Pomegra
Plato raised a $16M Series A led by Nava Ventures to build simulated workplaces where AI agents practice real jobs before touching live business systems.
Key Takeaways
- Plato closed a $16M Series A led by Nava Ventures, bringing total funding to $19M.
- Gradient Ventures, Transpose Platform and Microsoft AI CEO Mustafa Suleyman also invested.
- Valuation was not disclosed. Plato was founded in 2025 and is based in San Francisco.
Lead
Plato, a San Francisco startup founded in 2025, announced a $16M Series A on October 28, 2026. Nava Ventures led the round. The company builds simulated workplaces, replicas of real jobs and software, where AI agents can train and be tested without risking live money or records. Total funding now stands at $19M, which implies a seed round of roughly $3M. That seed was led by Gradient Ventures, and the company did not disclose its size.
What Does Plato Build?
Plato builds repeatable software environments that mimic specific jobs, so AI developers can train agents by trial and error. The environments cover customer support, finance, accounting, HR and healthcare administration. The company says its simulators span 10 domains, 36 roles and 210 applications.
Each environment resets after a run, so results can be compared across attempts and across models. Plato says it grades both the end result and the path an agent took, including whether it followed the rules. The company also publishes two benchmarks: Computer Work Val v1.0, a 300-task validation set, and Web Clone Bench, which tests how well a model can recreate an application.
The company pays professionals to rebuild their own jobs as simulations. That human labor is the core cost and, if it works, the core moat.
Who Backs the Round, and What Is Plato Worth?
Nava Ventures led the round, with Gradient Ventures, Transpose Platform and Mustafa Suleyman participating. Suleyman is CEO of Microsoft AI. Freddie Martignetti, Nava's founder, is joining Plato's board.
Plato did not disclose a valuation, and it did not name customers. For a company roughly a year old, that leaves the revenue picture open. A $16M A round on a $3M seed suggests investors are paying for the founders' résumés and the pace of the category as much as for traction.
Who Is Running It?
CEO Robert Farlow previously built an annotation platform at Yutori for training computer-use models. CTO Pranav Putta has a background in computer vision, robotics, reinforcement learning and computer-use models at MultiOn. Both came from the data-labeling end of agent development, which is the work Plato now tries to replace with full environments.
"Mastering real complex work comes from experience doing those tasks," Farlow said in the announcement.
Why Are Simulated Workplaces Drawing Capital?
Agent builders have run into a data problem: a model that can click through a browser still lacks experience in how an accounting close or an insurance claim actually unfolds. Static datasets capture outputs, not the sequence of decisions behind them. Environments capture the sequence.
Money is moving in this direction. In October 2026, micro1 announced $1B in spending on enterprise data for similar training environments. In June 2026, Patronus AI raised $50M for stress-testing agents in simulations. Plato's $16M is small beside both, which frames its pitch as a specialist play rather than a scale play.
The counterargument is sturdy. A simulation is only as good as its fidelity to the real job, and exceptions, messy data and office politics are hard to script. Frontier labs may also build environments in-house, which would shrink the market for third-party vendors. Plato's response is breadth and the use of paid domain experts, though neither guarantees that labs will buy rather than build.
How Will Plato Spend the Money?
Plato will use the funds to expand its library of work environments and hire more professionals to recreate their jobs. It also plans to deploy agents on real tasks to collect additional experience. A further goal is richer simulations with longer workflows, since multistep jobs spanning hours or days are where current agents fail most often.
Outlook
Plato enters a category where larger competitors already hold bigger checks and, in micro1's case, a stated $1B budget. The next evidence will be commercial: named AI-lab customers, benchmark results that other labs adopt, and the cost of paying experts per simulated role. Until a valuation or revenue figure surfaces, the round reads as a bet that realistic practice, not more text, is what moves agents from demos to jobs.
