Claude Cowork, AI hype, and its real impact on white-collar work

Before the holidays, Adam Conner began vibe coding. Like everyone else in the know, he was using Claude Code. Compared to popular chatbots, Anthropic’s advanced AI agent speaks the language of computers: code. Normally, you click buttons in browsers, open folders, and drag files. But you can also do so by coding—interacting with software by typing commands into a terminal, a text-based app.

Claude Code goes beyond such primitive tasks, though: an AI that can code can effectively do nearly anything on a computer. 

“We expected developers to use Claude Code for coding, but then something unexpected happened,” an Anthropic spokesperson tells Fast Company. “We started seeing the discovery arc where people would approach Claude Code to tackle a coding task, then have an ‘aha moment’ when they realized it could help with other tasks.”

The result of that ‘aha moment’ is a vibe coding phenomenon that lets developers—and, crucially, non-developers such as Conner—harness the AI agent to write code and create projects that could grow tomato plants, knit sweaters, and build fully fledged iOS apps in hours. 

Conner’s vibe coding wasn’t as dramatic as keeping living organisms alive, but no less impressive. He used it for his work: an AI labor market simulator, built from Bureau of Labor Statistics and Federal Reserve data. It projects the potential impact that AI could have on the economy. (Headline: it’s pretty big.) 

“I got it running in a day,” says Conner, who’s also the vice president for technology policy at the Center of American Progress (CAP), a think-tank. “It shows how this AI is more accessible and powerful. You can have limited programming experience, and still build something quickly.”

But the real revelation, says Conner, came using Claude Code to form his own policy council: 21 AI agents with competing ideologies and political agendas. Within minutes, it had generated 24-page proposal papers, 12-page draft legislations, and hundreds of policy ideas.

“You can now direct a small army of bots to do the tasks of humans relatively quickly,” says Conner. 

“It’s not yet the game changer that can fully automate someone’s job, but you can begin to see how AI could be transformational.”

We’re now in the fourth year of widely available generative AI tools. As adoption has ticked up—23% of U.S. workers are now frequent AI users, nearly doubling year-over-year—so too have AI-related layoffs.

Despite many companies citing AI as the cause for layoffs, though, actual mass job displacement due to AI has yet to materialize. But the capabilities of Anthropic’s Claude Code—and in particular its advanced AI workplace tool, Claude Cowork—could change that.

The new knowledge worker

So far, generative AI’s use case has been split fairly equally between work and personal needs, such as generating ideas, editing, and even companionship. In 2025, bots were more often leveraged for therapy, life coaching, and, increasingly, coding—the latter likely driven by the popularity of Claude Code: generating more than $1 billion in revenue, just six months after general availability, according to Anthropic. 

On January 12, the San Francisco-headquartered company launched Claude Cowork. It’s effectively a UI update to Code for the mass market. Whereas Code can only be utilized through a terminal’s command line—the barebones essence of human-to-computer interaction—Cowork adds the friendly, compliant chatbot layer to the Claude desktop app.

Currently, anyone with a $100-per-month Claude Max subscription can prompt the AI to complete nearly any computer related task. 

Compared to popular chatbots like ChatGPT, Perplexity, or the original Claude.ai, Cowork can, in theory, leverage users’ hard drives––their digital lives––in its working memory. 

In practice, as a workplace tool, Cowork can organize haystacks of files into neat, delineated folders; turn screenshots of invoices into actionable spreadsheets; pull material from multiple websites, synthesized and analyzed in a single document; and even action slide comments. 

In short, it’s a general AI assistant, built for knowledge work.

“It’s like having a junior researcher that’s proactive, makes few mistakes, and solves problems before you realize there even was one,” says Tomas Chamorro-Premuzic, professor of business psychology at University College London and Fast Company contributor who’s written books about AI. Compared to other AI platforms, Cowork is smarter and more autonomous, he adds. 

