Barclays' 5,000 cuts: "AI" shouldn't be a polite word for cost cutting

When a bank says AI makes its people more productive and, in the same breath, announces it needs 5,000 fewer of them, which part of that sentence are we supposed to believe? Productivity is a measurable claim. A layoff is a decision. Barclays just welded the two together, and we think the weld deserves a close look.

Barclays is cutting deep while still making money

On Thursday, the bank announced it will cut around 5,000 jobs, roughly 5% of a workforce of 93,000, as Expansión reports. About 1,000 exits have already happened, and the remaining 4,000 are due before year-end. Chief financial officer Anna Cross said around 60% of the cuts hit support functions, and argued that AI makes the bank's staff more productive. According to Cinco Días, CEO C.S. Venkatakrishnan said the move results from years of technology investment, and that AI has reduced the need for certain profiles, such as developers.

The timing is the interesting part. Nine-month attributable profit rose 17% to £5,837 million, and the bank launched a £500 million share buyback. Third-quarter profit still came in below analyst forecasts because of weaker fixed-income trading. The bank also expects cost savings of £1,100 million in 2027.

The numbers describe savings, not productivity

Look at what has been disclosed: a headcount target, a savings figure and a shareholder payout. What we haven't seen, at least in the material we've reviewed, is any measure of output per worker, error rates, delivery times or service quality. The claim that AI makes staff more productive is an assertion, and the evidence offered is the list of exits.

That matters because the two stories are different. If AI genuinely lets a smaller developer team ship the same work, great: show us. If the bank needed to trim costs after a disappointing quarter and a 12% slide in its share price this year, that's a legitimate business reason too. But then call it that. Our view is that "AI" is becoming the most flattering label available for cost cutting, because it sounds like the future instead of like a spreadsheet.

Notice also who benefits. The cuts fall heavily on support roles, including developers, in a company that is profitable. Shareholders get a buyback, and in our view the people in the affected roles bear the cost. One analyst quoted by Cinco Días said the cuts show how far AI and automation could transform banking in coming years. That may be true, but "could" is doing a lot of work.

What a credible AI layoff would look like

We'd suggest a simple standard. Before attributing job cuts to AI, a company should publish at least one verifiable metric: tickets resolved per engineer, processing time per transaction, defect rates before and after. It doesn't need to expose trade secrets. A bank is a regulated institution, so it already knows how to report numbers that matter.

There's a risk argument too. As negocios.com notes, automation isn't free: it needs infrastructure, data protection, training and control systems, and the savings only work if they outlast those costs without hurting service or raising operational risk. Cutting people first and discovering the gaps later is, in our opinion, a gamble, and the people who lose it are customers and the remaining staff.

Is Barclays wrong? We can't say, and that's the point. What has been disclosed is a plausible story, not a proven one. We also wouldn't extrapolate: Standard Chartered has announced its own AI-linked exits, while HSBC says some roles are disappearing but new ones are being created. Each bank has its own cost structure.

Our verdict

Treat Barclays as the first big test of a rhetorical move you'll hear a lot more. When a company ties layoffs to AI, ask three questions: what got measurably better, who got the savings, and what happens if the tool underdelivers. If the answers are vague, read "AI" as "cost cutting" and judge it accordingly. Employees, regulators and investors are entitled to that evidence. Until it appears, the efficiency story is a promise with a very real bill attached, and someone else is paying it.