Smar NetworkSmar.inSmargrid.comNewsInsuranceRoomsGamesMeasurements

SMAR Daily Industry News

Daily news across AI, technology, finance, insurance, healthcare, automotive, real estate, and government

Facebook Telegram
← Back to AI News

Plurai Secures $13M to Improve AI Agent Reliability

📅 2026-11-11 | 📰 TechCrunch AI
#plurai #ai agents #synthetic data #simulation #enterprise ai

AI agents are currently facing reliability challenges, struggling to perform complex tasks without human oversight. Recent benchmarks, such as LiveBench 2.0, indicate that models like Claude Fable 5.2 and ChatGPT-6 Astra successfully complete 82.9 and 75.6 out of 100 tasks, respectively. Additionally, EON’s simulation of corporate environments found that the top-performing model, Claude Opus 4.8, achieved 76.8% accuracy. To address these limitations, a new startup called Plurai has raised $13 million in seed funding to improve agent performance using methodologies originally developed for self-driving cars.

Plurai’s co-founders, Ilan Kadar and Elad Levi, bring experience from their time at Nexar, an autonomous driving technology firm. Kadar previously collaborated with Waymo on simulating and evaluating deep learning driving software. The startup aims to apply similar simulation techniques to digital systems, creating digital twins of a company’s knowledge base and internal tools. By mapping potential agent workflows, the platform prepares models to handle difficult edge cases that often cause failures, such as agents ignoring spending policies during travel bookings or struggling to synthesize information across multiple databases.

The startup’s platform is designed to integrate with existing tools like Claude Code, Codex, Copilot, and Cursor. Beyond simulation, Plurai focuses on using gathered data to post-train smaller, open-weight models customized for specific client requirements. According to the company, tests conducted alongside Nvidia and Nebius demonstrated that these specialized models could outperform mid-tier frontier models while offering significant cost reductions. Kadar noted that by decomposing tasks and using simulations to verify quality, they can achieve performance comparable to or better than frontier models at a fraction of the cost.

Google DeepMind researcher and angel investor Reah Miyara highlighted that Plurai’s strength lies in its ability to identify production failure modes and generate synthetic data for model tuning. As businesses grapple with the high costs and sustainability of using frontier-capable models that may not align perfectly with their specific data, Plurai’s approach offers a potential path toward more efficient deployment. The company draws a parallel between digital agents and physical robots, suggesting that while general AI models are powerful, models tailored for specific tasks are currently better suited for real-world, reliable application.

For IT leaders and enterprise decision-makers, this development signals a shift toward prioritizing specialized, cost-effective models over general-purpose frontier systems. As companies continue to integrate AI agents into corporate workflows, the ability to simulate and tune models against specific internal policies and data sets may become a critical factor in achieving reliable, production-ready automation.

Why it matters

Plurai's approach to using simulation for AI agent reliability could offer enterprises a way to reduce costs while improving the accuracy of AI systems tailored to their specific business policies.

FAQ

How much funding did Plurai raise?
Plurai raised $13 million in seed funding in a round led by Team8, with participation from U&I, Velocity, and Mercer.

What is the core technology behind Plurai's solution?
Plurai uses simulation techniques and synthetic data, similar to those used in autonomous vehicle development, to create digital twins of a company's systems and prepare AI models for edge cases.

Which existing tools does Plurai's platform support?
The platform works alongside existing agentic tools including Claude Code, Codex, Copilot, and Cursor.

Source: TechCrunch AI. See the original source for full details.