Meta's AI Agent Plans Included 60% Team Cuts

Meta's AI Agent Plans Included 60% Team Cuts

Meta's AI agent plans nearly cost thousands of employees their jobs. It was that close. Earlier this year, the company created a plan to reduce some teams by as much as 60 percent, all in service of becoming "AI native," and they didn't stop there with the internal push. The initiative, codenamed Project OT, explored scenarios where AI would handle much of the daily work currently performed by human staff, but it's the scale of those cuts that still stuns people today. So the threat was real, and it's not over.

The plan called for two rounds of layoffs. The first hit in May. The second was canceled. What happened in between tells a complicated story about the limits of artificial intelligence in the workplace.

What Project OT Actually Proposed

January brought change. Meta executives created Project OT, and CEO Mark Zuckerberg directed the management structure overhaul himself, according to internal planning documents that painted a picture of a future company. But those same documents described a workplace where "AI-ready tools and agents interact, workflows are automated, [and] new builds are AI-first." It's a bold bet. They're staking the whole organization on it. So the memo's language is clear: automation isn't optional, it's the new baseline.

Under these scenarios, small teams of people would oversee the AI systems. Some employees would get new roles. Others would be laid off. One HR executive reportedly said the company could have reduced headcount by 25 percent or more.

Meta confirmed the plan existed. But it told Reuters the project was canceled before determining how many layoffs it would've triggered. The company said it asked teams to conduct scenario planning exercises, which ultimately led to moving thousands of employees to new priority work on several newly-established teams, a process that spanned multiple departments and reshaped internal roles across the board.

The Financial Incentive

Money mattered. That's part of the story. Meta planned to use the savings from layoffs to funnel cash toward its highest performers, especially those with specialized AI engineering skills, because the company needed to stay competitive in a brutal talent war. But the logic was simple: cut costs in some areas, invest heavily in others. It's a classic trade-off, and they weren't shy about it.

June changed everything. Meta began selling AI agents to third parties, a key component of the "AI native" vision that would reshape how the company builds and ships products. A pilot program restructured engineering teams, research teams, and at least eight other teams into smaller groups. That's a lot of moving parts. An internal post called "AI-Native Playbook" outlined how the pilot would remove middle management and use "agent-assisted analysis" to prioritize daily tasks, but it didn't stop there, because the same playbook promised to shift decision-making down to the smallest possible unit, where agents could flag bottlenecks before humans ever saw them. So the whole thing hinges on speed. Don't expect a slow rollout.

Why Zuckerberg Pulled the Plug

The May layoffs happened. Then something changed. Zuckerberg canceled the November wave almost immediately after the first round, though Reuters said it could not determine exactly what prompted the shift.

Several factors likely played a role. Employee morale took a hit from March and April reports of impending layoffs and from Meta tracking keyboard and mouse input to train AI agents, a program the company has since paused.

But the bigger issue may have been performance. The numbers tell a sobering story.

Code changes to internal software platforms were up 220 percent year-over-year, according to a post by CTO Andrew Bosworth in early June. But changes that led to new or upgraded features reaching users were only up 36 percent.

That gap signals plenty of motion, yet surprisingly little in the way of concrete results. AI agents were making "large-scale, disruptive actions that humans are unlikely to execute," according to internal posts, and that's a sobering thought. Major technical and security incidents increased 40 percent compared to the prior year. But the real drain? Employee time spent resolving those problems jumped by as much as 70 percent, so the costs aren't just operational, they're human, and they're compounding fast. It's a mismatch that can't last.

The Productivity Question

In July, Zuckerberg acknowledged during a company meeting that the "trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected." That admission undercuts the entire premise of replacing workers with AI.

People, not AI, made those calls. Meta told Reuters that performance rating and promotion decisions "were and are made by people, not AI," so the company is holding its ground on that point. But it wouldn't touch the internal posts about disruptive AI actions and increased incidents. No comment. That silence speaks volumes, and it's a stark contrast to the firm denial they offered on the core question.

What Meta's Experience Reveals

Meta's story is a cautionary tale. It warns organizations rushing to implement AI, and that warning cuts deep. Many businesses, both larger and smaller than Meta, are exploring ways AI can save time, drive revenue, or reduce costs, including in human resources, yet they often overlook the messy reality of replacing people. So even the most eager companies may struggle to replace human workloads with AI. But Meta's experience shows it's not simple. It just isn't.

The risks of overzealous AI projects are real. They're expensive, too. Disruptive actions, technical incidents, and time spent fixing problems can offset any gains from automation, and that's before you count the lost focus and the morale drain that quietly eats away at a team's momentum. But consider this: the 220 percent increase in code changes that didn't reach users suggests busywork, not progress. It's a red flag. So we've got to ask what we're actually building.

The Human Cost

Meta's AI agent plans would have reshaped the company dramatically. A 60 percent reduction in some teams, with an overall headcount cut of 25 percent or more under certain scenarios, represents a massive reorganization. Thousands of employees would have faced redeployment, role closures, or layoffs.

Zuckerberg's direct involvement in crafting the plan was telling. But it's the sheer fact that such a blueprint existed at all that reveals how seriously Meta weighed automation against human labor, even as the company ultimately shifted thousands of employees to priority work on new teams instead. That's the real story.

Meta's case highlights the gap between AI's promise and its current reality. The technology can handle certain tasks. It can automate workflows. But it also creates new problems that require human intervention, potentially erasing the efficiency gains it was supposed to deliver.

Organizations watching Meta's experiment should pay attention. They had the resources, the talent, and the leadership commitment to make AI agents work, yet the rollout still strained internal workflows, rattled employee trust, and forced abrupt course corrections that no one had anticipated. But here's the hard truth. If Meta couldn't manage it without major disruption, smaller companies with fewer resources could face even tougher obstacles, and they don't have the same cushion to absorb the fallout. So don't assume you'll fare better.

Project OT's lesson isn't that AI is useless. It's that replacing people with AI is harder than it looks, and the costs of getting it wrong extend far beyond the balance sheet, bleeding into morale, trust, and the unglamorous reality of daily operations where small failures compound into expensive, human-sized problems. Meta's AI agent plans are dead. But the questions they raised about automation, productivity, and human work aren't going anywhere.