK rate would miss about a sixth of that, roughly £0.3bn, which is the £326m in the table once unrounded figures are used.

How wide the range is

58The estimate is uncertain. The size of the Welsh adoption gap is the first uncertainty. If the true gap is larger than the central case, the potential gain is larger, and if it is smaller, the gain is smaller. On the central assumption about productivity, a gap that lasted through the decade would cost Wales between £440m and £1.84bn of output over 2026 to 2035.¹¹⁰

59The size of AI’s effect on productivity is the second, and larger, uncertainty. If that effect is lower or higher than the Office for Budget Responsibility’s central case, the figure scales accordingly, and across all the cases modelled the cost of a lasting gap lies between about £130m and £3.29bn over the decade.¹¹⁰

60The model does not claim that any particular programme would close the gap, and it leaves out investment in data centres. It also produces a larger figure on which this report does not rely. If support brought Welsh adoption forward by a year as well as closing the gap, the model adds a further £2.0bn, but no evaluation in Wales shows how far support speeds up adoption.¹¹⁰

Who would gain

61Small companies would receive a substantial share of the gain. In the central case for 2035, small and medium-sized companies, including microbusinesses, would receive about 37% of the whole gain from AI, and microbusinesses alone about 15.5%.¹¹⁰ The gain does not stop in 2035, and output that Wales would otherwise have missed continues to be earned in the years that follow.

62For scale only, three years of spending at Northern Ireland’s rate, about £13.5m, is about 1.2% of the £1.1bn central gain. This is not a return on public spending, because extra output is not the same as extra tax revenue and the model does not link spending to adoption.¹¹⁰

Section 07

Issues for consideration and next steps

The issues the evidence raises

63The evidence in this report raises a small number of practical questions for the Welsh Government and its partners. They are set out here to inform discussion. The answers depend on choices that belong to ministers, the Senedd and the organisations that would deliver support.

64The first is funding. The AI plan has no budget of its own. The next opportunity to give it one is the 2027-28 draft budget, due on 17 November 2026 with its full line-by-line detail, followed by supplementary budgets and the budgets after that.^{10, 11} The UK Budget of 28 October 2026 added about £450m to Welsh funding over the spending review period, but the Welsh Government says that day-to-day funding will still be 0.5% lower in real terms next year, so any new budget line would have to be found from limited funds.^{12, 54} The record in Section 4 suggests that the length of any commitment matters as much as its size, because support has tended to stop when money stopped.

65The second is focus. Small businesses are 99.3% of businesses in Wales, about 2p in every £1 of public money awarded for AI work has gone to helping businesses adopt AI, and in the central case small and medium-sized companies would receive about 37% of the gain from AI.^{41, 7, 110}

Delivery in Wales

66Much of what makes the Welsh economy distinctive bears on how support is delivered. It has many microbusinesses, rural areas with weaker connectivity, and large tourism and farming sectors, and most people in Gwynedd and on Anglesey speak Welsh.¹¹⁴

67The record of earlier programmes is relevant here. Opportunity Wales reached nearly 12,000 small companies through local delivery partners.^{106, 107} Help to Grow: Digital, a national scheme, reached 20 Welsh companies, and its evaluators recommended local channels and one-to-one advice.¹⁰³ When Superfast Business Wales arrived as a free national service, a local IT company told the Senedd in 2019 that it had been “completely cut out” of the e-business advice it used to deliver, and that the programme had not signposted companies to it.¹²⁷ How a programme works with existing local providers appears to affect how far it reaches.

68Wales already has relevant capacity. M-SParc, a Welsh science park, is delivering the Business Wales programme, a Cardiff University hub has given small companies technical help with AI, and Bangor University and M-SParc have surveyed Welsh small companies on their use of generative AI.^{4, 95, 128} The survey project is recorded as finished, and no results had been published by late October 2026.¹²⁹ The UK Budget of 28 October 2026 included £2.5m, subject to a business case, to expand M-SParc’s space and support for small businesses and start-ups, although this is not money for AI adoption.⁵⁴ Business Wales’s Helo Blod service lists Welsh language technology for the workplace, including AI assistants such as Copilot, ChatGPT, Claude and Gemini that are starting to handle Welsh.^{119, 130} The Development Bank of Wales received a new remit for 2026 to 2030 on 30 September 2026, which asks it to help companies invest in “digital tools, technology adoption” but does not mention AI.¹³¹ Sbardun, a 24-month programme of intensive support for businesses with strong growth potential announced on 2 October 2026, starts with twenty of the Development Bank’s existing customers and also does not mention AI.¹³² These efforts are not yet joined into a single offer that reaches every part of Wales.

