American Express Cuts About 5% of Global Staff to Prepare for AI
American Express said on Friday 23 October 2026 that it will eliminate about 5% of its global workforce, with a $336 million pretax restructuring charge booked in the third quarter. According to CNBC, the company had almost 77,000 employees at the end of 2025. Banking Dive, citing a company spokesperson, put the number of roles at about 3,500. The CFO said the aim is fewer management layers, so that processes can be automated in the coming quarters and years. The CEO said AI agents will only reach the company’s systems through approved API or MCP channels. For other businesses, the practical lesson is about order: simplify the approval chain first, then automate it, and decide exactly which systems an AI tool may touch.
In short
- American Express plans to cut about 5% of its global workforce, with a $336 million pretax charge in the third quarter.
- The reported headcount is about 3,500, but the workforce base differs between 77,000 and 78,000, so the exact figure is still open.
- The company says it will reduce management layers first, then automate processes in later periods.
- The CEO said AI agents will only be allowed into its systems through approved API or MCP channels.
- Before automating any routine process, map the approvals, systems and access rules.
What happened
According to CNBC, American Express is cutting about 5% of its global workforce. CEO Stephen Squeri disclosed the plan to analysts on the company’s third-quarter earnings call on Friday, 23 October 2026. The charge is $336 million pretax, and the company said part of it is funded from the gain on its sale of its stake in Global Business Travel Group, according to Reuters as reported by Quartz.
The headcount is not yet settled. Banking Dive reported that the company will cut about 3,500 of roughly 78,000 workers worldwide, based on a spokesperson’s statement. Bloomberg also reported, citing Squeri’s remarks on Bloomberg TV, that the company has 78,000 employees and that the move affects about 3,500 jobs. CNBC uses the year-end 2025 base of almost 77,000. Five per cent of 77,000 is about 3,850, so the exact number depends on which base is used. Neither CNBC nor the other outlets have published a final, reconciled figure.
The numbers at a glance
| Item | Figure reported | Reported by |
|---|---|---|
| Share of global workforce being cut | About 5% | CNBC; Reuters as reported by Quartz |
| Roles affected | About 3,500 | Banking Dive; Bloomberg |
| Workforce base, end of 2025 | Almost 77,000 | CNBC |
| Workforce base, current | About 78,000 | Banking Dive; Bloomberg |
| Restructuring charge, pretax | $336 million, third quarter | CNBC; Quartz |
Why the company says it is doing this
Chief Financial Officer Christophe Le Caillec told CNBC that the restructuring is meant to make the company leaner and more agile, and to prepare it for AI. He added that the company does not want to automate processes that are poorly structured. The plan is to reduce layers of management and widen the number of people each manager oversees, so that automation can follow in later periods.
Squeri framed the change as a way to accelerate innovation and to invest in business priorities. Bloomberg described the move as an effort to keep rising costs under control. Publicly, the company has not put a figure on AI-driven savings from this plan. Coverage of the earnings call described AI as a reason to prepare, not as proof that AI has already replaced these roles.
Why it matters for businesses that automate work
The most useful point in the announcement is the order of steps. Automating a chain of approvals that has too many handoffs usually keeps the handoffs and makes them faster. American Express says it is changing the structure first. Any business planning automation for billing, reconciliations or follow-ups can apply the same test before buying a tool.
The second point concerns access. Squeri said the company will only let AI agents reach its systems through approved API or MCP channels. The Model Context Protocol is an open-source standard for connecting AI applications to outside systems. Its own documentation describes it as a way to connect AI tools to data and tools in a standard form. For a business, that means the question is not only which AI tool to use, but which of its systems the tool is allowed to read and change, and through which connection.
What changes in practice for a business of any size
A five-person office and a group with thousands of staff face the same first step. They need a map of the work before anything is connected. The table below sets out the questions to answer for one routine process, such as supplier invoice approval, refund handling or month-end reconciliation.
| Question | Why it matters |
|---|---|
| How many people approve or touch this step? | Each extra handoff adds delay and a place for errors to hide. |
| Which systems hold the data, and who owns them? | An AI worker can only act on systems it is connected to, so the owner must list them. |
| Does each system offer a supported connection? | Supported connections are easier to log, limit and switch off than screen-scraping workarounds. |
| Which actions need a named person to approve? | Payments, refunds and changes to vendor details usually need a human sign-off. |
| Where does the record of each action go? | Without a log, a finance or HR lead cannot check what the tool did. |
Some practical limits apply. Automation cannot fix a process that nobody understands, and it cannot decide policy questions such as who should be paid a bonus or which exception is acceptable. Results depend on each business’s own volumes and error rates, which need to be measured before any change is made. If any part of the plan reduces roles, pay or hours, the notice, consultation and severance rules of the country where those staff work apply, and the change should be agreed with the employee or checked with a lawyer.
What to watch next
- The final headcount, and whether the company confirms the base it used for the 5% figure.
- Whether American Express quantifies any savings from the plan in its later results.
- How the company describes its AI agent rules in future earnings calls, including any further channels it approves.
- Whether other card issuers and banks announce similar structure-first plans, as Barclays has reportedly been reported to be doing in separate coverage.
This explains the rule in general terms; it is not legal or tax advice for your case.
Sources
How AiStaffo would automate this
For a business that wants routine work automated without a restructuring programme, AiStaffo would start with one chain, such as supplier invoice approval, and connect only the systems that process needs through supported connections. The AI worker would read incoming invoices, match them against orders, route exceptions to the named approver and write a record of every action. The owner still sets approval limits, decides which exceptions need a person, and signs off any change to who can approve what. The free audit measures the current steps, systems and volumes first, so the owner sees what would run automatically before anything is connected. Book a free automation audit.
Questions people ask
How many jobs is American Express cutting?
Why is American Express cutting jobs if it is profitable?
What does the CEO mean by approved API or MCP channels?
Should a smaller business cut management layers before automating?
Has American Express reported savings from AI yet?
Book a free automation audit
Thirty minutes. We look at one process you run every week and tell you exactly what an AI worker would take off your desk, and what it would not.
