Many of your people already use AI at work, and far fewer have been trained in it. This is how a firm of 100 to 2,000 staff turns "get everyone using AI" into a programme it can run and fund.
How to Train a Whole Workforce in AI: Designing an AI Skills Programme for a UK Business
By James Cotton · Last updated · 13 min read
Part of our topic guides on AI Skills for Business and Data Literacy.
By James Cotton, Founder, iO-Sphere
A request to get everyone using AI tends to arrive as a purchase: a course, a licence, a number of seats. The question under it is what each person in the business needs to get done with AI, and which route, at what cost, gets each group there.
Where UK firms are with AI training
The Office for National Statistics article Artificial intelligence in UK businesses: 2023 to 2026, released 20 July 2026, finds in its Opinions and Lifestyle Survey that 55% of employed and self-employed people report using AI for work or education.
ONS's training measure comes from its Business Insights and Conditions Survey, live 15 to 28 June 2026, with 10,311 responses from a sample of 38,637. Among businesses with 10 or more employees that use AI or are unsure whether they do, 11% report that more than half of their workforce has received AI-related training, a level ONS calls "limited in practice".
In 2024, a Department for Science, Innovation and Technology (DSIT) survey of the public found that 84% of people in work had not undertaken any AI-related training in the previous 12 months, and that 21% of people in work felt confident using AI tools in the workplace.
Inside firms that already use AI, most staff do not yet. DSIT's AI adoption research (updated 13 February 2026), from spring 2025 fieldwork among private-sector businesses with five or more staff, found that in businesses using AI an average of 30% of staff were using it, with 18% at medium businesses (50 to 249 staff) and 20% at large ones (250 or more).
In the July 2026 ONS article, training or retraining existing staff is "consistently the most widely used" way to integrate AI skills "across all size bands", with "around 40% of medium- to large-sized businesses" reporting it, and "the adoption of AI is not yet associated with widespread changes in workforce size". For a mid-sized firm, the people who will do AI-assisted work are, for now, its existing staff.
What people will use AI for
What AI gets used for shifts with the size of the firm. In the same DSIT research, the 919 businesses that used or planned to use AI in spring 2025 reported these uses, with mid-sized firms (the survey's 50 to 249 band) and large firms (250 or more) further ahead on data and automation:
| Use | All users and planners | Where size moves it |
|---|---|---|
| Creative and content creation | 77% | 81% at micro firms (5 to 9 staff) |
| Administrative and support | 70% | 79% at large firms |
| Data and analytics | 56% | 68% mid-sized, 86% large |
| Automation | 47% | 58% mid-sized, 75% large |
Overall, creative and administrative work still rank above data and automation; at mid-sized and large firms, data and automation are two of the common uses. Asked in the same research why they adopt AI, 65% of businesses using, planning or actively considering it (n=942) named efficiency or productivity; the next reason, improving employee experience, drew 16%.
ONS finds the same in June 2026: improving business operations is the top purpose for AI in every size band, named by 65.7% of AI-using businesses with 100 to 249 staff and 63.8% of those with 250 or more, with "larger firms tending to focus on efficiency".
ONS's June 2026 figures also show where that work sits. Among businesses using large language model text generation, 45.9% name administrative or clerical roles as affected, 35.8% creative or design roles, 33.2% data analysis roles and 23.9% IT roles.
Employers interested in training pointed the same way. In 2024, DSIT's employer survey found 36% of employers interested in AI training; of those 344 interested employers, 86% wanted training on using AI to reduce repetitive tasks and improve efficiency, and 87% on keeping information safe and private.
In DSIT's qualitative interviews of spring 2025, most organisations understood AI as "a tool that can improve efficiency and automate routine tasks within their business", and on support, "Staff training and awareness were felt to be critical".
Our view, and the point where any whole-workforce design should begin:
When a company asks for AI skills, it is usually asking for people who can get more of their own work done with AI. Two of the most common meanings are automating the routine parts of a job and doing your own analysis instead of waiting for a data team; what it means in a given team depends on the work that team does.
Training that builds the capability on the person's own tasks is what changes how the business runs.
How one team learns this on a real task is the subject of our guide to AI adoption training for non-technical staff, and the routine-work side is taken further in AI automation for business processes.
What the programme should cover
Several public benchmarks describe what a foundation layer holds. Skills England's AI foundation skills for work benchmark, published 28 January 2026 for England, lists as technical skills writing clear instructions for AI tools and using AI tools to support routine processes and tasks; as non-technical, using simple software that automates work tasks and adjusting how an AI tool works to get better results.
Under responsible and ethical use, it lists understanding the risks and consequences associated with AI use, and analysing information using AI tools. Both everyday meanings of AI skills above, automating routine work and doing your own analysis, are already on that list.
