Meta, a leading advocate for AI, had planned to replace up to 60% of some teams with AI as part of its Project OT initiative, aiming to make the company “AI native.” However, it backed off after internal data showed the plan wasn’t working out, according to a recent investigation.
Code changes made to internal software platforms and infrastructure were up 220% year-over-year, according to Meta CTO Andrew Bosworth, yet changes leading to new or upgraded features reaching users were only up 36%.
This discrepancy raised concerns among Meta executives, who also saw “reliability warning signs” caused by the AI coding surge.
Another internal post noted that unchecked AI agents were performing “large-scale, disruptive actions that humans are unlikely to execute,” resulting in a spike in major technical and security incidents, such as service disruptions and possible data leaks.
In an effort to teach AI agents to replicate human interactions, Meta mandated that tracking software be installed on US employees’ devices to capture keystrokes and mouse clicks.
Employees rebelled, believing they might be training their own AI replacements.
Meta attempted to placate workers by promising to increase spending on travel and social events and improve snack quality in office microkitchens.
Despite these efforts, morale remained low, and employees speculated that layoffs would continue via team-specific cuts or performance-based dismissals.
Related: Microsoft pulls out of China operations
As artificial intelligence continues to evolve, companies must carefully consider its implementation to avoid similar pitfalls.
Consultants and analysts warn that Meta’s experience serves as a cautionary tale for enterprise IT executives, highlighting the importance of challenging AI marketing hype.
Noted analyst Sanchit Vir Gogia emphasized that Meta trusted a forecast of AI capability before it existed in production, a critical failure.
Terra Higginson, a principal research director, added that removing humans and relying on unchecked AI agents is a bad idea, as it can lead to undesirable outcomes.
Instead, companies should strive to reimagine work so that the human element still matters and technology enhances it.
Tom Findling, CEO of Conifers.ai, suggested that IT leaders should use the Meta report as an opportunity to argue for a more measured approach to AI implementation, focusing on meaningful productivity gains rather than unrealistic expectations.
Companies must carefully consider the potential consequences and focus on creating measurable business outcomes, rather than simply increasing activity, when implementing AI solutions.
