Anthropic CEO Dario Amodei (Associated Press/Alamy)
7 min read
The scale of the disruption to the labour market from AI is becoming clear; we cannot leave it to big tech or populists to frame the debate about what to do about it, argues Roa Powell
If AI is powerful enough to turbocharge Britain’s economy, it is powerful enough to disrupt our labour market. Ministers must reckon with this dual reality.
This government is taking rapid AI progress seriously. They have committed to a world-leading AI Security Institute, invested £500m in the UK’s sovereign AI capability, and plan for the UK to be the fastest AI adopter in the G7. Ministers have described AI as “the defining technology of our generation”, “the engine of economic power” and an “industrial revolution in a decade”.
But the more seriously this government takes AI, the harder it is to justify silence on AI-driven job loss.
If AI is really going to be an “industrial revolution in a decade,” surely we should expect disruption on a similar scale with backlash akin to the Luddites. If AI is going to help streamline the “flabby” civil service, with government suggesting 62 per cent of the most junior civil servants’ work is automatable, surely we can expect our own bosses to follow suit and cut headcount.
According to Public First, two-thirds of UK adults already expect AI to contribute to unemployment, and as AI’s impact on jobs becomes more prominent in the public consciousness, so does the political cost of doing nothing about it.
One excuse for inaction is that forecasts on AI-driven job loss still vary widely. Our own analysis at the Institute For Public Policy Research (IPPR) imagines scenarios ranging from eight million UK jobs lost to no jobs lost at all. When the International Monetary Fund said that 60 per cent of roles in advanced economies were exposed, economists pushed back, pointing out that AI being capable of performing a task tells you little about whether that person’s job will actually be scrapped. Even the tech CEOs disagree with Anthropic’s Dario Amodei predicting AI could eliminate half of entry-level white-collar jobs in five years, only for Nvidia’s Jensen Huang to push back, claiming “you’re not going to lose your job to an AI, but you are going to lose your job to someone who uses AI”.
But the forecasts are moving in one direction and the evidence that AI will bring significant change to our job market is mounting. AI’s performance at real-world job tasks has more than doubled in a year. Evaluations show AI matching or beating a human at a full-day task – that is a task that would ordinarily take a human eight hours to complete – 71 per cent of the time. At the same time “agentic” AI is taking us beyond the chatbot interfaces most people know.
Given a goal and control of a computer, AI can be left alone to browse the web, draft and send emails, edit files and book meetings. Meta, Salesforce, IBM, Microsoft and BT have all attributed significant job cuts to AI and entire disciplines have transformed overnight, with top engineers at Anthropic and Open AI saying AI now writes 100 per cent of their code.
Once the impacts are here it will be too late, risking an outcome where the state is grinding into gear just as millions are out of work, tax revenues have collapsed and techlash has peaked
We are also starting to get a sense of how uneven impacts will be. New data from the Financial Times shows 60 per cent of high earners use AI daily compared to just 16 per cent of low earners, with women also using AI less. This makes them less equipped to adjust to a world where our bosses expect us to use AI and maybe pay us more as a result. Young people are also set to feel the brunt of this as entry level jobs are most exposed, and without opportunities to learn on the job they will struggle to reach the next stage in the career ladder. Compared to the industrial revolution, the geography of AI’s impact is expected to be flipped, with high earners in cities more exposed to automation while rural areas can look immune on the surface but be left out of the economic upside.
We shouldn’t expect concrete predictions on this to arrive until very late in the day. The sequence from AI getting more capable, to AI getting adopted and then people becoming displaced doesn’t follow automatically from highly capable AI. Adoption and displacement depend on legal certainty, human preferences, and the relative cost of human versus machine labour. But by the time concrete data arrives, disruption could be well under way.
With AI, the timing trap is brutal. It is not realistic for our government to make major spending changes before bigger impacts from AI have arrived, but once the impacts are here it will be too late, risking an outcome where the state is grinding into gear just as millions are out of work, tax revenues have collapsed and techlash has peaked.
The real challenge for a country like Britain is whether we can capture the economic windfall AI brings, either by attracting firms to the UK so they are part of our tax base, or by considering new progressive tax structures for massive AI-driven profits. Across a wide range of AI labour market scenarios, this will determine whether we have the money to help those most in need.
The UK is especially exposed on this. First, because the sectors most vulnerable to AI disruption, like financial and professional services, currently bring in our biggest tax revenues.
Second, because the companies set to reap the rewards sit largely outside of our tax base. We don’t have any of the technology giants like Google or Microsoft who are already reaping the profits from AI, nor do we have any of the frontier AI companies who are seeing some of the fastest growing revenues ever, like Anthropic which just hit $30bn in revenue, up from just $1bn in January 2025.
In practice, this means we need to lay some serious groundwork now, both politically and practically.
Practically, we need to prepare multiple plans to capture the value from AI. If frontier AI companies hoover up all the profits, we should consider taxes that target them specifically. If the gains spread to any company that uses AI, we will instead need to raise corporation tax to reflect that revenue no longer accrues to workers. And in the meantime, government-backed wealth funds can help us reap some of the rewards whatever happens, by spreading our investment across the AI landscape and redistributing this to workers who need support most.
None of these policies are possible overnight and, for lots of this, the UK won’t be able to go it alone. We already struggle to effectively tax big technology companies, and international co-ordination is essential to allow everybody to take a fair share. For this to work, we need to start detailed scenario planning now.
Politically, our government needs to develop a stronger voice on this issue. The political ground is shifting fast. Just last month, OpenAI published a blueprint promoting robot taxes, a national wealth fund and a four-day working week. Meanwhile, Amodei has written that existing tax systems will no longer make sense and that progressive taxation on AI companies may be needed. These are not the demands of trade unions or left-wing think tanks. When AI’s biggest winners are calling for redistribution, it is beyond time for government to take that seriously.
The cost of waiting is just too high. Waiting would mean ceding the debate to AI companies designing rhetoric to suit their public image, or to populists who are faster in finding a way of riding anti-AI sentiment but would have our economy stall while other countries race ahead. If government remains absent from this debate any longer, its ideas will arrive just as disruption escalates, public pressure builds and simplistic solutions become dominant.