69Coverage is uneven. None of the in-person AI events in the Business Wales listings of October 2026 was in Powys, north-east Wales or the Valleys. On 29 October the nine in-person sessions listed for November and December were in Cardiff (four), Aberystwyth, Newport, Pembroke Dock, Gwynedd and Monmouthshire, and the Aberystwyth accelerator was already closed to booking.¹⁸ Separate AI workshops for tourism businesses, funded by the Welsh Government and run by MWT Cymru, were held or scheduled in these areas in October and November 2026, in Ebbw Vale, Hay-on-Wye, Rhayader, Machynlleth and Llangollen.²⁰ The Welsh Language Commissioner advises that AI tools be tested in Welsh before they are used.¹¹⁵ The Commissioner is also developing a framework with Business Wales to encourage the small businesses it supports to commit to Welsh language provision.¹³³

An illustration of what a programme might cost

70To give a sense of scale, this part estimates what a programme might cost. The figures are illustrations, not proposals. At Northern Ireland’s rate of about £1.43 per person a year, Wales’s population of 3,175,181 implies about £4.54m a year.^{75, 78} At Scotland’s 2026-27 rate the figure would be about £1.03m a year,⁷⁹ and at the Basque rate about £17m, although the Basque scheme is a large grant scheme for industry.⁸⁰

71A cost was also built up from the average cost per company of earlier programmes. Table 8 shows what about £4.5m a year would pay for at those averages, if support followed the pathway that the Business Support Review described, with a separate voucher for microbusinesses and sole traders. Over three years this would come to about 3,000 diagnoses, technical help for 210 companies, 120 grants and 600 vouchers.

The average costs per company are taken from earlier programmes and rounded. The cost of a diagnosis is based on the mid-point of the Digital Tourism Business Framework (£762) and Superfast Business Wales (£2,004), and the cost of technical help is based on the mid-point of Scotland’s first year (£6,250) and the Hartree Centre Cardiff Hub (£17,426).^{102, 100, 79, 83, 94, 95} Grants are costed at the Made Smarter maximum of £20,000, and vouchers at the Irish maximum of €5,000 at 50%, which is about £4,250.^{105, 86} Because grants are costed at their maximum, actual spending per company would usually be lower. The numbers of companies are set against past reach. Superfast Business Wales supported about 800 companies a year and the Hartree hub engaged 115 companies in 2024.^{100, 95} The illustration assumes that evaluation is paid for from within the total.

72For scale, £4.5m a year is about 20% of the £22.5m core Business Wales budget for 2026-27, and about £18 per business a year.² This compares with about £2.85 per business awarded for AI adoption in total over the past four and a half years.^{7, 43}

Questions of design

73Beyond the amount, the main design questions are whether support should follow the pathway that the Business Support Review recommended, from awareness events to a short expert diagnosis, then technical help and a matched grant,¹⁴ and whether microbusinesses and sole traders should have a route of their own, in Welsh and in English. In the Basque Country live online training in digital skills is open to the self-employed, and in Ireland vouchers are open to self-employed people who employ staff.^{91, 86}

74Two further questions would need no new money. One is whether councils should allow software, subscriptions and training in their business grants where these are for adopting AI.^{59, 60} The other is whether the AI Growth Zone money for local adoption should be delivered through existing Welsh support, so that companies meet one offer rather than two.⁵³ Regional Local Growth Fund calls are not designed for this either. The South West Wales call, which closes on 15 November 2026, is mainly for capital projects typically worth more than £500,000, with limited revenue funding, and it does not mention AI or digital adoption.¹³⁴

Measuring results

75Earlier programmes were judged mostly by activity. Measures that would show results include spending and take-up by company size and region, the share of companies that move from awareness to putting AI to use, and comparisons with similar companies that were not supported.

76The smallest companies are measured only by small samples. Weighted Welsh estimates of AI use from the Office for National Statistics business survey, of the kind Scotland already has, would help fill that gap. Scotland’s latest weighted figure is 36.9% of businesses with ten or more employees, for September 2026, and Wales has no such figure.¹³⁵ The Longitudinal Small Business Survey now asks about AI, but only of one part of its sample, with about 150 Welsh employers answering the AI question, so a larger Welsh sample and an AI question asked of the whole sample would make its Welsh figures more reliable.⁷¹ Earlier Welsh figures may also be obtainable from the 2022 returns of that survey.¹³⁶

Next steps for this research

77This initial report rests on published sources only. The next stage of work should hear directly from Welsh small businesses about how they use AI and what help they need, through a survey and interviews in Welsh and English. It should also seek unpublished figures on spending and take-up through Freedom of Information requests, test common AI tools in Welsh, and update the estimate in Section 6 as better measures of Welsh adoption appear.

