Feature

No employees and prototypes within days: AI agents supercharge start-ups

Simon Kuper reports on founders using AI agents for coding, research, prototyping, marketing and even lab construction. In Helsinki, Claudio Reiter and Leon Busch have built two automated biotech labs and now have an agent coordinating a third, managing a budget of about $500,000 including instrument buying and contractor instructions; other agents run simple experiments. Named examples: Rig (Toby Mather, data integration), who says he did 18 months of development in three months after Claude Opus 4.5 arrived on 24 Nov 2025; Structured AI (Isabel Greenslade, drawing review), which returned with a prototype two days after recording a customer conversation and asking Claude Code to build it; GetMint, Malt and Impact Advantage. Investors quoted include Phoenix Court's Ziv Reichert ("It's never been easier to start a company") and The Embassy's Helena Gagern, who describes a drug company with no employees whose candidate reached animal testing. Formation data: per Apollo, weekly US business applications never averaged 80,000 in the 15 years to 2019 but have almost always exceeded 100,000 since ChatGPT launched; Beauhurst/NatWest count UK application-software incorporations up 40% to 28,100 in the first half. Counterweights: output can be mediocre or repetitive, agents add supervision work, and Malt's Vincent Huguet and GetMint's Joan Burkovic stress that people still decide the outcome and that fast rises can mean faster falls. Caveats: the most striking zero-employee cases (the lab builder and the drug company) are anonymous, with no revenue, funding or investors, and the formation data is not shown to be caused by agents.

Why it matters

It goes straight to Dealroom's core data: if agents cut the team needed to launch, expect more and smaller company formations, earlier prototypes and later or smaller first hires, which changes how start-ups should be counted and screened. The +40% UK software incorporations and the US application surge are signals to check against Dealroom's own new-company data.

Read the full article: Financial Times

More top stories