The Public Value Ledger: Auditing Labour's AI Ambition Before the Boom Arrives

Labour did not come to office to preside over an economy that grows in the aggregate while fraying at the edges. Yet as the government makes artificial intelligence the centrepiece of its 2026 growth strategy, that is precisely the risk. The promise of new data centres, new clusters and new private investment is welcome. The question is whether the gains will be spread across the country and shared with the workers who create them.

Gordon Brown's Treasury was rarely shy about growth. It was, however, insistent that economic reform carry a measurable public purpose. Tax credits were judged by child poverty; regional policy by jobs and investment outside the South East; fiscal rules by credibility in the bond markets and fairness in the classroom. The same discipline now needs to be applied to AI.

Why growth needs a fairness ledger

The AI boom is not a single policy. It is a bundle of choices about compute, data, skills, procurement and tax. Each choice distributes advantage. A grant to a data centre in one region is a decision not to invest in another. A public-sector data deal is a decision about who owns the value created from patients' records or pupils' progress. A tax incentive for automation is a decision about the balance between machines and people.

Brown understood that markets need rules, but he also understood that rules need evidence. His Treasury habit was to attach conditions to public support and then publish the results. The AI economy needs the same treatment: not a regulator that stifles innovation, but a public value ledger that makes the distribution of gains visible before commitments are made.

Four tests for the AI economy

The first test concerns data ownership and public benefit. Public data is public infrastructure. When the NHS, local government or schools enter commercial partnerships, the public should know what is being shared, on what terms and for what return. A Brown-style settlement would require transparent licensing, a public register of material data deals and a clear principle that public value is captured, not merely gifted. Innovation should be encouraged; the family silver should not be sold by accident.

The second test is regional spread. Britain's AI footprint must not become a golden triangle writ large. Investment incentives, procurement and skills funding should be designed to build capacity in every nation and region, including places whose contribution to the last industrial revolution was written out of the ledger. That means conditions on major support packages, local supply-chain plans and a serious role for devolved administrations and combined authorities. Growth that clusters in a few postcodes is not a national growth strategy.

The third test is job quality. The number of new AI jobs matters, but so does the character of work. Labour should ask whether adoption raises pay, security and training, or simply displaces tasks and intensifies pressure. Public contracts can require workforce plans, union recognition where appropriate, and funded transition training. The Treasury's old insistence on making work pay should now become an insistence that the AI transition does not create a two-tier labour market.

The fourth test is public accountability. Algorithmic systems used in public services must be auditable, explainable and open to redress. Procurement should not outsource responsibility along with delivery. Parliament should receive an annual public value report on the AI economy, with independent evaluation of costs, benefits and distributional effects. The public should be able to see whether the promised gains are arriving, and for whom.

Credibility is not an obstacle to fairness

None of this is an argument against AI. It is an argument for specifying what AI is for. Brown's record reminds us that fairness without fiscal credibility is a slogan, and credibility without fairness is a machine for managed decline. The AI strategy must therefore be costed, evaluated and transparent. Public support should come with sunset clauses and independent assessment. Tax reliefs should be reviewed against evidence, not renewed by habit. The Office for Budget Responsibility and the National Audit Office should be able to scrutinise the claims made for the programme.

There is also a temptation to treat AI as a magic revenue source that will spare hard choices. It will not. The gains may be real, but they will arrive unevenly and over time. The public value ledger should guard against the assumption that a new technology can substitute for a fair tax system, a functioning skills settlement or sustained investment in public services.

A specification, not a brake

The fairest way to describe this agenda is as a specification. Labour should say to investors: Britain wants your AI investment, and Britain will help you build it. But public support will be judged by whether it spreads opportunity, protects data rights, improves work and remains accountable. That is not red tape. It is the condition on which a lasting coalition for growth can be built.

The Brown test was never about slowing enterprise down. It was about asking enterprise to serve a country, not just a balance sheet. As the AI boom gathers pace, Labour should apply that test now — before the contracts are signed, the clusters are fixed and the inequalities become another inherited failure.