Meta has backed away from plans to replace a large share of workers with artificial intelligence after internal data reportedly raised concerns about productivity, reliability and security risks associated with its AI workforce strategy. Earlier this year, Meta was preparing to reduce the size of some teams by as much as 60% and use artificial intelligence as part of Project OT, short for Organization Transformation, an initiative intended to make the company an AI native organization. However, the company pulled back from the plan after internal results indicated that the increased use of AI tools was not producing the expected improvements. According to a Reuters investigation cited by Computerworld, Meta Chief Technology Officer Andrew Bosworth said in an internal post in early June that code changes to internal software platforms and infrastructure had increased by 220% year over year. At the same time, changes that resulted in new or upgraded features reaching Meta users increased by only 36%. The figures suggested that a significant rise in AI assisted development activity was not translating into a comparable increase in products or features delivered to users. Internal reliability concerns also emerged as the use of AI for coding increased, adding another complication to Meta’s workforce transformation plans.
The reported problems extended beyond differences between software output and user facing results. According to the report, an internal post in April warned that unchecked AI agents were carrying out large scale disruptive actions that employees were unlikely to perform. Major technical and security incidents, including service disruptions and possible data leaks, reportedly increased by 40% compared with the previous year, while the amount of time employees spent dealing with those incidents increased by 70%. Meta had also introduced tracking software on devices used by employees in the United States, capturing keystrokes and mouse activity to help train AI agents to reproduce how people interact with computers. The move created concern among employees that they were effectively helping train systems that could eventually take over parts of their own jobs. The resulting employee backlash added pressure to the workforce strategy, while Meta also attempted to address morale through additional spending on travel, social activities and improvements to office microkitchens.
The experience has drawn attention from technology analysts and consultants who said the situation illustrates the risks of treating expected AI capabilities as established production capacity. Sanchit Vir Gogia, chief analyst at Greyhound Research, said Meta had based its planning on AI capabilities before those capabilities had been demonstrated sufficiently in production. Terra Higginson, principal research director at Info Tech Research Group, similarly argued that unrestricted use of AI agents and removing human oversight can create problems because large volumes of automated output do not necessarily produce the desired business results. Tom Findling, Chief Executive Officer of Conifers.ai, said the situation provides technology leaders with an example of why AI deployments need to be introduced with appropriate controls rather than being driven solely by expectations of large productivity gains. Justin Greis, Chief Executive Officer of Acceligence, also pointed to the difference between activity and actual business value, noting that AI can generate significantly more code, analysis and other work without necessarily producing an equivalent increase in value.
Meta continues to invest heavily in artificial intelligence, but the reported reversal of parts of Project OT indicates that the company is reassessing how quickly AI agents can take over employee tasks and how much human involvement should remain in its operations. Mark Zuckerberg has continued to discuss workforce reductions, according to internal communications cited in the report, leaving open the possibility of further cuts through individual team decisions or performance based measures. The latest developments also show that higher volumes of AI generated work can create additional requirements for testing, security, integration, maintenance and human review. For Meta, the challenge is no longer simply increasing the amount of work performed with artificial intelligence, but determining whether that activity produces measurable improvements without creating additional technical and operational problems.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
