On 9 April 2026, the Financial Times ran as a top story that BDO, the UK's fifth-largest accountancy firm, was cutting 31 partner roles. The framing in the headline was direct. Partners were being removed because AI pressure was growing and profits were falling. The two causes were placed side by side, as if their combined force settled the question.
Before accepting that framing, it is worth separating the two halves. Profits at BDO have been under visible strain for some time. Accountancy Daily reported last year that average partner pay at the firm had dropped by 14 per cent in a single year, from around £681,000 to £589,000, while partner numbers had grown to more than 500. That is a firm whose economics were stretched before any AI narrative arrived.
A Wider Pattern
BDO is not an outlier. It is part of a pattern that has been visible across 2025 and the first quarter of 2026, in which companies attribute job losses to AI capability gains that, on closer inspection, do not always hold up.
Klarna became the most cited case study of AI-driven efficiency in 2024, announcing that its AI system was doing the work of 700 customer service agents. By 2025 the company had quietly begun rehiring customer service staff. Salesforce cut around 4,000 customer support roles on the basis that AI agents were handling around half of interactions. Its own filings show total headcount rising across the same period, from 72,682 at the end of 2024 to 76,453 at the end of 2025. The story told to the public and the story told to the labour market were not the same story.
In April 2026, Sam Altman observed in a public interview that there was what he called AI washing, where organisations were blaming AI for layoffs they would have undertaken anyway. Marc Andreessen, writing separately, argued that the current wave of cuts was largely a correction to pandemic-era overstaffing that in some companies reached 75 per cent above steady state. According to data summarised by Tom's Hardware, tech companies laid off around 78,500 workers in the first quarter of 2026, of which roughly half were attributed in company communications to AI and workflow automation. Whether that attribution is accurate in half of those cases is a different question.
The Reasons Behind the Reason
Why does it matter which explanation we accept? Because the explanation shapes what is asked of the workforce that remains.
If the story is that AI is systematically absorbing work, the rational response for a surviving worker is to compete with the machine on its own terms: faster, cheaper, more automatable. If the story is that organisations are rebalancing from pandemic-era hiring, redirecting cash to data centre capital expenditure, and using AI as a narrative frame for decisions that have other drivers, then the rational response is different. It is to ask what value a human workforce actually provides, and to insist that the answer is articulated honestly.
IBM, whose early AI announcements were widely read as a forecast of white-collar job loss, has this year tripled its entry-level hiring. The company's stated reason is that AI can handle many entry-level tasks but still needs what it called a human touch. That is a significant shift in register from the public commentary of eighteen months ago.
What Machines Do Not Yet Do Well
In late 2024, researchers at MIT Sloan, Isabella Loaiza and Roberto Rigobon, published a framework they called EPOCH: Empathy and emotional intelligence; Presence, networking and connectedness; Opinion, judgment and ethics; Creativity and imagination; Hope, vision and leadership. The researchers' argument, supported by a mapping of these capabilities across occupations, was that these five clusters represent work that AI is poorly placed to substitute for and well placed to complement.
The EPOCH taxonomy is not a consolation prize. It describes the capabilities on which professional work, at its most valuable, actually depends. A partner at a mid-tier accountancy firm does not earn £589,000 for processing tax returns. She earns it for judgment, for relationships with clients who trust her, for the ability to hold a difficult conversation with a chief executive, and for the presence to know when a number that looks right is telling her something wrong. Those are EPOCH capabilities. They are also the capabilities that a headline about 31 partner cuts erases from view.
Opinion: The Story We Are Being Asked to Accept
Every round of job losses comes with a story attached. The story attached to this round is that AI is pressing down on the firm and that people have to go. It is a tidy story, and it has the advantage of making the decision look inevitable. Decisions that look inevitable do not invite scrutiny.
There are other stories available, and they are less tidy. Post-pandemic over-hiring. The need to free up cash for capital programmes. Margin pressure unrelated to AI. Partnership structures that have quietly grown beyond the revenue that can sustain them. Each of these is a more boring explanation than artificial intelligence, and each is probably closer to what is actually happening inside more firms than are willing to say so.
The EPOCH research points, in a different direction, to something that ought to change how we read announcements of this kind. The capabilities that are most difficult to automate are also the capabilities that give professional work its economic weight. Empathy. Presence. Judgment. Creativity. The imaginative capacity to hope. When an organisation says that AI is taking work previously done by human beings, the question to ask is which work. If the answer is a set of EPOCH-heavy tasks, scepticism is warranted. If the answer is a set of routine tasks, the more interesting question is what the humans will now be freed to do.
Has human talent ever been more valuable than it is at the moment when the unique contours of human capability are, for the first time, being measured against a serious alternative? Or are we living through a period in which the real shape of that value is being obscured by a story that is easier to tell?
Sources
"BDO axes 31 partner roles as AI pressure grows and profits fall" — Financial Times, 9 April 2026 (headline story, FT front page)
BDO partner pay down 14 per cent to £589k — Accountancy Daily, 2025
The EPOCH of AI: Human-Machine Complementarities at Work — Isabella Loaiza and Roberto Rigobon, MIT Sloan, November 2024
New MIT Sloan research suggests that AI is more likely to complement, not replace, human workers — MIT Sloan, 2024
Tech industry lays off nearly 80,000 employees in the first quarter of 2026 — Tom's Hardware, 2026
