Plurai wants to help AI agents work better using the same approach that helped get self-driving cars to roads

AI agents often fail to perform reliably in complex, real-world tasks, with benchmark tests showing success rates around 75-83%. Plurai, a startup founded by Ilan Kadar and Elad Levi, aims to address this by applying simulation techniques from autonomous driving. The company creates a digital twin of a company's systems, maps possible agent flows, and uses synthetic data to prepare models for edge cases. Plurai also post-trains small, open-weight models tailored to specific customer needs, claiming they outperform mid-tier frontier models at significantly lower costs. The startup has raised $13 million in seed funding led by Team8, with participation from U&I, Velocity, and Mercer. Investors like Google DeepMind researcher Reah Miyara believe Plurai's focus on failure modes and cost-effective tuning gives it an edge as businesses seek sustainable AI expenses. Kadar emphasizes that even advanced general models won't know a specific business's policies, highlighting the need for tailored simulation-based training.

(Source：TechCrunch)