Summary

  • HM Treasury announced an additional £23 million over three years for the AI Economics Institute on 24 October, ahead of the next Budget.
  • The institute's June prospectus says government policy questions shape its agenda and selected analyses go public; it does not promise a public copy of every dataset, dashboard or model.

Britain is putting money behind a question its AI speeches usually skate past: who gets the gains? The Treasury announced another £23 million over three years for the AI Economics Institute, or AIEI, on 24 October. The institute was introduced in June to measure what AI does to productivity, employment and living standards. A budget line gives the scorekeeper time to build its instruments. It also makes the rules for showing the score worth reading.

AIEI is a joint research organisation of the Treasury and the Department for Science, Innovation and Technology. Its prospectus gives it two jobs: assemble evidence from surveys, administrative records and potentially private model-usage data; then build economic models and scenarios. It wants indicators across workers, regions and firms, and it explicitly lists dependence on compute, energy, talent and global supply chains among the risks to study. This is a wider brief than counting how many offices installed a chatbot.

The interesting sentence is about who sees the work. AIEI says live policy questions will shape its research agenda, and that it will publish selected analysis externally where that supports open debate. Departments, not AIEI, make policy. The institute is designed to put evidence in their hands first. A minister can use a model built on private usage data; a worker, employer or MP may see only the findings chosen for release. The official papers do not give a publication schedule or promise access to the full dashboards, methods and assumptions.

That matters because the prospectus itself warns that current economic estimates rest on assumptions about adoption, labour substitution and demand. It asks how gains split between capital and labour, and how much value Britain captures while importing AI services and infrastructure. One model can tell a cheerful national story while hiding a bad wage or regional story in the average. The institute's own research questions make those differences central.

The £23 million is an announced allocation for the coming three years, not a published spending ledger or a measured AI dividend. The Treasury also announced £10 million a year for work on jobs and industry, but did not explicitly equate that line with AIEI's allocation. Keep the accounts separate. The next public test is simple: when AIEI produces findings, show the method, the distribution and the assumptions alongside the headline. A scorekeeper earns trust by opening the scorebook.

Sources