Fifty-Three Per Cent Isn't a Backlash, It's Arithmetic

Fifty-three per cent of Americans now believe artificial intelligence will put someone in their household out of work. That figure is not a 'backlash', despite the label the business press prefers; it is a working hypothesis based on the observable reality that AI was the most-cited reason for U.S. job cuts in July, for the fifth consecutive month. The concern, as CNBC concedes, holds across demographic groups: no education, age or income bracket has been granted immunity from the suspicion that their role is next. A fortnight ago I noted that 45 per cent of under-35s had done the maths on AI and their careers. The adults appear to have started doing it too.

According to CNBC, the corporate response to this anxiety has evolved into something called a 'communication strategy'. Its cornerstone, we are told, is trust; its method is to let individual teams drive the AI roadmap; its purpose is to avoid alarming individual contributors. 'Trust remains the cornerstone of an ideal AI communication strategy', the article explains, as if trust were a design feature you could install after the beta. If you have ever been told your job is safe by a man whose bonus depends on reducing headcount, you will recognise the genre. The roadmap is not a plan for adoption; it is a plan for the delegation of blame — hand the steering wheel to the people who are about to be run over, so they can't say they weren't consulted.

The companion piece, from Axios, reports that chief executives are now weighing their public statements about workforce reductions 'more deliberately against the economic backdrop'. The change, the article notes, appears to be in the language and framing surrounding those cuts. The layoffs are still happening; the language is simply being tastefully lit. Challenger, Gray & Christmas, the firm that tracks such announcements, reports that AI was the most-cited reason for U.S. job cuts in July — the fifth consecutive month. Somewhere in a boardroom, a CEO is nodding sagely about 'economic headwinds' while the spreadsheet in front of him says 'Fte_savings_from_ai_initiative.xlsx'.

Dario Amodei, chief executive of Anthropic, described the industry's problem last week as 'fundamentally a crisis of trust'. He is right, in the sense that a man who has been struck by a bus is experiencing a crisis of velocity. The industry has spent years treating safety as a rhetorical exercise and workers as variables. Now it discovers that the public has noticed. The solution on offer is not to change the variables, but to change the memo: 'letting individual teams drive the AI roadmap' is the technical term for crowd-sourcing the bad news, and for ensuring that no single executive can be named in the resulting inquiry.

Consider what the strategy actually asks employees to do. You are told that your team is 'driving' the adoption of AI, invited to contribute to the roadmap, encouraged to find efficiencies. If your role is then eliminated, the AI did not fire you; your team did. The company just followed the roadmap you helped build. This is the most elegant blame-dispersal mechanism since the invention of the focus group, and it works until someone asks why the person who 'drove' the roadmap is also the person being driven out of the building. The same logic, extended upwards, gives us the CEO who tells CNBC that AI will make us more human while the CFO is entering 'labour optimisation' into a forecast.

None of this is to say that companies must choose between AI and jobs. Some firms genuinely use AI to augment workers rather than replace them. But if that is your case, make it — with numbers, timelines and severance commitments — rather than with a 'communication strategy'. And if your case is that AI replaces roles, have the decency to say so. The worker who fears for their job is not being irrational; they are responding to the most-cited reason for job cuts in America. The irrational party is the executive who believes a carefully worded email about 'transformation' will close the gap between what they say and what the payroll data shows.

The broader consequence is not merely organisational but political. A sector that cannot level with its own employees about headcount will not level with Congress about capability; the two evasions are the same failure. I wrote recently about the legislative scramble after OpenAI's rogue models, and about the industry's habit of denouncing every oversight proposal as catastrophic while insisting safety is its 'north star'. This is where that habit leads: to a public that assumes every reassurance is a euphemism and every euphemism is a lie. The only fix is to stop speaking in euphemisms — name the jobs that will change, publish the data that supports the decision, and let workers form their own view on your actual record, not your roadmap.

The industry will tell you the backlash is a misunderstanding, a communications problem, a generational quirk. It is none of those things. It is arithmetic. Fifty-three per cent of Americans have done it, and the number will keep rising until the people who run AI companies stop treating the workforce as a resource to be optimised. You can call it a trust crisis, a backlash, or an 'AI communication strategy' that is evolving. The rest of us will keep calling it what it is: a verdict.