AI
Meta’s ‘AI for Work’ Transformation Lead Departs After Two Months
Meta VP Emily Dalton Smith, tapped to lead ‘AI for Work’ transformation two months ago, is leaving, deepening unrest inside Meta’s aggressive AI overhaul.
The Meta product executive leading the company’s ‘AI for Work’ transformation is leaving, two months after Chief Technology Officer Andrew Bosworth put her in charge of the unit responsible for the internal AI assistant Metamate. Emily Dalton Smith, with the Facebook parent since 2015, told employees she will stay briefly to help transition the team with Bosworth, according to a report from Reuters, which cited an internal announcement seen on Wednesday.
A Meta spokesperson declined to comment on whether Dalton Smith was leaving, but said Meta was continuing to incorporate AI tooling into Metamate. The exit lands in the middle of a sweeping restructuring that has cut roughly 10% of the workforce, pushed thousands more into new AI-focused roles, and installed mouse-tracking software on staff machines, per the Reuters report that first detailed her two-month tenure.
Two Months In, the Lead Departs
Dalton Smith’s exit lands at the top of the ‘AI for Work’ restructuring, an effort that has played out largely by memo, code, and headcount. She was named to lead a pod within the Agent Transformation Accelerator (ATA) plan, a company-wide push to build AI agents that can autonomously do work now done by Meta’s own staff. The role was announced in April.
That memo, also seen by Reuters, said Dalton Smith’s pod would be focused on ‘the interfaces, platform components, memory systems, automations and shared product experiences that make AI useful for everyone.’ Her remit also included Metamate, the enterprise AI assistant Meta wants as the single front door for everything from research to sales decks. She joined Meta in 2015 and had previously been a vice president of product management and head of product for Threads, the company’s microblogging app. The Reuters report is the most detailed public account of her two-month tenure.

What Metamate Was Built to Do
Metamate is the public face of Meta’s enterprise AI push, named for the company’s internal nickname for its employees. In a memo last month seen by Reuters, Dalton Smith told employees the goal was to consolidate the company’s fragmented collection of internal AI tools into Metamate. Her team was planning to pull in features such as chat-based coding coordination, file navigation, and ‘persistent memory’ of an employee’s work. She framed the consolidation in plain terms: make Metamate the single starting point for everything from deep research to prototyping a new feature to a sales presentation. New features were set for June 1.
Our goal is to make Metamate the starting point for all kinds of work.
The plan was to bring in polished dashboards and microsites from Manus, a Singapore-based AI agent startup Meta acquired in December. ATA is the spine of the broader restructuring, and Metamate is its most public-facing piece. Dalton Smith told employees, per the Reuters report, that the work would include features from Manus, even as Meta was working through a response to the Chinese government’s order. The Manus integration made the project especially sensitive, the Reuters report noted.
A Meta spokesperson declined to comment on whether Dalton Smith was leaving the company, but said Meta was continuing to incorporate AI tooling into Metamate. The Reuters report did not name a successor for the pod, or give a timeline for the Metamate rollout. New capabilities were expected to land inside Metamate as of June 1, per the same memo.
The Manus Thread Running Through the Project
The Manus integration made Dalton Smith’s project unusually sensitive. Meta paid about $2 billion for the Singapore-based startup in December. The Chinese government ordered the deal unwound in April, citing concerns over foreign ownership of an AI firm with Chinese roots.
Per the Reuters report, Meta subsequently cut off Manus’ access to its internal systems while it worked through a response. Dalton Smith’s team was still planning to bring Manus-style features into Metamate, even as the regulatory picture stayed unresolved. The complication adds an operational layer to a project already under internal strain.
Manus was founded in China before relocating to Singapore. The company launched its first general AI agent in March 2025, a release that saw the startup lauded as the next DeepSeek. Manus said it had passed $100 million in annual recurring revenue within eight months of launch, a milestone Meta cited when announcing the deal.
The Layoff and Reassignment Math
The Dalton Smith exit is the visible top of a much larger workforce shakeup that has played out across the spring. In April, Meta told employees it would conduct a major round of layoffs the following month. The cuts were framed in a CEO memo as a step the company had to take to compete in the AI era.
- 8,000 Meta employees laid off on May 20, 2026, about 10% of the workforce
- 7,000 Meta employees reassigned into new AI-focused roles
- 6,000 open positions canceled at the time of the April announcement
- 1,000 Reality Labs employees cut in January 2026
- 25% drop in Meta’s overall employee rating on Blind since Q2 2024, with a 39% drop in culture rating
In his May 20 layoff memo, Zuckerberg conceded that Meta had been less clear in its communication than it aspired to be. He also said executives ‘do not expect other companywide layoffs this year,’ and that ‘AI is the most consequential technology of our lifetimes. The companies that lead the way will define the next generation.’ The same memo carried the line ‘success isn’t a given’ in the AI era, the layoff-day memo seen by reporters. The job cuts hit numerous departments, though teams focused on AI infrastructure, foundation models, and AI monetization were largely protected.
