Lord O’Donnell calls for AI ‘winners’ to fund worker retraining

TL;DR:

• Lord Gus O’Donnell, UK cabinet secretary 2005–2011, told the FT that AI’s “dramatic” productivity boost in the public sector will put “lots of people out of work” and that the government should redirect AI-driven corporation and income tax revenue into mass retraining.
• He argued digital ID is “crucial” to harnessing AI productivity gains and criticised the Starmer government for initially framing the project around illegal immigration rather than service delivery.
• Resultsense view: the timing is striking — fresh YouGov polling out today shows 47% of Britons would back an AI tax against just 20% opposed. The political pressure to monetise AI’s winners and fund displaced workers is no longer a fringe argument.

Speaking at the inaugural lecture of Cambridge’s Downing Battcock Institute, the former Treasury permanent secretary set out how he believes the UK government should use AI: with sharp productivity expectations for the public sector, a workable digital identity system, and an explicit fiscal compact to compensate workers whose skills are made obsolete.

The retraining argument

Lord O’Donnell told the FT that AI would have “a dramatic impact on productivity” inside the public sector, “but along the way you may be putting lots of people out of work”. His proposal is straightforward in principle: as private-sector profits and individual earnings rise on the back of AI productivity gains, more corporation tax and income tax should flow to the Treasury, and a substantial share should be earmarked for retraining. He said he was not yet clear on the exact tax design, but stressed that “winners compensate the losers so that overall, you’re net better off”.

Digital ID as enabler

He also argued that digital ID is essential to making AI productivity work in government, because AI’s value depends on linking data sources to specific individuals. He criticised the early presentation of the Starmer government’s digital ID project around illegal immigration as having “set back the argument” for citizen-facing benefits, and welcomed the recent pivot to framing the scheme around service delivery.

He added that ministers themselves should receive AI training. Chancellor Rachel Reeves has publicly said she does not use AI; O’Donnell suggested ministers could even use Claude or ChatGPT to draft challenge speeches that test the civil service.

Looking forward

O’Donnell’s argument lands in a domestic environment where appetite for AI-targeted taxation is real but politically unsettled. The fresh YouGov data shows Labour, Lib Dem and Green voters running 55–58% in favour of an AI tax, with Reform UK voters split. Whether the Treasury constructs an explicit retraining levy, or simply leans on rising headline revenues, is now an open Whitehall question — and one that intersects directly with the digital ID rollout O’Donnell sees as the critical enabling infrastructure.