Britain has an AI strategy, a skills strategy and an investment pitch, but the picture is less settled than ministers would like to suggest. Recently, OpenAI paused its main UK data-centre project in April over high energy costs and what Reuters described as an unfavourable regulatory environment, a reminder that ambition and delivery are not the same thing. The wider UK-US technology relationship has also been less straightforward than the headlines first implied: a major bilateral tech pact was announced last year, then stalled before talks restarted this year. That matters because the government is asking the country to believe two things at once: that Britain can attract large-scale AI investment and that it can adapt its workforce quickly enough to live with the consequences. It has a plan for the first part – a plan is good, but execution may be a significant challenge as shown with OpenAI’s in-out investment messaging. However, it does not yet have a convincing tax plan for the second and this is a problem.
The government’s own January 2026 labour-market assessment is careful, but it is not comforting. It says the evidence still does not answer many of the biggest policy questions. It also says UK job postings in more AI-exposed occupations have shown signs of decline, with a one standard deviation rise in AI exposure associated with a 3.9% reduction in posting volume during the period studied, even though postings later returned to earlier levels. That does not prove an employment collapse is under way, indeed unemployment fell to 4.9% in the three months to February, the Office for National Statistics said, defying expectations that it would hold at 5.2% – a blip attributed to students going back to university. It does show, however, that the labour-market effects are real enough to justify something more serious than cheerful talk about productivity.
A recent BBC interview with former prime minister Rishi Sunak sharpened the argument. His point was not simply that AI may flatten hiring in law, accountancy, creative work and other service sectors. It was that the tax system still leans heavily on employment even as technology makes it easier for firms to expand output without increasing headcount at the old rate. His suggested answer was to abolish National Insurance over time and shift more of the burden towards corporate profits that rise with AI-driven productivity. Whether or not that exact prescription is right, the underlying question is hard to avoid: if labour becomes a smaller share of how businesses create value, can the tax base remain tied so heavily to labour? The government has not answered that question in a serious way. The official emphasis is still on diffusion and reskilling.
The current strategy is easy enough to describe. The government says it wants Britain to become both a major AI maker and a major AI user. The AI Opportunities Action Plan and its January 2026 update focus on compute capacity, adoption across the economy, public-sector deployment and workforce readiness. The flagship labour response is AI Skills Boost, which aims to upskill 10 million workers by 2030, alongside broader work on AI foundation skills, curriculum reform and place-based programmes such as TechLocal. That is not trivial, you could argue that it’s ambitious. It may help some of the labour market adapt more quickly than pessimists expect. But it is still largely a labour-supply answer to what may also become a tax-base problem. If AI and robotics allow employers to hold headcount flat while lifting output, the state risks losing labour-linked tax revenue at the same time as unemployment, retraining costs, welfare spending and wider regional adjustment pressures begin to rise. Once that starts to happen, the issue is no longer just productivity. It becomes a question of tax receipts, welfare costs and who carries the bill for adjustment.
There is another reason this matters now. Britain has recently made employing people more expensive, not less. Since April 2025, the employer Secondary Class 1 National Insurance rate has been 15% rather than 13.8%, and the secondary threshold was cut to £5,000 from £9,100. Ministers softened that move by raising the Employment Allowance to £10,500 and removing the old £100,000 eligibility cap, which helped many smaller employers. Even so, the broad direction is obvious. The tax system continues to place a direct charge on payroll while placing no equivalent charge on a machine, model or workflow that replaces part of that payroll. That may be defensible if automation quickly creates matching new jobs and fresh tax receipts elsewhere. It looks less secure if hiring at the bottom and middle of some professions starts to thin out while profits and productivity gains are concentrated higher up the corporate chain.
This is where the debate is often mishandled. A crude “robot tax” sounds simple, but simple is not the same as sensible. Taxing every machine that substitutes for a person would create obvious practical problems. It would also be a nightmare to define in law. Britain has already seen how absurdly technical tax classification can become; the long-running Jaffa Cakes dispute turned on whether they were cakes or biscuits because chocolate-covered cakes are zero-rated for VAT while chocolate-covered biscuits are taxed. A robot tax would invite the same kind of borderline arguments, is a self-scan machine at Tesco a robot? This type of nuance will only lead to confusion and delay only with far more money and complexity attached. A supermarket that uses robotic cleaners, shelf scanners or automated stock systems is not necessarily doing something socially harmful. The real fiscal question is narrower and harder: if a company reduces payroll growth through AI or robotics while lifting profits and output, should more of that gain be captured for transition support? That points less towards a per-machine levy and more towards a rebalancing of how labour, profits and automation-driven gains are taxed.