The 7,000 reassigned employees were moved into new AI-focused roles tied to the AI restructuring. Some workers inside the new AI units, in WIRED’s reporting, have described the work as menial, with one employee calling it ‘a gulag.’ The reassignments ran alongside the layoffs, even as the layoffs dominated the public narrative.
Across internal channels, employees have openly criticized executives in company meetings and on message boards, the same kinds of fora where complaints about the tracking software have stacked up. The unrest inside the AI team is part of a broader downward swing in morale at Meta in the wake of mass layoffs, worker surveillance, and other concerns, a record covered in Zuckerberg’s own admission.
The Mouse on Every Desktop
A separate program has caused its own share of internal anger: a software rollout that tracks employees’ mouse movements, clicks, and keystrokes on US staff machines. Called the Model Capability Initiative, the program is overseen by the Meta Superintelligence Labs team and captures periodic screenshots for context. The employee mouse-tracking software launched in April as part of the broader push.
In an internal memo, the program was framed this way: ‘This is where all Meta employees can help our models get better simply by doing their daily work.’ A Meta spokesperson, Andy Stone, told Reuters the data would help with tasks Meta’s AI agents struggle with, ‘things like mouse movements, clicking buttons, and navigating dropdown menus.’ Stone said the data would not be used to evaluate employees. European employees are not included, since the program would run afoul of national privacy laws.
Reuters reported that many employees see the software as tantamount to helping design their own bot replacements. Staff complaints about layoffs, transfers, and the tracking software have stacked up across the spring. The unrest has shown up in company meetings and on internal message boards.
An Admit-the-Mistakes Phase
Inside Meta, the leadership has begun to acknowledge the toll. Chief Technology Officer Andrew Bosworth conceded in a memo reported by WIRED that the company had done an ‘atrocious’ job rolling out its new AI division, the CTO’s recent memo on the AI reorg.
We obviously did an atrocious job explaining the vision, giving people a clear picture of how we would support them and their careers in the shift, and painting a picture of how it would change over time.
The Applied AI unit he was describing is a separate 6,500-person division of engineers and product managers formed in March. Workers inside it have described their work, in WIRED’s reporting, as menial, with one employee calling it ‘a gulag.’ Bosworth, in a follow-up comment, blamed himself and other executives for losing sight of the employee perspective while rushing to focus on broader strategy issues, in the memo reviewed by reporters. He announced a series of measures: managers capped at about 20 direct reports each, more stable team assignments, and improved ‘microkitchens’ in offices.
Bosworth’s memo also introduced a softer corporate motto, recasting ‘move fast and break things’ as ‘moving fast and fixing forward.’ A separate Friday memo from Maher Saba, vice president leading the Applied AI team, told drafted employees they could now apply for other roles. Saba said the new team is ‘initially focused on projects to increase the coding and agentic capabilities of Meta’s frontier AI models.’
In his memo, Bosworth said Meta doesn’t subscribe to the belief that AI will fully replace workers, but added that the company should heed the saying ‘AI won’t take your job but someone who knows AI might.’ The same memo pledged better communication, career growth, and even snacks. ‘I hope we can rekindle the best of the culture we joined,’ Bosworth wrote.
The AI Bet, Two Months Later
The Dalton Smith exit, the layoffs, and the unrest have piled up against the same backdrop: Meta’s bet that AI agents will do the work of a large slice of its own staff. According to data from the Blind anonymous professional network service, Meta’s overall rating by staffers sank 25% from a peak in the second quarter of 2024 to the current period, with a 39% drop in its culture rating. Bosworth’s memo, acknowledging that the rollout had been ‘atrocious,’ is now on the same shelf as Zuckerberg’s ‘we made mistakes’ note.
For now, the ATA plan continues. Dalton Smith is staying on briefly to help Bosworth transition the team to ‘what’s next.’ Meta has not yet named a successor for Dalton Smith’s pod, or given a public timeline for the Metamate rollout. The two-month tenure ended without a public successor, a public timeline, or a clear public read on the workforce’s view of the program.
-
CRYPTO1 month agoAndreessen Horowitz Bets $2.2B on Crypto’s Quiet Cycle
-
AI2 weeks agoVinRobotics’ VR-H3 Debuts at Vienna, VinFast Is Next
-
CRYPTO1 month agoCathie Wood Calls SpaceX IPO Demand ‘Voracious’ Ahead Of $1.75T Debut
-
NEWS1 month agoApple Strikes Preliminary Deal For Intel To Make iPhone And Mac Chips
-
APPS1 week agoDGO App Brings Rs 549 Mobile Pass for FIFA World Cup 2026 in Nepal
-
AI3 weeks agoAnthropic Hits $965 Billion Valuation, Edges Past OpenAI
-
NEWS2 weeks agoGoogle Search Profiles Build a Follow Graph Inside Discover
-
AI2 weeks agoTrump’s AI Memo Strips Vendors of Veto Power Over Military
