Uber Cut 10% of Its Support Team. The Handoff Is What Breaks.
· CX Pulse
Uber is cutting 10% of its support workforce to embrace AI. What small businesses should actually take from it, and a twenty minute test to run on your own front line.
Uber told its staff last Wednesday that it is cutting 10% of its customer service workforce. The two reasons it gave were "simplify operations" and "embrace AI."
Most of the coverage has been about the jobs. For anyone running a small business there is a narrower and more useful question. What happens to a customer after a company does that?
The detail sitting in the same report
Marketplace covered the cuts on July 23. In the same piece it quotes Jordan St. Laurent, a social worker whose job means calling hospitals, courts and schools all day. Every one of them now puts an AI agent in front.
What he says about those agents: "Nine times out of 10, it just ends up having me have to reach out to a live representative."
That is one person's account and not a statistic, and we are not going to dress it up as one. But it describes something familiar to anyone who has run a queue. The automation handles the easy half. The hard half still lands on a person.
Automation changes the mix, not just the volume
This is where teams get caught out, and it has nothing to do with whether the bot is any good.
When an AI agent absorbs the simple contacts, password resets, order status, hours and locations, it does not remove a random 30% of the queue. It removes the easiest 30%. What is left is denser: the angry ones, the edge cases, the billing disputes, the things with no help article to point at.
So the average ticket that reaches a human gets harder and takes longer. If you sized your team against last year's average handle time, you have just understaffed the hardest work in the building.
Uber can absorb that. It has scale, tooling and a workforce planning function. A ten person company running support out of a shared inbox does not.
The thing that actually fails is the exit
In most of the bad AI support stories people tell, the complaint is rarely that the bot gave a wrong answer. It is that there was no way out of it.
The bot could not answer, would not say so, and offered no route to somebody who could. The customer ends up repeating themselves into a form, or hanging up, or writing the review.
An escalation path is only real if somebody is standing on the other end of it. Cut the people without touching the routing and that path is still on the flowchart, still in the vendor demo, and dead in practice.
A test worth twenty minutes
None of us have 10% of a support team to cut. The same pitch is coming for our front lines anyway, usually as a chat widget with a monthly fee. So before buying anything, run this against what you already have.
Go to your own website, phone line or inbox the way a customer would. Then count.
- How many steps to a human? Count the clicks, menu levels and messages it takes to reach a person. Whatever that number is, it is what your customers are living with today.
- What happens when it does not know? Ask it something genuinely outside its scope. See whether it admits that, invents an answer, or just loops.
- Who is on the other side? When it does hand off, find out where that lands and how long it sits there. An inbox nobody owns is not an escalation path.
If any of those three answers surprises you, that is the work. It costs nothing to check and it does not require buying a thing.
The short version
Uber is removing 10% of the people who catch what the automation drops. Worth knowing how many of those people you have, and where they sit, before something drops.
Source: Marketplace, "An AI customer service agent will take your call now," July 23, 2026