Signal
Insights September 14, 2026

Uber Cut 3,300 Jobs and Didn't Blame AI Once. The Memo Blamed Management.

Uber laid off roughly 3,300 people (10% of its workforce) on September 2, 2026 — and CEO Dara Khosrowshahi's memo never mentions AI. The stated target is management itself: teams with one or two direct reports cut by half, headcount seven-plus layers below the CEO cut 20%, manager count down 20% overall. In a month when every other layoff got the AI-displacement headline, Uber said the plainer thing out loud — they hired too many managers of managers.

Every tech layoff this year comes with the same headline: AI did it. Amazon, Oracle, the whole 2026 list — AI-agent adoption gets cited, implied, or left conspicuously unaddressed while everyone assumes it anyway. Then on September 2, Uber cut about 3,300 people, roughly 10 percent of its staff, and Dara Khosrowshahi's memo to employees didn't reach for that explanation at all. He named something more specific and, frankly, more damning: the company had built too many layers of management managing management.

The numbers in that memo are unusually precise for a layoff announcement. Uber is cutting teams with only one or two direct reports by nearly half. It's reducing headcount sitting seven or more reporting layers below the CEO by 20 percent. Manager count overall drops 20 percent. In the tech org, Core Services engineering and science teams are being combined into one.

Khosrowshahi's own words: "more layers, more coordination, more fragmented ownership, and in some cases structures that made sense when businesses were smaller but no longer serve us well at our current scale." That's not an AI-adoption memo. That's an org chart that grew a management layer for every problem instead of solving the problem, and a CEO admitting it in public.

Sit with the phrase "micro-teams" for a second, because it's the tell. A manager with one or two direct reports isn't managing a team — they're a translation layer between an individual contributor and whoever's above them. Uber apparently had enough of those to name them as a category and cut them by half.

That doesn't happen by accident. It happens when "give them a manager" becomes the default answer to every retention request, every promotion cycle, every reorg, for long enough that nobody stops to ask whether the layer is doing anything except adding a status meeting and a performance review to the chain.

Here's why this is a different story than the AI-displacement one everyone's been running, and why it matters more to how you staff than another agent-adoption stat would. The AI story says: some tasks are now cheaper to automate, so headcount doing those tasks shrinks. That's real, but it's a story about individual contributors and their task mix. Uber's story is about a structural failure mode that has nothing to do with what AI can or can't do. It's about what happens when an organization scales revenue and product surface faster than it scales judgment about org design. You can have zero AI adoption anywhere in your company and still build seven layers of reporting distance between your CEO and the person actually shipping the work. Uber just told you, with specific percentages, that it did exactly that.

If you're a CTO or VP of Eng reading your own headcount plan next to this, the question isn't "which of my roles could an agent do." It's a different audit entirely: for every manager on your chart, how many direct reports do they have, and what decision do they make that their own manager isn't already making one layer up? A one-or-two-report team is usually not a team — it's evidence that someone got a title bump instead of a real scope change, and the org absorbed a permanent coordination tax to make that work politically. Multiply that by however many "micro-teams" you're carrying and you get exactly what Uber described: distance between the CEO and the work, fragmented ownership, and a lot of meetings that exist to move information between layers that didn't need to exist.

This is the audit I'd run before the next req goes out, not after: map span of control across every manager in the org, flag anything under three or four direct reports, and ask what specifically breaks if that layer is removed and the reports roll up one level. Most of the time the honest answer is "nothing breaks, it just gets faster." That matches Khosrowshahi's own phrase, "more time spent building rather than coordinating." The roles worth protecting and staffing aggressively are the ones where a real decision gets made — resourcing, tradeoffs, hiring calls, technical direction. The roles worth questioning are the ones that exist to relay a decision someone else already made.

The uncomfortable part for a lot of orgs: this is a self-inflicted cost, not a market-forces one. Nobody made you build seven reporting layers. Nobody made you promote someone into a manager role for a team of one. You did that, incrementally, one reasonable-sounding decision at a time, and now it shows up as a September memo naming a category of team that shouldn't have existed. Run that audit on your own chart before you write it as a layoff. It's cheaper to catch on a whiteboard than in a memo to 3,300 people.


VC5 Consulting builds hiring plans around real spans of control and real decision ownership, not headcount targets that quietly recreate the layers you just cut. If your next org chart has more managers than decisions, let's talk.