AI Layoffs at Barclays and Amex: Two CEOs, Opposite Stories

Barclays will cut about 5,000 jobs this year, and 60% of them fall in support functions that include software development. Over the same stretch, the bank plans to put Anthropic's Claude Code in the hands of half its software developers. Its chief executive says AI is a side story.

One day later, American Express announced about 3,500 cuts of its own, framed in the coverage as cuts made to prepare for AI. That puts 2 of the year's biggest AI layoffs in the same earnings week, at companies deploying AI at scale. Their chief executives tell almost opposite stories about why. The last time two companies cut jobs the same week, neither blamed AI. This time, one did, and neither story quite fits the math.

Barclays calls AI the icing

C.S. Venkatakrishnan described the bank's AI plans this way when Barclays reported third-quarter results on October 22:

"We will employ the AI at scale but AI is the icing on the cake not the cake."

The icing is thick. Barclays expects most of its developers to be using Claude Code by the end of 2027. More than 16,000 UK employees already use an internal AI knowledge assistant, and AI models in Global Markets process 120,000 client emails a day. The bank spent £307M on restructuring in the first 9 months of 2026 and expects about £300M more by year end. The cuts land hardest in the functions the AI is moving into.

Amex calls AI the reason

Stephen Squeri took the other line. Amex will cut about 3,500 of its roughly 78,000 employees, close to 4.5%, and recorded a $336M restructuring charge this quarter. Squeri said "we needed to become a more agile and efficient organization" to win over the next few years. He pointed to Brex and Ramp, fintechs that use AI agents to compete for corporate customers, and said more agents are coming to the Amex platform.

The AI he described in detail is ChatGPT writing personalized pitches for premium restaurant and travel customers who book through Resy. That is useful work. It does not obviously account for 3,500 people, and an Amex spokesperson declined to say which roles or divisions the cuts will reach.

Put the two side by side and the stories cancel each other out. One chief executive plays AI down while the cuts follow it into the org chart. The other credits AI while describing uses that would not require cuts this size. Underneath the messaging, both companies are making the same purchase. Kroger cut thousands of corporate roles this month on the same logic. Three companies, six weeks, one pattern.

The same AI is on its way to every rival

Claude Code is available to any bank that signs a contract. The agents Squeri watches at Brex and Ramp run on models anyone can license, and Capital One bought Brex this year. The distance between the best AI models and the rest keeps closing: Stanford HAI's 2025 AI Index found that the spread between the top-ranked and 10th-ranked model on the Chatbot Arena leaderboard fell from 11.9% to 5.4% in a single year.

If that pattern holds, an AI coding assistant becomes something every bank has, the way every bank has email. It raises the floor for the whole industry at once, which means it stops being anyone's edge. What a company still owns, after you subtract the AI its rivals can buy too, is its people. That means the developer who knows why a payments system written decades ago behaves the way it does. It means the servicing team that knows what a premium cardholder will forgive, and what she won't.

Everyone will have the same AI. Not everyone will have the same people.

That is the argument behind my Will AI Take My Job page, which I update every time a company makes this trade.

The people who check the AI are on the list

AI-written code still needs someone who can tell when it is wrong. At a bank whose roots go back to 1690, that someone is a developer carrying years of context about systems no model was trained on. Thin that bench far enough and Barclays ends up renting the model and renting the opinion about whether the model is right. Call it rented judgment. It is the hardest dependency to price, because it never shows up on the invoice.

The clocks do not match either. A restructuring charge clears in a quarter. The judgment that walks out with it took a decade to build, and no vendor sells it back. I spent 20 years at Cisco, where I shaped a $1.1B innovation portfolio, and the lesson still holds: the technology is always the easiest part for a rival to match. The people who make it work inside your own systems are not for sale.

What this week means for your own job

If you are a developer or a support analyst reading this week's news, you are looking at the relevance cliff. That is the moment a skill that pays your mortgage starts to look like something your employer can rent. It is a hard week. It is also survivable, and the people who come through it are the ones who move toward checking the AI's work. Ask your manager a plain question: which part of my job will the AI do, and which part of the AI's work will I be trusted to check? The second half of that answer is where your next decade lives.

For leaders, the question comes before the cut list, and it is the one the book I am writing now, The Great Relearning, keeps returning to: once every rival has bought the same AI, which of our people are still the reason a customer chooses us? Check whether those names are on the list. If they are, the restructuring spends the hardest advantage to copy on one every competitor can buy too.

Barclays and Amex will both have strong AI by 2027. So will the companies they compete with. Which of them will still have the people who can tell when it is wrong? Take that question into your next leadership meeting. And if you see it differently, I'm easy to find.

Here is what makes Alex a credible voice on this topic: Alex spent 20 years at Cisco. There he built a $1.1B innovation portfolio that generated more than $400M in revenue. He now advises leaders on a hard question: which people to keep when every rival can buy the same AI.

Weighing AI spending against headcount at your company? Talk with Alex →