In contrast to how people typically use ChatGPT—often for their personal lives, such as travel plans, self-therapy, or recipes—Anthropic’s latest agent can be more entrusted with their work. “I use Cowork to digest research: ‘Upload this folder and create a review of these articles’ main points, then look online for everything I’ve written on this topic.’ So you can give it multi-layered tasks and receive highly accurate outputs,” says Chamorro-Premuzic.

These sorts of tasks—bite-sized chunks of reading, researching, and writing—are traditionally delegated to entry-level employees, says Conner. It’s how knowledge workers’ judgment and expertise is honed on their gradual climb up the career ladder. “If AI takes over some of those junior-level processes, it could pull the ladder up from them,” he adds.

The risk is that the effectiveness of AI agents like Cowork, and their widespread adoption, could impact entry-level hiring, continues Conner. This has long-term consequences for the economy. And in some ways, it’s already impacted how many entry-level jobs are available right now; some data suggests a 35% plummet in entry-level openings in the U.S. since 2023.

“Fewer junior roles eventually leads to fewer qualified candidates for mid-level and senior roles,” Conner says. “So, while short-term financial focus means some companies may want to adopt AI more and hire fewer junior folks, it’s possible we look back years from now and realize it hurt the talent pipeline.”

Coding the future

Cowork has already deeply impacted the economy, just over a month after launch.

On February 3, $300 billion was wiped from software and data stocks. It swiftly followed Anthropic’s release of plugins that can tailor Cowork to specific roles across sales, legal, financial analysis, and other industries. “The worry is that if in-house lawyers can just use Cowork to do their work, they no longer need legal software,” says Chamorro-Premuzic. 

He believes Cowork is an incremental upgrade, rather than exponential. But the next iteration will likely come soon. On February 5, Anthropic released Claude Opus 4.6. It has four times the effective recall of Opus 4.5, its predecessor that made Claude Code go viral—launched just 73 days before.

The pace of release is accelerating (Sonnet 4.6, Anthropic’s free model, was released on February 17). That’s not just because of the AI race—Claude Code already writes 90% of its team’s code. On December 27, its lead developer said all 40,000 lines of new code he’d contributed in the past 30 days were written by the agentic tool. An autonomous, self-perpetuating feedback loop means the next upgrade is always imminent. 

“We’re going to see AI involved in improving and building itself more and more,” says an Anthropic spokesperson. “In the closer term, Claude Code helps people build faster, with fewer resources, and the result is innovation across entire industries. We’re in the early innings of what’s possible.”

Claude Cowork isn’t perfect. It requires oversight, particularly with high-stakes tasks. Like all AI, it can also make mistakes—and given it can have access to personal files, apps, and tools, those errors can be more widespread than a chatbot’s text output. It’s also not compliant first: conversations histories are stored locally, rather than within tightly regulated workflows.

However, mass adoption of advanced AI tools like Cowork will likely be inevitable. So will labor market displacement—it’s only the extent that’s unclear. 

There are already signs that junior workers and college graduates are disproportionately affected by the onward march of AI. A November 2025 Stanford University study found a 16% relative decline in employment for early-career workers in occupations most exposed to the technology since 2022, such as software developers and customer service representatives.

Not all entry-level jobs will be decimated by AI. Companies will always want to hire the best talent, and train them over time to accumulate institutional knowledge, workplace culture, and project history in ways technology can’t.

But there are concerns over how roles may be reshaped.

If knowledge work merely becomes entering and fine-tuning prompts, it risks workers automating themselves, hampering their development and soft skills. “Generative AI can tell you what to say in giving critical feedback, but it’s not the same as having a hard conversation and learning from it,” says Chamorro-Premuzic. “It becomes judgment without experience.”

The power of Claude Cowork means we’re one step closer to that future.

“It’s like seeing the Wright brothers fly for the first time,” says Conner. “You wouldn’t have understood the concept of a Boeing 767 flying across the ocean, but you’d have grasped the idea of aviation. With Claude Cowork, you can see the future more clearly, and how this technology will have a major impact.”

View the full article