Mark Zuckerberg Says AI Agents Have Not Advanced as Quickly as Expected

Meta CEO Mark Zuckerberg told staff this week that AI agent development has not progressed as quickly as hoped. The admission comes months after the company cut about 8,000 jobs and moved thousands more employees into AI-focused teams.

The statement was made during an internal all-hands meeting, as first reported by TechCrunch. It highlights the gap between earlier internal expectations and current results. Meta had aimed to accelerate agent capabilities through a new AI-centered structure, yet those expected benefits remain unrealized. TechCrunch reported Zuckerberg saying the pace of agent progress “hasn’t happened as quickly as we would have hoped.”

Layoffs and Restructuring Did Not Deliver Fast Change

Meta reduced headcount by roughly 10 percent earlier this year. At the same time the company reassigned another 7,000 people to multiple AI groups, including the Agent Transformation team. Zuckerberg described the layoffs as not clean enough because senior leaders feared the firm could not adapt fast enough to industry shifts.

The restructure was meant to align resources with AI priorities. Instead, the company is still waiting for visible gains from those moves. The internal message framed the slower pace as a temporary setback rather than a permanent problem.

Heavy Spending Continues Despite Slow Progress

Meta plans to spend up to 145 billion dollars on AI infrastructure this year, according to Reuters. The investment covers data centers, chips, and training systems needed for agent development. Executives continue to back the spending even while acknowledging the current shortfall in agent performance. Reuters noted the figure reflects Meta’s continued commitment to scaling AI systems despite internal timelines slipping.

The high outlay reflects a bet that infrastructure scale will eventually translate into better agents. At the same time, the lack of immediate acceleration raises questions about how quickly returns will appear. The three-to-six-month window cited by Zuckerberg is now the key period observers will monitor.

Structure Changes Have Not Yet Produced Clear Wins

Zuckerberg noted that the benefits expected from placing AI at the center of company operations have not shown up on schedule. The new setup was designed to speed decision-making and reduce friction between product and research teams. Results so far suggest those adjustments need more time to take effect.

Employees in the newly formed AI groups face pressure to show measurable advances. The internal timeline set earlier this year placed greater weight on rapid iteration. The latest update indicates that timeline has slipped.

Agent Development Faces Technical and Organizational Limits

Building reliable agents requires progress on planning (the ability to break goals into sequenced steps), memory (retaining context across extended interactions), and tool use (correctly invoking external functions or APIs). Current systems still struggle with consistent multi-step execution over long horizons - for example, an agent asked to book a multi-leg business trip may successfully reserve the first flight yet fail to re-check hotel availability after an airline schedule change, or may attempt to use an expired corporate travel code on the return leg. The gap between prototype demonstrations and production-grade performance remains wide.

Meta's decision to consolidate talent was intended to close that gap. The slower pace suggests coordination challenges persist even after the reassignments. External competitors face similar hurdles, yet Meta's public scale makes the shortfall more visible.

What to Watch in the Next Three to Six Months

Three concrete signals will show whether agent velocity is improving. First, any new internal benchmarks shared in future all-hands meetings will indicate whether the three-to-six-month outlook is materializing. Second, updates on the Agent Transformation team's output, such as new agent frameworks or measurable task success rates, will reveal concrete progress. Third, Meta's quarterly infrastructure spend reports will show if capital deployment continues at the planned pace or adjusts downward.

If the first two signals remain weak while spending stays high, the timeline for agent gains will face renewed internal scrutiny. If the signals strengthen, the earlier layoffs and reassignments will be viewed as having delivered the intended shift. Readers tracking Meta's AI efforts should focus on these three checkpoints rather than broad industry commentary.