For thousands of Meta employees, it’s now the AI road, or highway.
On Wednesday, as 8,000 Meta employees read their layoff letters, 7,000 others opened an entirely different email. They were selected to participate in a new AI initiative directly launched by CEO Mark Zuckerberg, which was critical to accelerating Meta’s position in the AI race.
Many employees learned they would be moving to a group Meta founded earlier this year called Applied AI (AAI). The group is led by Maher Saba, vice president of engineering, and reports to Andrew Bosworth, chief technology officer.
Others were recruited into groups that focused more specifically on AI agents. These include a group led by Bosworth called “Agent Transformation Accelerator” and a team called “Agent Data and Optimization,” according to internal messages shared with Business Insider.
“This is a reflection of your influence,” one email read, telling recipients they were chosen for their “outstanding performance” and technical ability, according to a copy reviewed by Business Insider.
“You have been identified as someone who can make a big impact on this team,” he added.
With Meta’s 78,000 employees being cut this week, these reassignments to the AI team were a lifeline for some staffers who were worried about being laid off, even if they weren’t yet sure what their new roles would be.
Many companies are restructuring around AI, with some tracking usage like Disney and others like Visa rewarding teams that develop products faster. The meta has undergone particularly drastic changes. It created a superintelligence lab, organized its teams into small AI-native “pods,” and changed the job titles of some employees to “AI builders.”
“I got drafted,” one person said on a Discord server where Meta employees discussed layoffs.
“Welcome to the draft,” another posted.
Meta’s AI models have so far lagged behind those of competitors like OpenAI and Google in performance, giving leaders an opportunity to better train Meta’s employees on the technology.
A Meta spokesperson declined to comment.
Inside Meta’s AI Draft
Participation in the new initiative turned out to be non-negotiable, and various new hires received the news with either relief, fear or confusion, according to Business Insider and four employees who responded to posts on several forums.
On the anonymous workplace app Blind, a section reserved for verified meta employees was flooded with posts from people announcing they had been drafted. Some asked what AAI actually does.
Two Meta employees said they expect the new AI task force to include data labeling, the tedious work of manually tagging images and modifying chatbot responses.
This is an area in which Meta’s head of AI, Alexandr Wang, has extensive experience. Mr. Wang is the co-founder and former CEO of Scale AI, one of the world’s largest data label companies. Scale AI relies on a number of contractors to train its AI.
Meta is rolling out other AI initiatives internally. Business Insider previously reported that the company has launched an internal tool, the Model Capability Initiative, to track employee keystrokes and mouse movements.
An internal announcement about the software in April said AI models still lack the knowledge to outperform humans at technical tasks such as coding.
“Models need to be trained on real-world examples to help agents understand how people actually use computers to perform everyday tasks,” the post said.
In a leaked audio recording of an internal meeting that same month, Zuckerberg said that Wang knew the world of data labeling well, and then offered a candid follow-up: The average employee at Meta has “substantially more” intelligence than those contractors, so he said he wanted to “collaborate” with top employees across Meta to train them in AI instead.
“I think this would be a huge benefit if we could make it happen,” he said, adding that it remains a “hypothesis” for now.
Have something to share? Email Hugh Langley. hlangley@businessinsider.com Or text us at 628-228-1836. Email Charles Rollet. crolet@businessinsider.com or on Signal and WhatsApp 628-282-2811. Use a personal email address and non-work device. Here’s a guide to sharing your information securely.