Meta's CEO told staff AI agent progress has stalled. An HR analyst says the writing was on the wall
Image from: TechCrunch, CC BY 2.0 <https://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons
Meta cut 10% of its workforce, reassigned thousands more, and committed $145 billion in AI spending this year. Last week, Mark Zuckerberg told employees the agents at the center of that bet aren't keeping pace.
At an internal town hall on July 2, the Meta CEO admitted that the "trajectory of the agentic development over at least the last four months hasn't really accelerated in the way that we expected," according to a recording obtained by Reuters. He also conceded that the company's recent job cuts weren't as "clean" as intended, and that the anticipated upside of Meta's new AI-focused structure hadn't "come to fruition yet."
The context behind that admission matters. Meta began laying off roughly 8,000 employees in May, about 10% of its corporate workforce, while reassigning another 7,000 to AI-focused teams, moves designed in part to help fund as much as $145 billion in AI infrastructure spending this year. Zuckerberg has acknowledged mistakes in the overhaul before, but this was his most direct concession yet that the technology at the center of it all is running behind.
Stacey Harris, chief research officer and managing partner at Sapient Insights Group, wasn't surprised.
"I think he forgot about the importance of context and the importance of the human interaction with those agents, and the auditing that needed to be required, like any other resource inside your organization," she said.
Speeding up bad processes
Harris's assessment is grounded in data. Sapient's Annual HR Systems Survey, the industry's longest-running independent study of HR technology, found that 36% of organizations were using AI in some way in their HR or workforce processes last year. Of those, about 15% had adopted too early, according to Harris. They moved before their processes and tools were standardized, and their business outcomes had decreased considerably as a result.
"Just so you know, we don't ask them their outcomes. We get those outcomes through third party analysis," Harris said, referring to measures such as profitability, market share and workforce readiness.
What separated the organizations losing money from those seeing gains was remarkably consistent.
"The biggest difference between them is that they did not have efficient or effective core HR processes in place already when they had put in place the AI. So basically they just sped up really bad processes."
Most organizations getting value from AI, about 60%, remain in what Sapient calls the automation phase, where the payoff is process efficiency rather than headcount reduction. Another 17% had reached better decision making, which required foundations like skills and talent management. Only about 7% were doing true work transformation, redesigning work environments and business models around the technology.
Agents remain the exception
Zuckerberg's frustration centers on agents specifically, and here Harris's numbers are stark. Only 4% of organizations had implemented full HR agents in their service delivery models last year. The expectation was that adoption would grow two to three times this year. Preliminary data from Sapient's current survey suggests it will only roughly double, to about 8 or 9%.
"As much as we were all talking about agents, really good organizations aren't jumping headfirst into that, because agents have autonomy, which is very different from a workflow or an assistant that's giving you advice or something that's basically pulling up information for you," Harris said.
Her caution tracks with independent research showing AI agents rarely complete professional work to a human standard, and with federal data suggesting adoption remains far more modest than headlines imply. The US Census Bureau's Business Trends and Outlook Survey found that just 17% to 20% of American businesses reported using AI from December through early May.
Harris argues agents fail without human context and oversight, and that the function best equipped to provide both is often the last one consulted, especially when restructuring is driven by tech teams rather than workforce strategy.
"The only audience that understands how to audit and manage a workforce that doesn't always do what you tell it to do is HR. So if he did not engage HR and he made this an IT thing, he'd already lost before he started."
A stock market story with a long tail
Meta is hardly alone in miscalculating its AI timeline, Harris noted, and the pattern usually ends the same way.
"Almost all of them have come back and rehired a good portion of their workforce," she said of companies that cut deeply in AI's name.
"This is very much a stock market driven conversation. It looks better to lay off to pay for some of this. And nobody's looking at the long term costs, and the stock market isn't willing to understand the long term cost because the stock market keeps going up every time we do it."
The bill, she believes, eventually comes due.
"I think long term, those companies will definitely pay the price. They will not probably be able to move as fast as they would have if they would have done it the right way."
None of this makes Harris an AI skeptic. She expects the technology to fundamentally reshape work, and the organizations doing it well treat their investments in talent management and HR technology as the foundation that makes AI viable.
"The companies that are really effective at this are realizing that the foundation they built through talent management and HR technology investment is what makes AI viable."
