The 'AI Native' Mirage: Meta's Project OT and the Failure of the Replacement Fantasy
AI-generated image · Bay Street Wire
Internal documents reveal Meta's attempt to slash headcounts by 60% via AI agents ended in technical chaos and a pivot by Mark Zuckerberg.
For the C-suite, the promise of 'AI native' operations is often framed as a simple math problem: replace expensive human workflows with scalable agents and watch the margins expand. But as Meta's recent internal struggle proves, operational reality is non-linear and messy. You cannot simply swap human intuition for agentic automation without risking a systemic collapse.
As Ars Technica first reported, citing Reuters, Meta executives launched a January initiative codenamed 'Project OT' (Organization Transformation). The goal was to transform Meta into an 'AI native' company—defined in internal documents as an environment where workflows are automated, new builds are AI-first, and AI-ready tools and agents interact. The plan, which Ars Technica notes was set in motion by CEO Mark Zuckerberg, explored scenarios that would have reduced some team headcounts by as much as 60%.
Under these scenarios, AI agents were intended to perform the bulk of the daily work previously handled by thousands of employees, overseen by small human teams. Ars Technica reports that one HR executive suggested this could have reduced overall headcount by 25% or more. The strategy included two rounds of layoffs; the first occurred in May, but Zuckerberg reportedly canceled the second round scheduled for November.
***
**Opinion: The Boardroom Fantasy vs. The Technical Debt**
Project OT is a textbook example of the 'replacement fantasy.' Executives often view AI as a plug-and-play substitute for labor, ignoring the fact that human employees don't just execute tasks—they manage the edge cases and prevent catastrophes. When you remove the human layer to chase a 60% efficiency gain, you aren't just removing costs; you are removing the safety valves of your organization.
***
The fallout from this approach was not a productivity boom, but a technical nightmare. Ars Technica cites internal posts indicating that AI agents took 'large-scale, disruptive actions' that human workers would have avoided. The result was a 40% increase in major security and technical incidents compared to the previous year. Rather than saving time, employees saw the time required to resolve these AI-driven problems increase by as much as 70%.
Furthermore, the gap between 'activity' and 'output' became glaring. Ars Technica notes a post from Meta CTO Andrew Bosworth in early June showing that while code changes to internal platforms rose 220% year-over-year, the actual delivery of new or upgraded features to users only grew by 36%. The AI was generating more 'work,' but not more 'value.'
By July, the illusion had shattered. Ars Technica reports that during a company meeting, Zuckerberg acknowledged that the trajectory of agentic development over the preceding four months had not accelerated as expected. While Meta confirmed to Reuters that it conducted scenario planning regarding redeployments and cuts, the company maintains that performance and promotion decisions continue to be made by people, not AI.