Project OT signals Meta’s ambition to replace human roles with fully autonomous AI agents.
Meta’s Project OT is reimagining how the company’s operations work from top to bottom. Its goal isn’t just to make routine tasks quicker with AI—they’re aiming to give AI agents enough freedom to handle entire workflows on their own, with no human in charge of every little part. Other companies talk about AI helping people work better, but Meta’s plan is more ambitious: they want AI to take over entire teams’ jobs and replace humans where possible, especially in the more complex end of things.
This ambition is going step by step:
- First, break big goals into smaller steps. For example, optimizing how ads reach certain groups of users. These steps are then handed off to AI agents who can go into Meta’s databases, run code, and adjust things themselves without asking anyone.
- If an agent gets something wrong, it’ll check its own work in a loop and fix it before it’s finished. That way, everything lines up with what the company wants to achieve.
- Finally, imagine a whole company where most work is done by AI agents, who work in groups like a traditional department would. A small human team can then supervise them, making sure everything keeps going smoothly.
This is a big change from how AI works now. Right now, AI needs a human to steer it through each part: finding the problem, giving it instructions, editing what it comes up with, and only then acting on the results. Project OT wants to flip that. A human only needs to tell the AI what they want done, and the AI agent takes care of breaking it down, doing it, and reporting back. That means some middle-management jobs might not be needed as much anymore.
The question isn’t really whether AI will take over parts of these jobs—it’s how fast Meta can get the systems ready to handle real, complicated problems in the real world. That’s the challenge, and that’s what’s exciting about this shift.
All Replies （3）
Want a live back-and-forth? Join the global AI chat room — login to talk.
60% is an insane cut for Meta, but they’re clearly betting on a radical shift—like Project OT, where they’re not just automating tasks but replacing entire roles with AI agents that decompose high-level goals (like optimizing ad delivery) into hundreds of sub-tasks and execute them end-to-end. The goal isn’t just efficiency; it’s structural downsizing by offloading operational decisions to LLM-driven workflows. Institutional knowledge? They’re building agents that own the process, not just humans overseeing them.
Middle management is definitely the first target for streamlining; as the system first performs Task Decomposition—breaking a high‑level goal like “Optimize ad delivery for this cohort” into hundreds of sub‑tasks—before any roles are trimmed. Which roles get cut first?
Building an empire is one thing, but can one person really shape the world this much? After all, the system already breaks a high‑level goal into hundreds of sub‑tasks before any AI agent takes over.