The City’s AI Skills Compact Is a Start — But Britain Needs More Than Voluntary Pledges

By Mei-Ling Zhao, Opinion Columnist, Technology and Civic Life

Chancellor Rachel Reeves will announce today that major financial firms have signed a “skills compact” committing them to retrain thousands of City workers whose roles are being transformed by artificial intelligence. Barclays, HSBC and others are on board. The goal is straightforward: keep London’s financial engine competitive as AI rewrites the rules of everything from compliance and risk modeling to client advice and trade execution.

It is a sensible, pragmatic step. It is also nowhere near enough.

The announcement arrives at a moment when AI adoption in finance is accelerating faster than most institutions — or governments — are prepared to handle. McKinsey estimates that generative AI could automate tasks accounting for 20-30% of work hours in banking and insurance. Goldman Sachs has projected that up to 300 million full-time jobs globally could be displaced or fundamentally altered by the technology. The City of London, which employs roughly 500,000 people in financial and professional services, sits squarely in the crosshairs.

Reeves’s compact is an attempt to turn that threat into an opportunity. Firms that sign on will pledge to invest in reskilling programs focused on AI literacy, data ethics, prompt engineering, and the new hybrid roles that combine domain expertise with technical fluency. The government will provide coordination, some funding for training infrastructure, and political cover for companies that might otherwise fear accusations of cutting jobs while spending on technology.

That framing is important. In an era when “AI will take your job” has become a populist rallying cry on both sides of the Atlantic, a high-profile compact led by a Labour chancellor signals that the state intends to be a partner rather than a spectator. It also gives companies a shared language and a loose accountability mechanism — something the UK has historically lacked compared with Germany’s more structured vocational systems or Singapore’s SkillsFuture program.

Yet the limitations are obvious from the outset. The compact is voluntary. There are no binding targets for how many workers must be retrained, what quality standards the programs must meet, or what happens if firms simply re-label existing training budgets as “AI skills” spending. There is no mechanism to ensure that the workers most at risk — administrative staff, junior analysts, mid-career compliance officers — are the ones who actually receive the new opportunities rather than senior managers who already have access to expensive executive education.

More fundamentally, the compact treats AI primarily as a skills problem when it is also a power problem. The same technologies that create demand for new competencies are concentrating value in a handful of hyperscale cloud providers and foundation-model companies, most of them American. British banks are becoming sophisticated users of tools they do not own and cannot fully audit. A workforce that is merely “AI literate” will still be dependent on foreign platforms whose pricing, data policies, and model updates are set elsewhere.

The compact also sidesteps the question of who pays when entire job categories shrink. If AI reduces the need for large teams of document reviewers or basic financial modelers, firms have every incentive to let natural attrition handle the transition while claiming credit for the workers they do retrain. The social cost of displacement falls on individuals and the public purse; the productivity gains accrue to shareholders. Without stronger conditionality — tax incentives tied to actual hiring outcomes, portable training accounts that follow workers between firms, or requirements that companies disclose AI-driven headcount reductions — the compact risks becoming a public-relations exercise.

There is a better model available if Britain is willing to look beyond the Square Mile. The Netherlands’ “Human Capital Agenda” for the digital transition combines public funding with mandatory employer contributions and individual learning accounts that workers can take with them. Denmark’s flexicurity system pairs generous unemployment benefits with aggressive active labor-market policies, including paid retraining leave. Neither country pretends that voluntary corporate pledges will suffice when the technological shift is this large.

For London specifically, the stakes are not just economic but civic. The City has long been one of Britain’s most successful export industries and a magnet for global talent. If the transition to AI-augmented finance is handled well, it could reinforce London’s position as a sophisticated, high-value hub. If it is handled poorly — if thousands of mid-career professionals find themselves under-employed while a small cohort of AI specialists captures the gains — the political backlash will be swift and the social fabric of one of the UK’s most dynamic regions will fray.

Reeves’s announcement is therefore best read as an opening bid rather than a finished policy. It correctly identifies that the state has a role in coordinating the response to technological change. It correctly frames reskilling as a shared responsibility rather than a purely individual burden. What it lacks is the scale, the accountability, and the recognition that AI is reshaping not only skills but bargaining power between labor, capital, and the platforms that increasingly sit between them.

The coming months will show whether the compact expands into something more ambitious — perhaps a statutory training levy with real teeth, or a national AI transition fund financed by a modest tax on large-scale automation — or whether it remains a well-intentioned but limited gesture. For the workers whose jobs are already changing and for the country that needs their productivity, the difference will matter enormously.

In the meantime, the City’s firms should treat the compact as the floor, not the ceiling. The ones that move fastest to give their people genuine agency over the next generation of tools — not just training them to use today’s chatbots but involving them in shaping how AI is deployed — will be the ones that retain institutional knowledge and public legitimacy. The rest will discover that voluntary pledges, like the technologies they are meant to tame, are only as strong as the incentives behind them.

---

Mei-Ling Zhao writes about technology policy, business culture, and civic imagination for Shadowfetch. She is based in Irvine, California.