A San Francisco startup that builds pretend workplaces for artificial intelligence agents has raised a $16 million Series A, the company said Wednesday. Nava Ventures led the round. Gradient Ventures, Transpose Platform and Microsoft AI chief Mustafa Suleyman also joined.
What does Plato build?
The company, Plato, copies the software and tasks of real jobs in fields such as customer support, finance, accounting, HR and healthcare administration. AI labs use those copies to train agents through trial and error, then test how well the agents perform. The simulations keep agents away from real money and real records.
“Mastering real complex work comes from experience doing those tasks,” said Robert Farlow, Plato’s co-founder and chief executive.
How does Plato test whether an AI agent is ready?
Plato says it does not judge an agent only by its final result. The company also checks how the agent reached that result, including whether it followed the rules. That approach, Plato says, keeps an agent from scoring well by gaming a test instead of doing the job properly.
Suleyman, one of the new backers, has argued that AI should be provably under control before it is released.
Plato was founded in 2025 and has raised $19 million in total. The new funding will go toward paying professionals to rebuild their own jobs as simulations. It will also be used to send AI agents out on real tasks so they can gather more experience.
Why is demand for AI agent training environments growing?
Demand for these training grounds is climbing. Earlier this month, AI training firm micro1 said it would spend $1 billion on company data to build similar practice environments for agents. In June, San Francisco-based Patronus AI raised $50 million to stress-test agents in simulated settings.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.