Meta Reverses Manager Cuts in Its Applied AI Division
Meta is asking individual contributors in its Applied AI division to volunteer for manager roles, reversing a 2023 flattening drive. Here's what happened and why it matters.
Meta is inviting individual contributors inside its Applied AI (AAI) division to voluntarily return to manager positions, according to reporting from Business Insider and Fortune. The opt-in program partially reverses a multi-year "flattening" effort Mark Zuckerberg launched in 2023, when thousands of managers were pushed into contributor roles or out the door. The move affects a unit that only recently took shape: roughly 7,000 employees were reassigned into the AAI division in 2026 after May layoffs cut about 10% of Meta's workforce, around 8,000 people.
What happened
| Data point | Figure |
|---|---|
| Employees cut in May 2026 layoffs | ~8,000 (roughly 10% of workforce) |
| Open positions scrapped at same time | 6,000 |
| Employees reassigned into AAI division | ~7,000 |
| Meta headcount at end of Q2 2026 | 75,472 (down 3% quarter-over-quarter) |
| Q2 2026 revenue | $60.8 billion (up 28% year-over-year) |
| Q2 2026 expenses | $42 billion (up 55% year-over-year) |
Meta is running an opt-in program that invites individual contributors in its Applied AI division to step back into team-lead roles. Business Insider broke the story; Fortune relayed it. Some of the people now being asked to manage again previously held manager titles before the company stripped those titles through its internal “flattening” initiative.
That flattening effort dates to 2023, when Zuckerberg declared a “year of efficiency” and pushed many managers and directors to either move into individual contributor positions or leave the company. The current ask is a partial walk-back of that policy, at least within one division.
The AAI division itself is new. It was formed in 2026, after Meta cut around 8,000 employees in May and cancelled plans to fill 6,000 open positions. About 7,000 workers landed inside AAI through reassignment rather than direct hire. Earlier this year, Wired reported that a portion of those reassignees were dissatisfied with the placement; some were later allowed to look for other roles inside Meta.
Neither Fortune nor Meta has named a target number of managers to bring back, and the reporting gives no indication the recall will spread beyond AAI.
Why it matters
The numbers tell a tension: revenue is growing fast (up 28% year-over-year), but expenses are growing faster (up 55%). Zuckerberg has said AI investments are shaping every major part of Meta’s business. If the company is now rebuilding a management layer inside its flagship AI unit, it suggests the flat structure created coordination problems that cost more than the management overhead it was meant to eliminate.
For anyone watching AI org design, this is a useful data point. Flat structures work well when individual contributors have clear, self-directed mandates. Reassigning 7,000 people into a brand-new division and expecting them to operate without team leads is a different challenge. The friction was apparently visible enough to prompt a reversal within the same calendar year the division was created.
It also fits a pattern. Meta has seen other notable AAI-adjacent changes this month, including the departure of superintelligence researcher Andrew Tulloch and the debut of its Muse agent. The division is clearly still being shaped in real time, which adds uncertainty for people inside it and for observers trying to gauge where Meta’s AI product priorities are heading. You can follow the broader story of how AI labs are structuring themselves in our AI news coverage.
Our take
The “year of efficiency” framing was always partly a financial story and partly a story about optics during a bad ad market. Meta is now in a very different financial position: $60.8 billion in quarterly revenue and a stated commitment to AI spending that is visibly pushing expenses up 55%. Rebuilding management in a high-priority AI unit is a rational move, but the framing as “opt-in” and the lack of a target headcount suggests Meta is being careful not to officially declare that the flattening experiment failed.
For business owners thinking about their own AI initiatives, the lesson is straightforward. Reorganising people into a new AI team without clear leadership structures creates drag, not speed. If you are considering integrating AI into your operations, the structural side (who owns what, who reviews outputs, who escalates problems) matters as much as the tooling.
What to do about it
- Watch Meta’s Q3 2026 expense line. If costs keep climbing at 55% year-over-year while this restructure continues, the “efficiency” era will look definitively closed.
- Track AAI product output. Manager count is only meaningful if it correlates with shipping speed and product quality.
- If you run an internal AI team, audit whether your contributors have clear ownership. A flat structure without clear mandates produces the same friction Meta is now correcting.
Structure your AI team before you scale it, not after.
Frequently asked questions
What is Meta's Applied AI division?
Meta's Applied AI (AAI) division was created in 2026. About 7,000 employees were reassigned into it after Meta cut roughly 8,000 people in May 2026 and cancelled plans to fill 6,000 open positions.
Why is Meta reversing its manager flattening policy?
Meta has not given an official reason, but the opt-in program asks individual contributors in its AAI division to return to manager roles. Earlier reports noted some AAI reassignees were unhappy with their placements, suggesting the flat structure created coordination problems.
How many employees did Meta lay off in 2026?
Meta cut roughly 8,000 employees in May 2026, about 10% of its workforce, and also scrapped plans to fill 6,000 open positions. The company ended Q2 2026 with 75,472 employees.
What were Meta's Q2 2026 financial results?
Meta reported Q2 2026 revenue of $60.8 billion, up 28% year-over-year, with expenses of $42 billion, up 55% year-over-year. Headcount fell 3% from the prior quarter to 75,472 employees.
