Better artificial intelligence skills across the UK workforce could add as much as £80bn to economic output by 2035, but researchers warn that shallow adoption and uneven training are limiting the productivity gains available from the technology.
Learning and Work Institute and the Rigby Foundation modelled the potential impact of broader AI adoption and workforce capability, concluding that comprehensive upskilling could support an economy-wide productivity improvement of around 2.3% by 2035 under the central economic scenario used in the research.
The report argues that Britain’s opportunity depends less on the small number of companies developing frontier AI models than on whether employers across the economy can use existing technology effectively in everyday operations.
Current adoption appears broad but relatively shallow. A YouGov survey used in the research found that 85% of respondents said their organisations used AI to some degree, while only 18% described its use as extensive. Larger organisations were more than twice as likely as small businesses to report extensive adoption, at 25% compared with 11%.
The gap appears at the other end of the scale as well. Around 26% of small organisations in the institute’s research were not using AI, compared with 3% of large enterprises. Smaller employers therefore face lower adoption in some cases and less intensive deployment among many of those already experimenting with the technology.
Stephen Evans, chief executive of Learning and Work Institute, said: “AI has transformative potential, and the gains are within reach.”
The distinction between access and capability is increasingly important. Generative AI products can be adopted with little upfront infrastructure, meaning an employee using a chatbot for occasional drafting may technically count as business adoption. That does not necessarily translate into redesigned workflows, shorter process times, higher output, better decisions, or measurable returns.
Productivity gains become more substantial when organisations integrate AI into repeatable processes and give employees enough technical and operational knowledge to use it safely. That can involve redesigning workflows, checking data permissions, defining where human review is required, measuring output quality, and training managers to identify tasks suitable for automation.
The skills requirement is therefore broader than prompt-writing or familiarity with individual software products. Businesses need people who understand processes well enough to redesign them, technical specialists who can connect systems and data, managers who can assess commercial value, and employees able to recognise unreliable or inappropriate output.
The report calls for a more active approach from government and employers, including incentives for investment in skills, faster updating of technical qualifications, stronger leadership capability, and greater flexibility within accredited training. Its central argument is that spending on AI software and infrastructure will not produce the full expected return if workforce investment fails to keep pace.
Different surveys produce different headline adoption rates because samples and definitions vary, but a consistent pattern is emerging: access to AI is spreading faster than intensive operational use. That leaves a substantial implementation gap between experimentation and measurable productivity improvement.
The economics can be particularly difficult for smaller companies. Large enterprises can establish specialist teams, run controlled pilots, and spread training costs across thousands of employees. Smaller employers may have less access to technology specialists and fewer opportunities to release employees from revenue-generating work for structured training.
The divergence creates a risk that AI widens existing productivity gaps between large and small organisations. If better-capitalised businesses move from experimentation into integrated deployment more quickly, they may capture a disproportionate share of the benefit even though the underlying software is broadly accessible.
Skills spending also competes with other investment. Businesses considering AI projects have to fund cybersecurity, data preparation, software licences, integration, change management, and training before the eventual productivity return is known. Weak implementation in any of those areas can reduce the value created by the technology.
The £80bn estimate is a modelled opportunity rather than a forecast that will automatically materialise. It depends on adoption, skills, investment, and the wider economic environment over the remainder of the decade.
The research places workforce capability at the centre of that equation. Acquiring AI tools is relatively straightforward; reorganising work so that the technology consistently improves productivity is a more demanding business change programme.
You must be logged in to post a comment.