Plato Technologies, Inc. said Wednesday it raised $16 million in a Series A round led by Nava Ventures, with participation from Gradient Ventures, Transpose Platform and Mustafa Suleyman, the CEO of Microsoft AI, who invested personally. The San Francisco startup builds simulated workplaces where AI agents practice real jobs before developers put them to work.
The round brings Plato's total funding to $19 million, according to the company's Oct. 28 press release. It did not disclose a valuation, and rivals in the same category are already raising and spending at a considerably larger scale.
What Plato builds
The company describes its product as "simulated workplaces where AI can learn to do real jobs." Each environment reproduces the tools, decisions and constraints of a given task, and is used to train and evaluate models and agents in contained settings, according to the release.
The Next Web reported that early focus areas include customer support, finance and accounting, HR and people operations, and healthcare administration. The appeal, in theory, is that AI labs can train and test agents without exposing live money or records to mistakes.
Plato's evaluations also examine process, not only outcome. The release said they test whether a model "followed instructions, respected constraints, and completed the task," and they flag reward hacking, a failure in which an agent games its scoring rather than doing the work. A purchase that blows the budget, for example, does not count as a success.
"Mastering real complex work comes from experience doing those tasks," CEO Robert Farlow said in the release. "We're building simulations that capture the tools, decisions, and unexpected situations those tasks involve, so AI models can gain experience in safe, contained environments before developers put them to work."
Investors and use of funds
Freddie Martignetti, founder of Nava Ventures, is joining Plato's board. Gradient Ventures led the company's earlier seed round, per the release, which gave neither the date nor the amount of that financing.
Plato said it will expand its library of real-world work simulations, pay professionals to recreate their jobs as simulations, and deploy AI agents on real tasks to collect additional experience.
"Robert and Pranav are deeply talented researchers and developers from Waterloo and Georgia Tech," Martignetti said in the release. He said the founders "have developed a unique way to create high fidelity environments at scale," and that Nava "quickly learned that Plato is really a data lab."
Founders and team
Farlow previously built an annotation platform at Yutori to support the training of computer-use models, according to the press release. Pranav Putta, the chief technology officer, has a background in computer vision, robotics and reinforcement learning, and earlier trained computer-use models at MultiOn, the release said. Plato was founded in 2025.
Headcount is hard to pin down. LinkedIn lists 11 to 50 employees, while Wellfound lists 1 to 10. The company's site shows open roles, including a research scientist for safety environments and a Head of AI Quality.
Plato named no customers or pilots, and Citybiz noted that none were identified in coverage of the round.
Benchmarks and competition
Plato's website lists two benchmarks. Computer Work Val v1.0, posted Oct. 27, covers 300 worlds drawn from the work of people earning more than $100,000 a year. Web Clone Bench v1, posted Oct. 5, tests whether models can rebuild web apps such as Gmail, Jira and Slack within set time budgets.
Competitors have moved quickly. Patronus AI announced a $50 million Series B on June 25 to build what it calls "Digital World Models" for training and evaluating agents, TechCrunch reported. On Oct. 9, micro1 said in a company blog post that it would spend $1 billion over 12 months on enterprise data acquisition to build reinforcement learning environments.
Against that backdrop, Plato's plan is comparatively modest: hire, build out its simulation library, and put the new money toward paying professionals to recreate their work.
