Plato raised $16 million to build simulated workplace environments where AI developers can train and evaluate models on realistic multistep tasks.
San Francisco-based Plato announced a $16 million Series A on October 28, led by Nava Ventures, with participation from Gradient Ventures and Transpose Platform. Mustafa Suleyman, Microsoft’s AI chief, also participated. The round brings Plato’s total funding to $19 million; Nava Ventures founder Freddie Martignetti will join the board.
Plato builds simulated work environments where AI models can practice tasks without affecting live business systems. The company says its simulations recreate tools, changing circumstances and job-specific requirements, and can be reset for repeated training. Its initial focus includes customer support, human resources, finance, accounting and healthcare administration.
The environments are intended for reinforcement learning: models attempt tasks and receive feedback. Plato says its evaluations consider not only whether a model reaches an outcome, but also whether it follows instructions and respects constraints. For example, a simulated purchase would not count as successful if the agent exceeded its budget; the evaluations also look for attempts to exploit weaknesses in a simulation.
That approach addresses a practical risk in training agents directly in production software: an agent could spend real money or change real records while learning. Simulations let developers test multi-step behavior in a contained setting, though results in a simulated workplace do not by themselves establish how an agent will perform in an actual organization.
Plato says it will use the new capital to expand its library of work simulations. Its plans include paying professionals to recreate job tasks in simulated environments and deploying agents on real tasks to collect additional experience. The company was founded in 2025 by Robert Farlow and Pranav Putta, who have backgrounds in computer-use model training and machine learning research.