A San Francisco-based artificial intelligence start-up, Mechanize, has announced its intention to automate white-collar work across industries as quickly as possible, the New York Times reports.
The company’s founders say they envision a “fully automated economy,” in which intelligent systems take on responsibilities traditionally carried out by humans at computers.
Mechanize was founded by Tamay Besiroglu, Ege Erdil, and Matthew Barnett, who previously collaborated at Epoch AI, a research organization focused on the capabilities of advanced AI systems. The new start-up has already attracted support from notable tech leaders, including Stripe co-founder Patrick Collison and Google’s chief AI scientist Jeff Dean.
“Our goal is to fully automate work,” Besiroglu said at a recent event hosted at a San Francisco cocktail bar. “We want to get to a fully automated economy, and make that happen as fast as possible.”
Mechanize is part of a growing wave of companies leveraging reinforcement learning — a machine learning method popularized by systems like DeepMind’s AlphaGo — to teach AI agents how to perform complex, multi-step tasks. These new systems are designed to move beyond single-purpose models by simulating real-world work environments.
One of Mechanize’s current projects involves building a virtual workspace to train AI models for software engineering. In this simulated environment, the system interacts with digital tools like email, Slack, and web browsers to complete assigned tasks. Success is rewarded and failure is penalized, creating a feedback loop to refine the AI’s performance.
Besiroglu described the process as “creating a very boring video game,” with the goal of teaching AI systems to behave like human workers over time. While the company is starting with programming, it ultimately hopes to expand to other professions that rely on computer-based tasks.
The vision of a fully automated workforce is not without precedent. Economists like John Maynard Keynes speculated nearly a century ago about machines taking over most labor, potentially freeing people to pursue more fulfilling lives. Recent advancements in AI have reignited those discussions, with industry leaders such as Anthropic CEO Dario Amodei warning that AI could displace half of all entry-level white-collar jobs within five years.
Still, Mechanize’s approach — and its candid goal of replacing rather than augmenting workers — stands out in a tech industry where automation is often framed in more reassuring terms. Unlike earlier rhetoric about “augmenting” human workers, Mechanize is openly pursuing labor replacement.
Some critics have raised ethical and practical concerns about this ambition. While the company believes AI could one day lead to “radical abundance,” potentially supporting laid-off workers through universal basic income, it has not proposed specific policies for how to manage the transition.
Matthew Barnett, one of the co-founders, said he believes the benefits of automation will ultimately outweigh the downsides, arguing that higher economic growth and technological progress would improve societal well-being, even if many current jobs become obsolete.
Mechanize estimates it could take 10 to 30 years to reach full automation, depending on the role and complexity of the tasks involved. Despite these long-term projections, the company is moving quickly to train AI models capable of performing increasingly sophisticated work.
