Mechanize wants to automate all jobs with AI agents — and controversy is already brewing

New AI startup wants to automate all human labor — and not everyone is cheering

Mechanize, founded by AI researcher Tamay Besiroglu, fuels debate over trust, technology, and the future of work

When AI researcher Tamay Besiroglu announced his new startup Mechanize last week, social media lit up — and not just because of the company's mission. With a vision for “the full automation of all work,” Mechanize aims to build the data, digital environments, and evaluation tools to make AI agents capable of performing any job currently done by humans.

In other words: replace human labor, everywhere.

The reaction was swift and polarized. Supporters praised the bold ambition; critics questioned both the feasibility and ethics of the plan. And some were alarmed by the blurring lines between private ambition and public trust — especially because Besiroglu also co-founded Epoch, a non-profit AI research institute previously viewed as neutral.

Full automation of the economy?

In an announcement post on X (formerly Twitter), Besiroglu laid out Mechanize’s goal: enabling the full automation of the global economy. He framed the company’s total addressable market not in terms of industries or verticals, but in total global wages—estimated at over $60 trillion per year.

“Our immediate focus is on white-collar work,” he told TechCrunch, clarifying that Mechanize is not yet targeting manual labor or robotics-based tasks. The company’s early goal is to build AI agents that can perform knowledge-based work typically done by office employees.

A trust crisis for AI science?

The backlash wasn’t just about the company’s vision—it also centered on its founder’s existing role in AI governance. Epoch, the AI research non-profit Besiroglu helped establish, is known for producing impartial economic analysis and performance benchmarks for large AI models. That perceived neutrality is now under scrutiny.

“Alas, this seems like confirmation that Epoch research was feeding into frontier AI capabilities,” wrote AI observer Oliver Habryka on X, reflecting concern that academic research is being repurposed to serve private business goals.

Epoch was already facing skepticism after revealing last year that OpenAI supported the development of a benchmark used to promote its GPT-4 successor model, o3. The new controversy only deepens doubts about whether privately affiliated AI research can truly be independent.

Who's backing Mechanize?

Despite the controversy, Mechanize boasts an impressive lineup of investors. Nat Friedman, Daniel Gross, Patrick Collison, Dwarkesh Patel, Jeff Dean, and Marcus Abramovitch are among the early backers. Several have stayed silent amid the public backlash, but Abramovitch—an “effective altruist” and hedge fund partner—defended the investment, citing the team’s depth of thought and technical expertise.

“The team is exceptional,” he told TechCrunch. “They’ve thought deeper on AI than anyone I know.”

Will full automation enrich or impoverish us?

Besiroglu argues that automating all labor could usher in an era of “explosive economic growth”, resulting in higher living standards and novel goods and services. He believes that humans, freed from work, could thrive via dividends, rents, and social programs.

Critics are more skeptical. “If people don’t have jobs,” one commenter wrote, “they won’t have income to buy what AI is producing.”

Besiroglu contends that even if wages drop, economic well-being can improve through non-wage income sources. But that presumes a system in which humans retain ownership—or at least benefits—from the AI-driven economy. Without equitable redistribution, critics warn, such a future could worsen inequality.

The tech is real, the problem is hard

Mechanize’s vision may sound extreme, but the problem it addresses is grounded in today’s AI agent shortcomings. Current AI tools often fail at retaining memory, completing tasks autonomously, or executing long-term plans. Besiroglu is focused on solving these gaps by building agent environments where skills can be trained, tested, and scaled.

Others are chasing similar goals. Salesforce, Microsoft, OpenAI, and a host of startups are working on agentic platforms, especially in fields like sales, finance, and coding.

For now, Besiroglu is optimistic—and hiring. But the deeper question remains: Should we trust frontier AI development more when it comes from billion-dollar startups and well-known investors?

Perhaps not. As debates around AI, labor, and ethics intensify, Mechanize may be less an outlier than a preview of the philosophical and practical battles ahead.

Stay tuned to The Horizons Times for critical coverage of emerging AI, labor automation, and the future of human work.