American Express plans to cut about 3,500 jobs as it expands AI

What happened

American Express plans to cut roughly 3,500 jobs, about 4.5% of its workforce, as it aims to become more agile and efficient while expanding its use of artificial intelligence, according to chief executive Stephen Squeri.

Banking Dive reported the reduction and tied it to the company's wider effort to change how it operates. The reporting centres on Squeri's comments on the direction of the business. It does not provide a breakdown of which roles, teams or business units will be affected, so those details remain unconfirmed.

The decision places a major card issuer's workforce plan directly alongside its AI strategy. That link is what makes the announcement relevant beyond banking, because it shows AI being discussed as a reason for structural change in headcount rather than only as a tool for individual tasks.

Why it matters

This is primarily an operating-model story. When a company connects a reduction of this size to greater use of AI, it signals that the technology is being treated as a lever for how work is organised and delivered. For leaders in financial services, the question is shifting from whether AI will change service delivery to how redesigned processes will be governed, measured and staffed.

For customers, the experience stakes sit in the places where capacity is most visible: servicing, complaint handling and high-stakes interactions such as disputes or fraud. Any change to these journeys needs to be tested against real demand, not against the capacity a planning model assumes. The reporting does not say how Amex intends to manage service levels during the transition, so that remains an open question for customers and for the wider market.

By the numbers

• 3,500 roles: the approximate number of jobs the company plans to cut.
• 4.5%: the reduction as a share of the company's workforce.

The Renascence take

Headcount reductions framed as efficiency gains are easy to announce and much harder to deliver well. The experience cost usually appears later, in longer resolution times, more repeat contacts and a gradual erosion of trust, rather than on the day the decision is made.

Reducing headcount changes the service system, not just the cost base. Remaining staff absorb the knowledge and goodwill that departing colleagues held, and customers feel that loss through slower, less joined-up service. Operators who treat AI as a substitute for judgement, rather than a support for it, will find that the savings arrive faster than the service problems they create. Before cutting, map every journey that depends on the roles being removed, set explicit service-level guardrails, and measure whether automated interactions actually resolve the customer's problem.