Barclays Is Trading Developers for Coding Agents and Turning Payroll Into an AI Budget

Barclays will cut about 5,000 jobs this year, with software developers among the roles going, as Britain's banks move money out of payroll and into AI.

Barclays says AI is making the staff who remain more productive. In July, its own research desk told clients the evidence for that wasn't there yet.

On October 1, Anthropic announced that Barclays expects half of its software developers to be using Claude Code, its coding agent, by the end of this year, and most of its engineers by 2027. Craig Bright, one of the bank's two group co-COOs, said Barclays would use it to modernize legacy platforms and improve the quality of its software.

Three weeks later, Barclays told investors it is cutting about 5,000 jobs in 2026, roughly one in every 20 people it employs. About 60% of those cuts fall in support functions, and CFO Anna Cross named developers among them, Reuters reported. Her explanation was that AI is raising the output of the people who stay.

Banks have spent two decades automating tellers, call centers and back-office processing. The engineers who wrote that automation were never on the list. At Barclays they are now, and the savings are being promised back to technology and shareholders. Across European banking, payroll is turning into the AI budget, and the people who build the systems are no longer exempt.

What Barclays announced

The cuts came out with third-quarter results. Pre-tax profit rose 12% to £2.3 billion, which still missed forecasts after a weak quarter in fixed income trading. About 1,000 roles were cut in the first half of the year, and the other 4,000 are due to go by December, according to Banking Dive.

Barclays now expects to spend about £600 million on restructuring this year, roughly twice what it planned, Quartz reported. It expects more than £1.1 billion in gross savings in 2027 and announced a £500 million share buyback the same morning. The results presentation credited the reductions to a simpler organization and to efficiency from AI in how the bank changes its technology, and described the whole program as lowering the cost base to free up money for investment.

CEO C.S. Venkatakrishnan tried to keep AI in proportion. He called it "the icing on the cake, not the cake," and said years of infrastructure work made the savings possible. On the same call, he said the bank would deploy AI at scale and that role reductions in favor of machines would speed up from here.

The automation has reached the people who build it

For most of the past 20 years, a bank's software engineers were the ones doing the automating. Their code closed branches and shrank call centers. Reuters called the Barclays disclosure one of the clearest signs yet from a big bank that AI is letting firms cut roles like theirs.

Barclays has roughly 30,000 technology staff, spread across hubs in Pune, Bengaluru, Glasgow, Manchester and New Jersey, by one estimate. The agent it is putting in front of them goes after the work that keeps big engineering teams busy inside old banks: maintaining legacy code and slowly replacing the systems underneath it. That work used to justify headcount. Now it justifies a license.

The developers who stay will write less code and check more of it. In a recent developer survey, 42% of respondents said AI now writes at least half their code, up from 12% a year earlier. The hours they saved didn't turn into lighter weeks. They went into review. A Meta engineer said this month he is already shipping code he doesn't fully understand.

That leads to a problem banks haven't worked out. Reviewing code and designing systems are senior skills, and engineers learn them by doing junior work first. Cut the junior layer and the savings show up this year, while the shortage of people who can catch an agent's mistakes shows up around 2030. It is the same trap we described when Wall Street started automating its own succession in the analyst class. Barclays has now extended it to engineering.

Payroll is becoming the AI budget

Barclays didn't describe this as a cost cut. It described it as moving money from salaries into technology, with a share going back to investors. Standard Chartered made the same case in May when it said it would cut 15% of its corporate function roles by 2030. Reuters put that at more than 7,000 jobs, and the bank paired it with a target of about 20% more income per employee by 2028, according to ARY News.

HSBC has been reported to be weighing around 20,000 cuts over three to five years, mostly in jobs that don't face customers. Reuters keeps a running list of layoffs tied to AI. By FourWeekMBA's count, the entries that give a number add up to about 102,800 cuts since October 2025, and seven banks account for roughly 39% of them.

When money moves from people to agents, the cost behaves differently. A salary is fixed and easy to forecast. Agent usage is metered, and it can run away from you. Uber capped engineers' monthly spend on AI coding tools after blowing through its budget, and Accenture now funds just $29 a month of Claude per employee. A bank that trades developers for agent capacity swaps a cost it understands for one it is still learning to predict.

It also changes where the risk sits. Engineering work that used to be spread across thousands of employees and several outsourcing firms starts to run through a handful of model providers. Regulators already watch how much banks depend on a few cloud companies. As AI vendors take on core engineering work, expect them to get the same attention.

Every restructuring is an AI story now

Barclays has made a cut this size before. In 2023 it removed about 5,000 roles, mostly in its support services arm and its UK operations function, City A.M. reported. The reasons given then were fewer layers of management and better technology and automation. Same number of jobs. This time the explanation leads with AI.

The bank's own economists have been careful about the productivity case. In July, Barclays analysts told clients they could find little sign that the US industries using AI most heavily had become more productive than they were before. Venkatakrishnan's line about icing points the same way. Much of what Barclays is saving comes from simplification and modernization that started long before anyone at the bank had heard of a coding agent.

None of that makes the cuts less real for the people losing their jobs. Unite general secretary Sharon Graham accused Barclays of using AI as an excuse to discard its workforce, according to the BBC. Investors, for their part, tend to reward any restructuring that arrives with an AI story attached.

That gives every company an incentive to call its cost program AI, whether or not AI did the work. It also gives unions and regulators a reason to start asking which jobs were actually automated and which were simply eliminated. Klarna has already shown how that can go wrong. It went back to hiring customer service staff after its AI cuts went further than the technology could carry.

What 2027 will show

Barclays has set itself a public test. By the end of 2027, most of its engineers are supposed to be working with a coding agent, and the bank is supposed to be booking more than £1.1 billion in gross savings. If it gets there without outages, stalled modernization projects or a quiet rise in contractor spending, every bank board in Europe will want the same plan.

Barclays will need to show results in its engineering as well as its headcount, meaning software shipped faster and old systems switched off. If HSBC, Lloyds and NatWest start naming developers in their own cut lists, this stops being one bank's bet and becomes how the industry staffs its technology.

The number in Thursday's headlines was 5,000. The bigger change is that a major bank now treats the people who write its software as a cost that software can replace. Whether that holds up depends on the one thing Barclays' own analysts say they haven't seen yet: evidence that the people left behind are getting more done.