Meta has begun mandatory transfers of its top software engineers into the Applied AI Engineering unit, as stated in an internal memo from Maher Saba.
What began as a voluntary process has now become compulsory, reflecting the unit’s emergence as one of the company’s top priorities.
This transition is part of a larger reorganisation in anticipation of layoffs, with Meta refraining from commenting on the details of the plan.
The Applied AI unit is responsible for developing tools and assessments for the creation of autonomous AI agents that can code, test, and deliver products with minimal human intervention.
CEO Mark Zuckerberg has indicated that 2026 will be a pivotal year for AI, fundamentally transforming work at Meta, with streamlined teams and AI-integrated tools enhancing productivity.
This marks a strategic shift from the metaverse to AI as a key driver of efficiency and cost management.
A vice president in the Reality Labs division and a long-time associate of Chief Technology Officer Andrew Bosworth, Saba had previously encouraged volunteers to join the organisation when he announced its formation last month. In his latest memo, he stated that participation was no longer optional.
Meta planning next step amid competition
The company is “now moving to the next phase: scaling the team” and has collaborated with leaders throughout Meta to “identify strong SWE (software engineer) talent” for this initiative.
Should Meta’s AI agents fulfill their potential, the company could greatly reduce its workforce while speeding up product development cycles, potentially surpassing competitors in cost efficiency and innovation pace.
On the other hand, if the performance of the agents falls short, similar to the delayed Avocado model, Meta risks increasing sunk costs and facing operational disruptions amid fierce competition from OpenAI, Google, and Apple.
The enforced talent transition indicates a significant gamble on AI as the cornerstone of Meta’s future competitiveness.
Read more: Meta widens global crackdown on teen content
