Keir Starmer is right to gamble on an AI revolution, but it might not pay out in time | Rafael Behr

5 hours ago 1

Keir Starmer made two predictions at the start of his week. He said that artificial intelligence will transform Britain’s economy in the coming years and that Rachel Reeves will continue to run the Treasury. Those were safe bets, but not guarantees. One is a forecast the prime minister makes eagerly, the other was bullied out of him. He would have preferred to talk about AI improving productivity, generating jobs and improving services, without being asked if he plans to sack the chancellor.

He doesn’t, and wouldn’t say even if he did. The question isn’t serious. It is a contrivance, a lobby ritual for turning speculation into news. Demand official comment on an improbable scenario, then interrogate the answer until it surrenders a headline. Starmer isn’t poised to jettison Reeves, but economic pressure on the pair is real. There is no growth. The pound is depreciating and debt costs are rising, gobbling resources that can’t then be used to upgrade the public realm.

The chancellor has already tested the limit of how much revenue she can tap without breaking pre-election pledges not to raise taxes on “working people”. Fiscal rules (self-imposed, but enforced by predatory bond traders) forbid unfunded borrowing. In the absence of a lucky windfall, the spending review in June will be brutally austere. It will demoralise Labour MPs and stir cabinet dissent.

That context gave a restless urgency to the launch of the prime minister’s AI action plan on Monday. The independent report that has been adopted wholesale as government policy was submitted last September. Back then, no one had Starmer down as a techno-utopian.

The prime minister has a reputation as a methodical man who ponders evidence, absorbs submissions, takes his time over decisions, but then fully commits to his chosen path. People close to the discussions say he is now sincere and vigorous in the conviction that Britain must be propelled headlong into a state-sponsored AI revolution. It is a big call about the long term. Also the right one, regardless of present Treasury constraint. If you know for sure that a colossal wave is about to break on the shore, and you have time to prepare, it makes more sense to harness that force to serve your preferred ambitions than to crouch low and hope the things you care about aren’t swept away.

This is a significant moment in the evolution of Starmer’s project in two ways. First, it channels the will of the prime minister into a specific industrial policy, from which concrete spending priorities will flow. That clarity of intent has so far been absent, feeding a sense of drift and incoherence. Second, it describes a theory of progress – a concept of how politics can improve people’s lives – that isn’t measured in traditional social democratic terms by the size of departmental budgets for good causes.

That digital vision has been forced on the prime minister by a fiscal impasse that puts analogue tax-and-spend levers out of reach. But force majeure doesn’t necessarily invalidate the insight. And it is refreshing to hear a centre-left leader talk with enthusiasm about the future, citing beneficial collective applications of technology. It’s about time someone contested an intellectual space that is otherwise dominated by anti-government libertarian dogma and creepy, conspiracy-minded nationalist machismo.

It is easy to list hazards and contradictions on the road to implementing the AI action plan. There are massive unanswered questions about ownership of data, copyright protection, accountability for automated decisions in public bodies, energy usage and how to mitigate its impact on the climate. It isn’t clear yet what the practical difference is between a sovereign government designing its own bespoke, globally competitive regulatory framework and a desperate government pleading for foreign investment and capitulating to the demands of tech companies.

But the difficulty of getting that stuff right isn’t an argument for not trying, given that the alternative is to dither and end up taking whatever answers other regimes come up with first.

There is a more imminent political problem. Money allocated for wiring Britain up for supercomputers and jumbo datacentres has to come from someone else’s budget. The dividend from that investment won’t immediately compensate the losers. The test of whether an AI-focused industrial strategy really is the key to decrypting Starmerism will be how No 10 arbitrates in those disputes; whether the cabinet can unite around the plan when Labour MPs are kicking off about cuts and voters are wondering why cyber-modelling of future potholes seems to be taking priority over filling in present ones.

The prime minister might be right when he says the benefits of AI will be felt much sooner than sceptics appreciate. But the timeline doesn’t map neatly on to a political cycle accelerated by voter impatience and global volatility. In the first phase of the revolution, exciting new plans for digitising the state will look suspiciously like the miserable old routine of shrinking it. Reserves of goodwill are already depleted and neither the prime minister nor the chancellor is gifted at the art of public policy evangelism.

In a more benign economic setting, Starmer’s speech launching the AI action plan would probably have received less attention – treated as a nerdy digression – but it could also have been judged as policy in its own terms. It might have been analysed as a declaration of strategic intent and not belittled as a symptom of frantic fishing for growth, or drenched in speculation about an embattled chancellor.

It is important, interesting stuff. But it has been overburdened with political heavy-lifting because Reeves and Starmer are stuck in a fiscal trap they laid for themselves in the past and badly in need of something upbeat to say about the future.