Fixing Britain’s welfare state and the limits of AI policymaking

This article is an on-site version of our Inside Politics newsletter. Subscribers can sign up to get the newsletter delivered every weekday. If you’re not a subscriber, you can still receive the newsletter

Good morning. I’m Chris Smyth, the FT’s public policy editor, and in covering for Stephen today I was tempted to write about welfare reform. After all, following my story last week on recommendations for a separate youth benefit system, both Andy Burnham and Alan Milburn have been publicly making the case for change this week.

But given the title of this newsletter, I hope you’ll forgive me for focusing less on the policy itself than on how it is being made and presented. And if you stick with me, I promise I’ll actually end up talking about AI policymaking.

The starting point for this year’s welfare reform process is last year’s failed attempt at disability benefit cuts, which helped mark the beginning of the end for Keir Starmer. In response, ministers have stopped talking about the rising cost of sickness — in reality central to the desire for reform — and more about what Burnham has termed “the moral failure” of denying young people a chance in life.

The focus on young people was a smart choice by Pat McFadden, the Labour veteran brought in as work and pensions secretary to clear up Starmer’s mess. Politically, the sense of a “lost generation” consigned to a life without purpose has a visceral power. Economically, this is the group with the greatest potential long-term cost to the state. And practically, policies developed for young people can be extended if they work.

McFadden in turn brought in Milburn to build a case for change. This would be done through an in-depth review of the causes of the Neet problem, to be followed next month by recommendations. Along with other reviews — one on the rise in mental distress in younger people by professor Peter Fonagy, and another on personal independence payments by Sir Stephen Timms — Milburn’s work has both expanded the evidence base and recast the problem in a way that is helpful to reform.

Alongside an intensive outreach effort to Labour MPs, this approach has put the government in a much stronger position to set out difficult reforms than the abrupt cost savings that emerged last year.

Inside Politics is edited by Georgina Quach. Follow Chris on X and Georgina on Bluesky. Read the previous edition of the newsletter here. Please send gossip, thoughts and feedback to insidepolitics@ft.com

No prompt solution

Which brings me, naturally, to AI. Like the private sector, the government is desperate to harness the power and cost savings that the technology brings. Some permanent secretaries talk excitedly about radically slimming down the civil service, and AI is due to be a key theme of cabinet secretary Antonia Romeo’s review of Whitehall, as authoritatively outlined by my colleague Lucy Fisher.

Yet in a fascinating paper for the Harvard Kennedy School, former shadow chancellor Ed Balls and former civil servants Nyasha Weinberg and Dylan Evans demonstrate some of the risks of this approach. Yes, they find commercial agents are excellent at rapidly summarising evidence and joining up disparate datasets.

More interesting is what they find AI is bad at. A simulation of how civil servants might use it to generate policy proposals on difficult problems such as housing supply and NHS productivity finds that AI swallows government spin and churns out “policy slop” — technocratic recommendations that offer “beige” variations on existing solutions. Crucially, it fails to grapple with political trade-offs or to produce anything genuinely novel — and is terrible at telling which policies actually work.

The irony is striking — the world’s most transformational technology currently risks deepening the tendency towards complacent mediocrity in Whitehall that Starmer memorably dubbed “the warm bath of managed decline”.

Which brings us back to welfare. The government is making a deliberate effort to change the inputs and framing of the problem, and so come up with solutions that were not — unlike last year’s cuts plan — left lying around Whitehall by the previous administration. It is an interesting speculative question whether such different source material would lead to different policy recommendations (my experiments with ChatGPT suggest it would; but still less radical ones than what Milburn is likely to propose).

The more fundamental point is the difference made by wily political operators who have been around long enough to look beyond the narrow policy issue and work out how to build a long-term political argument. This is the kind of skill that could end up being atrophied by over-reliance on AI.

In time, perhaps, we will be able to point the models to instances where deep policy expertise has been combined with political nous and deft communication to build a durable consensus on reform of a contentious policy problem. Perhaps by next year, welfare reforms will be a case study. But it’s not the least of Britain’s problems that, right now, we are somewhat short of training data.

Now try this

The BFI London Film Festival finished last week, so if you missed it, you will have to wait to see A Long Winter. But it will be worth it. Andrew Haigh’s delicate and unsettling film encompasses growing up, choosing your path in life and the strange freedom of bereavement. Set against the snows of Colorado, it manages to be sad and exhilarating at the same time. Look out for it when it arrives on general release next year.

Top stories today

• Warning from Office for Students | Poor-quality degrees are “tarnishing the reputation” of England’s universities, the higher education regulator has said, as it pledged to crack down on franchise courses and teaching that does not improve students’ prospects.
• Hope in every heart? | To gauge if Andy Burnham is winning over the public — and to see if any of his rivals are doing any better — the FT worked with Merlin Strategy to talk to 13 swing Labour voters in Sheffield.
• Career costs of non-competes | Andy Burnham this month pledged to introduce legislation to ensure non-compete clauses, which prevent workers moving to a competitor for many months, can “no longer be a barrier to hiring for our most promising start-ups”. Martin Arnold and Kieran Smith spoke to the businesses and workers affected.

Warning from Office for Students | Poor-quality degrees are “tarnishing the reputation” of England’s universities, the higher education regulator has said, as it pledged to crack down on franchise courses and teaching that does not improve students’ prospects.

Hope in every heart? | To gauge if Andy Burnham is winning over the public — and to see if any of his rivals are doing any better — the FT worked with Merlin Strategy to talk to 13 swing Labour voters in Sheffield.

Career costs of non-competes | Andy Burnham this month pledged to introduce legislation to ensure non-compete clauses, which prevent workers moving to a competitor for many months, can “no longer be a barrier to hiring for our most promising start-ups”. Martin Arnold and Kieran Smith spoke to the businesses and workers affected.

Recommended newsletters for you

The Week Ahead — Start every week with a preview of what’s on the agenda. Sign up

Newswrap — Our business and economics round-up. Sign up