Andrew Bosworth has long served as one of Mark Zuckerberg’s most trusted lieutenants at Meta. The chief technology officer, known inside the company as Boz, carries a reputation for plain speech that sometimes lands with the force of a sledgehammer. This week that bluntness turned inward. In an internal post reviewed by WIRED, Bosworth conceded that Meta handled its recent artificial-intelligence reorganization poorly. He called the rollout “atrocious.”
The admission landed amid months of rising frustration. Meta created an Applied AI unit of roughly 6,500 engineers and product managers in March. The goal was to harness the company’s scale and talent to accelerate generative models. Instead many workers found themselves assigned to tasks they viewed as menial. One engineer described the environment as “a gulag.”
Bosworth did not sugarcoat the damage. “We’ve undermined the trust you have that your specific expertise and contribution will be valued, that you will grow and advance your career, and that this will be a place where you can actually have an impact,” he wrote. The rapid shifts in strategy, combined with earlier hiring surges followed by cuts, left teams unstable. Managers changed. Priorities flipped. Confidence eroded.
Yet the problems run deeper than one reorganization. The Wall Street Journal reported last month that Meta planned to lay off about 8,000 employees while pouring tens of billions of dollars into AI infrastructure. At the same time the company began installing software on U.S. employee computers to record keystrokes, mouse movements and screen snapshots. The data would train AI agents meant to handle more of the actual work.
Employees reacted with alarm. Some launched a petition demanding an end to the tracking. Others posted angry comments and reaction emojis. When asked about an opt-out, Bosworth replied simply: no. To those worried about privacy he offered practical advice. Don’t open personal email on company devices. The stance reflected his broader vision. As he put it in earlier memos, the future involves AI agents that “primarily do the work” while humans direct, review and improve them.
The surveillance push and layoffs together intensified a broader slide in morale. Worker petitions, internal protests and reports of surveillance have compounded the sense that speed now trumps consideration. A departing staffer even posted a biting anti-AI video that name-checked Bosworth and the monitoring program known as the Model Capability Initiative, according to Mother Jones.
Bosworth’s latest message attempts to address the fallout. He promised greater stability. Better communication. A return to some familiar perks. Microkitchens would improve. Travel budgets for team events would rise. Social gatherings would come back. Managers would receive clearer expectations, capped around 20 direct reports, and fewer abrupt switches. The company would roll out AI coaching tools and give employees more personalized attention to career growth.
“I hope we can rekindle the best of the culture we joined,” Bosworth wrote. He also reminded staff of a competitive reality. “We should heed the saying, ‘AI won’t take your job but someone who knows AI might.’” Performance reviews, he suggested, would increasingly reflect impact on AI initiatives. Compute resources remain constrained. Trade-offs are constant. He encouraged teams to escalate bottlenecks rather than accept them quietly.
The Applied AI team’s leader, Maher Saba, echoed some of the corrective tone in his own earlier post. The initial decision to draft engineers into the unit drew on Meta’s unique advantages in scale and expertise. Now the approach would shift. Employees could once again apply for roles that interested them. The team would keep “moving fast and fixing forward,” Saba wrote, without the heavy reliance on traditional roadmaps.
But words alone may not restore trust quickly. Meta’s transition to an AI-first organization has been messy. Billions spent on data centers and models have not shielded the workforce from uncertainty. Earlier this year Bosworth told employees in an internal forum that those who could not accept policy changes should consider leaving. The message, first reported by Business Insider, underscored a culture that values commitment over consensus.
That stance carries risk. Talent remains mobile. Engineers with AI skills have options at rival labs that may offer clearer direction or less intrusive monitoring. Meta’s scale still provides leverage, yet repeated reorganizations and surveillance have created friction that no amount of snacks or microkitchens can instantly erase.
Bosworth has survived past controversies. A 2016 memo that justified aggressive growth tactics even at the cost of potential harm leaked years later and sparked internal outrage. He stayed. He rose. His current role overseeing both technology direction and the push to embed AI across the workforce gives him unusual reach. The New York Times described the company’s AI embrace as awkward at times and downright ugly at others. The internal temperature reflects that assessment.
Executives including Zuckerberg have also posted messages acknowledging employee sentiment and pledging adjustments. Whether these efforts produce lasting change remains uncertain. For now Bosworth’s candor stands out. He named the failure. He outlined fixes. He stopped short of apology but signaled recognition that the path taken so far has cost goodwill.
The coming months will test whether Meta can convert that recognition into renewed energy. AI projects continue at pace. Layoffs have trimmed headcount. The remaining staff must deliver results while adapting to new tools that, by design, aim to automate parts of their own roles. The tension is obvious. So is the ambition. Meta bets that its people, once properly supported, will outpace competitors who lack the same combination of data, distribution and determination.
Bosworth’s memo ends on a note of optimism. He wants to rebuild the culture that once drew top engineers. He wants clearer vision and steadier careers. The question is whether employees, after months of disruption, still believe the company can deliver on those promises. The answer will shape Meta’s AI trajectory as much as any model or data center.
WebProNews is an iEntry Publication