Conclusion

78Wales has begun to act. It has an AI plan with a clear aim for business, an Office for AI, free AI events delivered across Wales by a Welsh science park, and a training subsidy. The new government has promised a more coordinated national approach.

79The evidence also shows gaps. The plan has no budget of its own, no spending has yet been reported from the £2.1m package, and about 2p of every £1 of public money awarded for AI work in Wales has gone to helping businesses adopt it, although small and medium-sized businesses are 99.3% of Welsh businesses. Larger Welsh companies already use AI less than companies in any other UK nation, the smallest are measured only by small samples, and Northern Ireland spends about four times as much per person as Wales has committed.

80Wales’s earlier programmes show how support has reached small companies in large numbers, and why it has sometimes ended before it showed lasting results. The risks of AI fall hardest on rural places, Welsh speakers and the smallest companies. The central estimate in this report is that catching up with the UK’s rate of adoption by 2030 could add about £1.1bn of output over ten years, with gains continuing after that.

81These findings are a starting point. They are offered to inform the decisions ahead, including the 2027-28 draft budget, and to be tested and built on in the next stage of this work.

Appendix A

How the research and the estimate were done

This report was prepared by desk research from published sources. Policy documents, budget tables, Senedd records, council papers and evaluations of earlier programmes in Wales, the rest of the UK, Ireland and the Basque Country were reviewed. The main public records of money were also analysed, including the Welsh Government’s budget tables and payment files, Innovate UK’s records of funded projects, UK Research and Innovation’s grant records, procurement notices, Senedd written questions and the published records of the 22 councils.

To measure public money for AI, every item that could be found for April 2022 to mid-October 2026 was recorded, each with its source. Duplicates and parts of larger items already recorded were removed, so that each sum of money appears once. The money was grouped by purpose, and each sum was recorded at the furthest point it had reached, whether announced, set aside in a budget, awarded to a named recipient or paid out, and these stages were never added together. The main figures include only grants and contracts whose purpose is clearly AI, and they leave out wider digital programmes and all private investment. Figures per business use the Welsh Government’s figure of 246,500 businesses active in Wales in 2025. Every item was checked against its original source. The figures take account of events up to 29 October 2026, including Innovate UK’s records of 12 October 2026, the second supplementary budget for 2026-27 and the UK Budget of 28 October 2026.

The economic model follows the Office for Budget Responsibility’s method. For each industry in Wales, it estimates the share of work that AI could do, how much of that work could move to AI within a decade, and the saving in cost that would result. With UK figures, this approach reproduces the Office’s estimate that AI raises the level of UK productivity by 2.3% over a decade, which the model treats as the same rise in output. The model adjusts for the mix of industries and occupations in Wales, which gives a Welsh equivalent of about 2.16%. It assumes that Welsh companies adopt AI at between 74% and 94% of the UK rate, the range given by the two available measures, with 84% as the central case, and that adoption follows a reading of the Office’s central path over time, in which about 20% of the gain is realised by 2030 and about 94% by 2035. The Office does not state a percentage path itself. It expects the gain to add about 0.2 percentage points a year to productivity growth by the end of its forecast period, around 2030, between its J-shaped and S-shaped paths, and the mid-point of those two paths gives about 19% of the full gain after five years, which the model rounds to about 20% by 2030. It then compares this path with one in which Wales closes the gap in equal steps and reaches the UK rate by 2030. Output in Wales is taken as £87.4bn in 2024, growing by 1% a year in real terms. Estimates of jobs displaced scale published UK estimates to Wales. The model leaves out the AI industry itself, investment in data centres and AI Growth Zones, and wider effects on demand and wages.^{110, 111}

The research has limits. No Freedom of Information requests were sent and no survey was run. The size of AI’s effect on productivity is the largest uncertainty in the estimate, and the size of the Welsh adoption gap is the second, because the two available measures differ and neither covers sole traders. The estimate is not a forecast, and it does not show that any particular programme would close the gap. Quotations from ministers who spoke in Welsh use the Senedd’s official English translation.