The Skills England annual skills report 2026 says most workers will require practical AI literacy, "the ability to use, verify and safely integrate AI tools", while "a smaller share will need specialist technical skills". What AI literacy training has to cover in law is a separate question, answered in our guide to AI literacy training in the UK.
The AI Skills for Business Competency Framework, version 3 (released 9 March 2026), describes personas from Citizen, Worker, Professional and Leader to Citizen Advocate and Change Champion, the last being "trusted peers within teams who help translate an organisation's AI ambition into everyday practice". It says personas "are not intended to map directly onto job titles".
On structure, Dr Nisreen Ameen's research Skills for AI: what works for AI upskilling in the UK (10 June 2026), produced through the British Academy Policy-Led Innovation Fellowship in partnership with Skills England, ties scale to a clear core: "Role-based modules and use-case libraries can allow the same training to be used across teams, with changes only where needed."
In SMEs (small and medium-sized enterprises) and frontline services, the same research finds "very short (30 to 90 minutes), stackable modules are often more practical than longer sessions"; for professional and business services it describes "tiered, role-appropriate pathways from foundational literacy to advanced technical build capability".
Free material exists at national scale. In June 2025 the previous government announced a government-industry partnership to train 7.5 million UK workers in essential skills to use AI by 2030, with the training to focus on "enabling workers to use and interact with AI systems such as chatbots and large language models", and partners including Google, Microsoft and IBM making materials available free of charge.
The same government's AI Skills Boost, set out in January 2026, offered free courses "taking as little as under 20 minutes", and in a GOV.UK release of 8 June 2026 it reported the programme had passed "over 1.7 million AI skills courses completed". Use courses like these as a floor for everyone and give the programme's own hours to each team's tasks.
Choosing the shape of the programme
Our principle under every programme is to start from what the company, team or person needs to accomplish, then find the best way to get there with the funding available, commercial and levy (the apprenticeship levy, which only employers with an annual pay bill over £3m must pay, and which funds apprenticeship training). Two shapes come up most often, and neither is the only right one.
Big transformation programmes take the first shape: custom foundations training across a lot of people, bought commercially, plus champions upskilled through the levy. The champions go deep enough to lead the change in their teams, and one may own its governance; our guide to building AI champions takes the role further.
Smaller boutique programmes, which can be quite advanced, are the second shape. They fit where people are across geographies, do not want to commit to an apprenticeship, or are not eligible for one, and they are sized to the group. For a firm that wants the capability now, built on its own work, a paid custom programme is a first choice in its own right.
Our recommendation is a provider that can run both commercial and levy-funded training, since that is the provider that can carry the whole plan.
Our big commercial programmes are data literacy or data foundations, and AI foundations, run on the team's own tasks. We run them as an open cohort, where people from different companies train together, or a closed cohort for one organisation, where we look for 10 to 12 people; neither is the budget option, and team training is priced per engagement.
Since 2022 we have trained 1,000+ learners in data and AI, including teams at PwC, BAE Systems and Dunelm, and we are rated 4.8 out of 5 on Google. Champions go deeper through AI Transformation (Level 4), and the person who will own governance through Data & AI Governance (Level 4); on the apprenticeship side, iO-Sphere is listed on the Apprenticeship Providers and Assessment Register (APAR).
Paying for it
The levy at a mid-sized pay bill
The glossary of the apprenticeship funding rules for August 2026 to July 2027 says the levy is "charged at 0.5% of an employer's pay bill", with "an allowance of £15,000 to offset against their levy payment". Every sum in this section is our own arithmetic on published figures.
At a £5m pay bill, 0.5% is £25,000, so the levy bill after the allowance is £10,000 a year; at £20m it is £85,000. Skills England lists a funding band for each standard, the most levy or government money pays for one apprentice: £18,000 for ST0117 Business analyst (Level 4) and £13,000 for ST0795 Data technician (Level 3).
A levy bill of £85,000 a year covers four full ST0117 bands with £13,000 over, or six full ST0795 bands, though bands are drawn over the whole programme, so the comparison is a ceiling only. At these pay bills the levy funds depth for a few people, which is the champions half of a big programme.
Where levy funds run short, or the employer does not pay the levy, the government pays all or most of each apprentice's cost up to the band, depending on age, as our guide to AI training for small businesses lays out. The rules also name cohort size as a basis for negotiating the price within the band (paras 39.3.3 and 210.1), which helps when several champions start together.
Funds that enter accounts from August 2026 expire sooner: "After 12 months, funds become unavailable to use and are removed from accounts", says GOV.UK's apprenticeship funding guidance. Funds already in accounts before then keep the 24-month expiry, and money leaves on a "first-in, first-out basis", so plan champion starts before the money lapses.
Existing staff, paid time, units and transfers
Champions are usually people you already employ, and GOV.UK's guidance on employing an apprentice says "Apprentices can be new or current employees"; for each one, the rules look for "a clear and substantial link" between the apprenticeship and their day-to-day duties, a test our guide to apprenticeships for existing employees works through.
