American Express cut 5% of its workforce as it prepares for AI
A card company’s restructuring charge puts a sharp question before small teams: which routine tasks should software handle, and which still need judgment?
The 5% that landed on Stephen Squeri’s earnings call
American Express became the latest large employer to put AI beside a workforce cut, and Reuters reported Friday that the plan will eliminate about 5% of its global workforce.
The easy reaction is to treat that as a distant corporate story, involving people, budgets and systems far larger than a restaurant, studio or retail shop will ever have.
That would be too comfortable.
A business with two trainers and 200 members does not have a restructuring department. It has a booking page, an inbox, a WhatsApp thread and someone trying to keep all four from becoming a pile of missed follow-ups.
The pressure is closer than it looks.
American Express tied its restructuring plan to AI preparation
Reuters reported Friday that American Express took a restructuring charge in the third quarter for a plan expected to eliminate about 5% of its global workforce, after chief executive Stephen Squeri told analysts on a post-earnings call. Reuters reported the plan here.
CNBC reported Friday that American Express is eliminating 5% of its global workforce and that Squeri said the company is preparing to automate some internal processes using AI. CNBC also reported a $336 million restructuring charge in the quarter. CNBC’s account is here.
The distinction matters. This was not a software company selling a promise about what AI might do one day. It was a large card company tying a workforce decision to internal automation in the same breath.
That’s the receipt.
A card company is not a small service team
Large companies can make a mess at scale. They can also run an experiment worth noticing, because their internal work contains the same unglamorous friction that drains a small team: repeat questions, copied details, slow routing and people chasing status updates.
We think the useful part of this news is not the workforce cut. It is the admission that internal process work has become a serious enough target for automation to shape a major company’s planning.
Ignore the $336 million charge. It is a corporate accounting detail, not a lesson for a local operator deciding whether an enquiry should receive a reply while the team is busy serving customers.
The unknown is which internal processes American Express will automate first, and whether those systems will make customers’ experience better or simply make the company cheaper to run.
That gap is the whole story. Automation that clears a backlog can be useful. Automation that sends a confident wrong answer can turn a small complaint into a public one.
The sensible lesson is narrower than the headlines: copy the discipline, not the scale. Decide which work is repetitive enough to standardise, then keep a person responsible for the moments that need judgment. For a longer view of where the money around AI is flowing, read Is AI a Bubble? What Smart Investors Really Think.
The $336 million charge is the part to ignore
Big employers can spend heavily before they learn what works.
Small teams cannot.
But they can choose one narrow process, make its rules visible, and stop letting routine enquiries depend on whoever happens to check a phone first.
That is the turn.
Your booking page, your inbox and the work only you should keep
Start with the places where customers already wait. Not a grand AI plan. Not a new dashboard that nobody opens after the first week.
Use your Friday afternoon to trace three surfaces:
- Your booking page: submit a test enquiry, then write down what the person receives, when they receive it and who owns the next reply. If you skip this, you may discover that a customer gets no answer until the next day, after they have booked elsewhere.
- Your WhatsApp: list the five questions that appear most often, then prepare approved replies that give a clear next step and flag anything unusual for a person. If you skip it, staff will keep writing slightly different answers, and customers will keep receiving different promises.
- Your inbox: create a simple rule for enquiries that need a response, a booking link or a human call. If you skip it, the urgent message will continue to compete with newsletters, supplier receipts and internal chatter.
This is where AI automation for small business is worth trying: drafting a first reply from information you have already approved, sorting messages by intent, and reminding a person when a lead has gone quiet.
It is not worth handing an AI assistant the final word on refunds, health questions, a booking dispute or a complaint that has emotional heat. Those are not repetitive transactions. They are relationship moments, and the cost of getting them wrong is higher than the time saved.
You probably will not do this because it sounds too basic. A new tool feels like progress; opening your own booking form on a busy afternoon feels like admin. But the basic work is where the leak usually is.
We would set one standard: every incoming enquiry gets either a useful immediate response or a clearly assigned human owner. That does not require a developer, a large subscription or a complicated rebuild. It requires deciding what the first reply may say and where uncertainty must stop the automation.
If fast replies are the real problem, Speed to Lead Australia: Turn Enquiries Into Bookings Faster is the useful next read. If the question is whether reviews reveal the damage from missed follow-up, How Google Reviews Decide Business Success gives that issue more attention.
Watch whether American Express names the internal processes it automates
Watch for a specific signal: American Express naming the internal processes it actually automates, rather than repeating a broad commitment to AI.
If the company points to routine service, document handling or internal routing, that would strengthen the case for small teams to automate narrow, repeatable work first. If it stays vague, the workforce cut will remain clearer than the capability behind it.
And watch the customer-facing result. Faster resolution and fewer handoffs would matter more than corporate language about efficiency.
The test is simple: does automation remove waiting, or just move it somewhere customers cannot see?
Frequently asked
What do American Express AI layoffs mean for small businesses?
The reported plan concerns American Express’s internal processes and workforce. Small businesses can use the news to review repetitive tasks such as enquiry replies and booking follow-up.
Is AI automation for small business safe to use with customer messages?
It can be used for routine replies based on approved information. Messages involving complaints, refunds, health questions or disputes should remain with a person.
What should I automate first in a small team?
Start with repeat questions, initial enquiry responses and reminders for unanswered leads. Keep a clear handoff point for requests that need judgment.