Startup Plurai raises 13 million dollars for a simulation platform for more reliable AI agents
The company Plurai raised 13 million dollars in a seed round led by the firm Team8. It is developing a platform that uses digital twins of enterprise systems and synthetic data to test and improve the reliability of AI agents deployed in enterprise processes.
The company Plurai announced it raised 13 million dollars in seed funding. The round was led by the firm Team8, with participation also from U&I, Velocity, and Mercer. Founders Ilan Kadar and Elad Levi previously worked for the company Nexar, which develops autonomous driving technologies, with Kadar having collaborated with the company Waymo on simulation and evaluation of its software for end-to-end driving.
The Plurai platform creates a digital twin of a company's electronic systems, including the knowledge base and tools that the AI agent will have access to in production. It maps out all possible agent flows and, based on this graph, prepares evaluations and fixes for potential edge-case situations. According to the company, the tool is designed to work alongside existing agentic tools such as Claude Code, Codex, Copilot, and Cursor. The company also uses the collected simulation data to fine-tune smaller open-weight models customized for a specific customer; according to the company, tests conducted in collaboration with Nvidia and Nebius showed that smaller models fine-tuned in this way achieve comparable or higher quality than mid-tier frontier models at significantly lower costs.
As context, the source cites results from the LiveBench 2.0 benchmark, where the model Claude Fable 5.2 scored 82.9 out of 100 tasks and the model ChatGPT-6 Astra scored 75.6 out of 100, as well as a test by the company EON in a simulated corporate environment, where the best-tested model, Claude Opus 4.8, achieved an accuracy of 76.8 percent. According to investor Reah Miyar, a researcher at the company Google DeepMind who invested in Plurai as an angel investor, the company's advantage lies in understanding failures in production deployment and generating synthetic data for fine-tuning models.
Why it matters
Companies deploying AI agents in enterprise processes (customer service, travel booking, database queries) run into agent unreliability beyond simple scenarios. Plurai's approach offers a way to identify and fix such edge cases before production deployment, and according to the company, it also makes it possible to replace expensive frontier models with cheaper fine-tuned open-weight models without a loss of quality.
Two audiences, two different impacts
What this means
For individuals
Benchmarks cited in the source show that even the best AI agents built on top-tier models achieve only 75-83 percent accuracy on complex tasks, so agent outputs still need to be verified.
For a business
For companies planning to deploy AI agents in enterprise processes, an alternative to expensive frontier models is emerging: simulation testing of edge-case scenarios and fine-tuning smaller open-weight models on custom data, which according to the company Plurai reduces operating costs while maintaining comparable quality.
ProcessesCheck the original
Event sources
only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.