Off-the-job training, the apprenticeship's own training time, happens "during the apprentice's normal working hours", which the rules define as "the hours for which the apprentice would normally be paid, excluding overtime" (para 82.1). Each standard publishes its minimum (para 85); Skills England's record for ST0117 lists 348 minimum hours for compliance. Budget that time as well as the fee.
Senior leaders have a shorter funded route in apprenticeship units, programmes Skills England lists with their own hours and funding maximum. Between them, its three Level 5 AI leadership units (AU0009, AU0010 and AU0011) cover AI strategy and opportunity; adoption, procurement and governance; and delivery and organisational transformation. Each needs at least 30 hours, carries a £750 maximum and is open only to people "working in a leadership position".
The apprenticeship unit funding rules fund these units in full, up to the £750 maximum, for employers that do not pay the levy (para 92) and, where a levy payer's funds run short, in full for learners aged 19 to 24 and 75% at 25 or over (paras 93 and 94). A unit is not funded for anyone already on an apprenticeship or another unit (para 24.4 of the unit rules).
A levy payer can also transfer up to 50% of its annual funds to other employers, for apprenticeships and units alike (para 234), as GOV.UK's guidance on transferring your apprenticeship levy to another business describes. The sender commits for the whole apprenticeship (para 237) and may attach no conditions, "such as choosing the receiving employer's main provider" (para 240).
Licences and published prices
When we read the market's published prices on 8 October 2026, DataCamp's Teams plan showed £12 per user a month billed annually, labelled "Special Price", with £23 struck through beside it, and Coursera's Team plan showed £295.18 per user a year for up to 499 users, with 500 or more on an Enterprise plan priced through sales.
Five hundred DataCamp seats come to £72,000 a year at the special price, or £138,000 at the struck-through £23. Coursera's 499 Team seats come to £147,294.82 a year, and a 500th seat moves the firm to Enterprise.
Among training providers, We Call Shotgun lists a half-day executive briefing from £3,500 for up to 15 attendees, a two-day department intensive from £12,000 for up to 20, a 30/60/90-day adoption programme from £45,000, and a UK multi-site rollout "quoted on scope, typically £80,000 to £250,000 for 500 to 5,000 employees", all net of VAT.
Hartz AI prices multi-week team programmes from £5,000 + VAT for four weeks, £8,500 + VAT for eight and £12,500 + VAT for twelve, with groups capped at 20. Elansio quotes "£980+VAT for up to 20 people".
Multiverse lists AI programmes as apprenticeships, labelling "AI-Powered Productivity" and "AI Productivity and Workflows" Level 3 Digital Support Technician and "AI Solutions Builder" Level 4 AI & Automation Practitioner; its prices are not published. Grant Thornton UK sets out a tiered offer, from "AI & Digital Skills for Business Impact" for operations, finance and customer support staff to "AI Leadership short courses" for the C-suite.
Per-head cost across these formats is the business of what AI training for staff costs in the UK, and the wider field of providers is mapped in data and AI training providers for UK businesses.
How to tell it worked
Businesses in DSIT's AI adoption research (spring 2025) reported improved productivity of the workforce (75%) and new or improved processes (57%), while 77% had not yet seen a change in revenue; the report says the figures are self-reported and "should be viewed as estimates".
Our view is that a programme should be measured on work product, the analyses people now run themselves and the routine tasks they no longer do by hand, counted before and after; our guide to measuring workplace training outcomes goes further.
Frequently asked questions
Does everyone in the business need AI training?
Yes, everyone whose work AI will touch needs a foundation, with depth set by what each person must accomplish. The AI skills for life and work summary of 2024 surveys warned that concentrating AI skills in a small, highly educated share of the workforce risked a divide. BCG (26 June 2025, 11 countries) found people with five or more hours of training "significantly more likely to become regular users".
Can the levy pay for the foundations course everyone takes?
No, unless the course is an apprenticeship or an apprenticeship unit. GOV.UK's guidance lets employers spend levy funds on their "own apprenticeship training and assessment costs" or their "own apprenticeship units", or transfer them to another employer, and the funding rules add that "The entire apprenticeship must not be fully delivered by self-directed distance learning" (para 91.1). A short course outside those categories is a commercial purchase.
Can staff outside England join the same programme?
Yes, through a commercial programme, which the firm pays for directly; the funded routes on this page are England's. The levy is charged on UK employers, but the apprenticeship funding rules for August 2026 to July 2027 cover apprenticeships in England, and Skills England's AI foundation skills benchmark is marked "Applies to England". What else is open for staff elsewhere depends on the nation they work in.
Upskilling a whole team?
Tailored data and AI training for organisations, from data literacy to technical upskilling, built around your team's real work.