Internal Meta data analyzed by Reuters shows that restructuring the workforce into smaller AI-supported teams led to a 220% increase in the number of software changes. The data came from a broader company effort to reorganize engineering operations under the premise that lean teams with AI tools would be more productive.
The same survey, however, found that the growth in changes resulting in user-facing features was slower than the overall change rate. Meanwhile, internal records show increased reliability and security problems as employees accelerated their delivery pace with the help of generative AI.
Meta tested the reduced-team model across multiple software divisions, delegating code review and testing tasks to AI tools. The expectation was that automating development cycle stages would compensate for the staff reduction, but the data suggests the volume of changes grew faster than human or automated validation capacity.
For operations and engineering teams that adopt or are considering similar structures, the lesson is clear: productivity metrics based solely on change volume may hide quality erosion. Those running systems need to include stability and security indicators in the same report that measures delivery speed, or they risk replicating Meta's pattern.
This publication recognizes the value of testing new organizational models, but notes that technology still cannot replace human oversight at critical points. Companies that accelerate delivery without corresponding validation tend to suffer more incidents. Meta serves as a warning, not a benchmark.