Last year, Anthropic CEO Dario Amodei warned that artificial intelligence could eliminate half of entry-level white-collar jobs and that in six months it would be writing 90% of code. Neither of those predictions came to pass.
Now, in a report Anthropic published on Thursday, the company said it was hiring fewer junior technical workers and that human attention has become the new “bottleneck” as it increasingly relies on its Claude AI agents.
Anthropic says AI has changed the way work gets done, and as a result, restructured employment at the company. The report suggests its changes foreshadow the future of work more broadly: as more companies adopt AI agents, software engineers could face a tougher job market while entry-level tech workers struggle to build necessary skills and experience.
“As AI diffuses and improves, the organization of work could shift,” Anthropic’s report reads before offering a potential future. “One possibility is single-person startups, each with a human founder steering swarms of agents to rake in billions in revenue.”
In its report, Anthropic touts its “deep adoption of AI”, which it claims resulted in “producing more code per person, closing more sales per rep, and processing more contracts”. By outsourcing some coding and sales-related tasks to AI agents, Anthropic said it raised productivity within a few months.
From May 2026 onwards, AI agents logged more work hours than human employees, the report said. Employees now spend about 22% of their time managing Claude agents, more than double the amount a year ago, based on Anthropic’s analysis of their chat summaries.
It also says since March 2025 – soon after the release of its agentic AI coding tool Claude Code – entry-level technical hires dropped from nearly 30% to 9%, and headcount dropped from about 21% to 14%. It attributes the shifts to AI usage, and described its workers as merely “a thin layer atop a growing contingent of AI agents”.
While the report provides a small glimpse of the inner workings of the AI heavyweight, some AI researchers are skeptical that it will herald big changes across companies or industries. Instead, they say the findings are likely Anthropic’s attempt to drive its valuation ahead of its initial public offering – similar to how SpaceX boasted about proposed space data centers to boost its market debut in June.
Anthropic’s claims seem to back some tech executives’ view that AI is behind recent waves of corporate layoffs, but research complicates that.
Morgan Frank, a Case Western Reserve University researcher on how AI and technological changes affect workers, recently posted a paper that’s in peer review showing that AI-exposed job markets deteriorated well before the arrival of ChatGPT in November 2022 and Claude a few months later. State unemployment claims reveal that job markets tanked starting at the beginning of 2022, months before ChatGPT’s release, according to the paper.
“It doesn’t appear that the launch of ChatGPT corresponds to any major increase in loss or worsening outcomes,” he said.
Frank said several factors could explain recent job losses, including cutting pandemic-era hires and “AI-washing”, where some bosses credit AI for layoffs they might have made anyway.
Julián Posada, a Yale social scientist, also questions Anthropic’s motives for publishing its report.
“It is written from the perspective of a company that wants to sell shares, a company that really wants people to invest in it,” said Posada, author of the book Platform Extractivism: Data Work and the People Powering Artificial Intelligence.
Unlike an economics study, the report lacked an independent assessment by outside auditors, Posada said. Its figures and graphs only tell part of the story, without revealing how Anthropic’s workers feel about their work or how they’re actually engaging with the models, he said.
Dylan Baker, lead research engineer at the non-profit Distributed AI Research Institute, said he was also curious about how Anthropic’s developers feel about the changes. Shifting some software development to AI, while hiring few entry-level workers and having the remaining engineers wrangle AI agents seems “shortsighted”, Baker said.
Baker is also surprised by Anthropic’s human “bottleneck” characterization.
“That’s such a strange way to put it,” Baker said. “It fundamentally misunderstands how software is written, architected and designed. It’s such a social, interpersonal and human process.”
Anthropic’s report quotes an unnamed engineer, who described facing “an army of Claudes, with every one asking for my attention”. That doesn’t sound like a fruitful work environment, Baker said. “I wouldn’t think really thoughtful, well-designed products are coming out of that team.”
As generative AI products proliferate, a majority of Americans are wary of AI, with particular concerns about AI-related job losses, according to Pew Research Center. In its report, Anthropic suggests that tech employers should offer their workers stock ownership, posing this as a potential solution to rising concerns. “It is difficult to imagine a version of pro-worker AI adoption in which workers do not hold a financial stake in successful rollout,” it said.
Anthropic did not respond to the Guardian’s request for comment.
Anthropic, which could debut on the stock market as soon as November, released its employment report one day after pledging to commit 5% of its shares to the “public good”, though it didn’t provide details. Critics noted that philanthropy has long doubled as a reputation builder used by tech titans like Bill Gates and Mark Zuckerberg.
But Anthropic, and its claims it’s improving how work gets done, could face tougher scrutiny as it nears its IPO. A draft of the company’s S-1 filing revealed that Anthropic had a $80bn net loss over the first half of this year, and that it received a junk credit rating in June. Considering Anthropic’s slowing revenue growth, investors now project a valuation of $1.5tn, down from earlier estimates of $2tn or more. At the same time, Anthropic’s cloud computing partner, CoreWeave, has become swamped in AI debt and its stock is plummeting, fueling concerns that the AI bubble may begin to deflate.
Regardless of when or whether the bubble pops, software engineers and other tech workers will surely be affected, Baker said. The question is how that will play out.
“My hope is that there is a reckoning, a reassessment or re-evaluation of how much this has always been about human work,” Baker said.