That is why Sunak’s focus on National Insurance is more serious than it first sounds. National Insurance is, in effect, a tax on employing people. If the economy is moving towards flatter payrolls, then a tax system centred on labour becomes more brittle. The challenge is not only one of fairness. It is one of fiscal durability. If a law firm hires fewer junior staff because document review and drafting are increasingly automated, the Exchequer loses income-linked revenue from work. If a retailer automates cleaning, stock checks or front-of-house tasks, it may save costs while reducing payroll growth. The technology can improve productivity and still leave the Treasury with a weaker labour tax base. If displaced workers then struggle to find replacement work quickly, the state also faces a higher welfare bill. That is the full problem: lower employment-linked receipts on one side, higher support costs on the other. Skills policy may help workers adapt. It does not solve that fiscal squeeze on its own.
The government’s own evidence base is awkward enough to justify starting that discussion now. The January 2026 labour-market assessment does not claim that mass unemployment is a settled outcome. It stresses uncertainty. It also points to projections that jobs directly involving AI activities could rise from 158,000 in 2024 to 3.9 million by 2035, roughly 12% of the current workforce, and says a broader group of 9.7 million people may end up in AI-related occupations. That sounds encouraging until one reads the detail. Much of that increase is expected to come not from entirely new roles, but from existing roles acquiring AI-related tasks. The headline growth in “AI jobs” does not settle the question of what happens to entry-level and mid-skill work that can be automated faster than workers are reabsorbed. The same package of government research also says many stakeholders felt the UK is not currently supporting people enough to develop future-relevant AI skills.
That leaves three broad policy options, none of them especially comfortable. One is to keep the current approach: push adoption, back skills, and trust that the labour market adjusts without a major fiscal redesign. That is the least disruptive route, but it also assumes the most. A second is to rebalance gradually away from taxing labour and more towards profits, as Sunak suggests, on the logic that AI-driven efficiency should help fund the transition costs it creates. That has the advantage of targeting the economic gain rather than the physical machine, but it would raise difficult questions about investment incentives, tax competition and how to identify AI-linked profit growth in practice. A third is to create a dedicated transition mechanism: not a robot tax in the cartoon sense, but a fund financed from a broader slice of corporate gains in sectors where automation materially reduces labour demand. That could support retraining, wage insurance, local adjustment or temporary hiring incentives. The government has not chosen any of these paths explicitly. At present it is still speaking the language of readiness rather than redistribution.
For business readers, the important point is not that a new levy is inevitable. It is that the tax neutrality of automation will not stay politically untouched if visible displacement rises while payroll taxes remain high. Companies using AI to improve output per worker should expect the argument to shift from enthusiasm about productivity to questions about who captures the upside and who absorbs the disruption. The state is encouraging firms to adopt AI, funding skills and building compute, yet it still relies heavily on taxes linked to human work. That contradiction is manageable while displacement is uncertain and uneven. It becomes harder to defend if the flattening of entry-level hiring spreads across finance, legal services, retail and administrative work.
The right conclusion is not that Britain should slow AI or slap a blunt charge on every robot. It is that an AI growth strategy without a tax strategy for displacement is unfinished. If ministers believe AI and robotics will raise productivity, they need to decide how much of that gain should be recycled into cushioning job loss, retraining workers and protecting the tax base. That could mean the kind of longer-term NIC reduction Sunak floated, funded by greater reliance on profits that rise with automation. It could mean a more explicit levy on firms where labour substitution is clear and large-scale, especially if robotics begins to replace lower-paid physical work as well as white-collar entry roles. But it cannot mean pretending the present system will remain balanced if payroll growth weakens while automation gains compound. Britain is preparing workers to use AI. It still has not shown how it would pay for the damage if AI and robotics also leave fewer of them in work.
Rishi Sunak’s BBC Interview
More from Finace Monthly: UK has a growth problem
More from Finance Monthly: Anthropic Amazon Compute Deal: What the $100 Billion Commitment Says About AI Infrastructure Power
#Answer #Jobs #Shock #bring