The Anticipation Tax: Paying for AI That Does Not Work Yet

Sixty percent of organisations have already cut staff in anticipation of AI. Only two percent tied those cuts to AI that was actually working.

That gap, between expectation and evidence, is not a rounding error. It is the defining feature of the current labour market. And it has a cost that falls almost entirely on the people least able to absorb it.

The numbers do not add up

A December 2025 survey by Thomas Davenport (Babson/MIT) and Laks Srinivasan, published in Harvard Business Review, asked over a thousand global executives about their AI-related workforce decisions. The headline finding is striking: 60% had reduced headcount based on AI expectations, while just 2% connected large-scale layoffs to AI systems that were actually deployed and functioning. A further 29% said they were simply hiring fewer people than usual, again based on expectations rather than results.

Read that again. The vast majority of AI-motivated workforce reductions are not responses to technological change. They are bets on technological change that has not happened yet.

Oxford Economics puts a finer point on it. In the first eleven months of 2025, roughly 55,000 US job cuts were attributed to AI, but that figure represented only 4.5% of total reported job losses. Standard market and economic conditions accounted for four times as many. Oxford Economics characterised the AI layoff narrative as "corporate fiction," a way for companies to frame cost-cutting as forward-thinking strategy rather than admitting to weak demand or post-pandemic over-hiring.

The anticipation tax

I have started thinking of this phenomenon as the anticipation tax: the real, immediate cost that workers pay today for efficiency gains that may never materialise.

Taxes, in theory, fund something useful. You pay now, you get services later. But the anticipation tax offers no such guarantee. The PwC 2026 Global CEO Survey found that 56% of CEOs say they have gotten "nothing out of" their AI investments. Only 12% reported that AI both grew revenues and reduced costs. An NBER study of 6,000 executives across four countries found that 90% of firms reported zero impact on employment or productivity over three years of AI adoption. Average executive AI usage was just 1.5 hours per week.

Robert Solow's 1987 observation, "You can see the computer age everywhere but in the productivity statistics," has been dusted off so often in the past year that it deserves its own LinkedIn profile. The Federal Reserve Bank of St. Louis has identified only 1.9% excess productivity growth since ChatGPT launched in late 2022. MIT economist Daron Acemoglu projects a "disappointing" 0.5% productivity increase over the next decade from current AI applications.

Workers are losing their jobs to fund a future that the data says is not arriving on schedule.

The Klarna parable

No company illustrates the anticipation tax more clearly than Klarna. Between 2022 and 2024, the fintech cut approximately 700 customer service roles, replacing them with an OpenAI-powered chatbot. CEO Sebastian Siemiatkowski was candid about the strategy, and the company was widely celebrated as proof that AI could replace entire departments.

Then quality dropped. Siemiatkowski later admitted the company had "focused too much on efficiency and cost," resulting in "lower quality" service. By mid-2025, Klarna was rehiring human agents under a hybrid model targeting students, parents, and rural workers on flexible schedules.

The AI was never as capable as the layoff announcements implied. But those 700 people still lost their jobs. The disruption to their lives, their finances, their career trajectories, that was real and immediate, even if the AI revolution that supposedly caused it was not.

Gartner now predicts that by 2027, half of companies that attributed headcount reductions to AI will rehire staff to perform similar functions under different job titles. Forrester Research expects a similar reversal, but with a darker twist: the rehiring will happen offshore or at significantly lower salaries. The company gets to pocket the savings. The workers bear the downside. The AI was just the justification.

The hidden bottleneck

There is an irony buried in the productivity data that deserves attention. Workday's 2026 research found that 37 to 40% of time supposedly saved by AI gets consumed by reviewing, correcting, and verifying AI-generated output. For software teams, AI adoption increased task completion by 21% and pull requests by 98%, but PR review time increased 91%.

AI is not eliminating work. It is shifting work. It turns production tasks into review tasks, and the review tasks still require human judgement, domain knowledge, and accountability. The net productivity gain, once you account for the verification overhead, is far smaller than the press releases suggest.

This matters because it undermines the core premise of anticipatory layoffs. If AI creates nearly as much oversight work as it eliminates production work, the business case for cutting headcount collapses. You still need the people. You just need them doing something slightly different.

Who actually pays

The anticipation tax is not distributed equally. Stanford and Dallas Fed research using ADP payroll data from 25 million workers found a 13% relative decline in employment for workers aged 22 to 25 in AI-exposed occupations since late 2022. Older workers in the same roles saw 6 to 9% employment growth.

The mechanism is not dramatic. Companies are not firing 23-year-olds and replacing them with chatbots. They are simply not hiring them in the first place. Entry-level positions are disappearing, not because AI can do those jobs, but because executives believe it soon will. An entire generation of workers is being locked out of the starting positions they need to build careers, skills, and experience.

This is where the anticipation tax becomes something uglier. When Wharton professor Peter Cappelli observes that companies announce layoffs "expected to occur due to AI" while admitting they "hadn't done it yet," we are looking at a labour market shaped by vibes rather than evidence. The workers who cannot get hired today, the customer service agents who lost their jobs at Klarna, the entry-level analysts who never got an interview, they pay a concrete price for an abstract projection.

The uncomfortable question

Block CEO Jack Dorsey laid off 4,000 employees in February 2026, nearly 40% of the company, and attributed the cuts directly to AI efficiency. He predicted that "the majority of companies will make similar structural changes" within a year. Oxford Economics and McKinsey noted, rather pointedly, that most firms are still experimenting with AI and have not scaled it.

I keep returning to a simple question: if 56% of CEOs say AI has delivered nothing, and 90% of firms report zero productivity impact, why are workers losing their jobs over it?

The answer, I think, is that "we are replacing you with AI" sounds like progress. "We over-hired during a boom and now demand is soft" sounds like a mistake. One story raises your share price. The other lowers it. The anticipation tax is not really about technology. It is about narrative, and the people who pay it are never the ones telling the story.