AI infrastructure startup Plurai has raised $13 million in seed funding to improve the reliability of AI agents used in business operations. The funding round, announced on November 11, 2026, was led by Team8, with participation from U&I, Velocity, and Mercer.
Plurai develops simulation technology that replicates a company’s digital systems, allowing AI agents to be tested and improved before deployment. The startup aims to address reliability problems that prevent businesses from using AI agents independently for complex tasks.
AI Agents Continue to Face Reliability Challenges
Despite improvements in AI capabilities, existing models still struggle to complete complicated workflows consistently. According to LiveBench 2.0, Claude Fable 5.2 and ChatGPT-6 Astra achieved success rates of 82.9% and 75.6%, respectively, on long-running agentic tasks.
A separate study by AI infrastructure company EON tested agents in simulated corporate environments. Its highest-performing model, Claude Opus 4.8, completed tasks with 76.8% accuracy.
These limitations can affect business applications involving customer support, travel bookings, and database operations. Agents may overlook company policies, struggle to combine information from multiple sources, or produce incorrect responses during extended conversations.
Plurai Uses Digital Twins to Test AI Agents
Plurai was founded by Ilan Kadar and Elad Levi, who previously worked at autonomous driving technology company Nexar. Kadar also worked with Waymo on simulations used to evaluate its autonomous driving software.
The founders are applying similar testing methods to AI agents operating in digital environments. Plurai creates a digital twin of a customer’s business systems, including internal knowledge bases, software tools, and workflows.
The platform then maps possible agent interactions and generates simulated scenarios to identify potential failures. These evaluations help developers identify and address problems before agents interact with real customers or business systems.
Plurai’s technology is designed to work alongside existing AI development tools, including Claude Code, Codex, Copilot, and Cursor.
Startup Targets Lower AI Operating Costs With Specialized Models
Beyond testing, Plurai plans to use simulation data to train smaller, open-weight AI models for specific business tasks. These models could handle specialized workflows without relying entirely on larger, more expensive frontier models.
According to tests conducted with Nvidia and Nebius, Plurai’s specialized models outperformed certain midtier frontier models while operating at substantially lower costs. The results were reported by the company and do not establish performance across all business applications.
Kadar said the approach involves breaking complex workflows into smaller tasks, training specialized models, and verifying their performance through simulations.
Google DeepMind researcher Reah Miyara, who invested in Plurai, said the company’s approach could help businesses address AI failures while managing growing infrastructure expenses. The startup is focusing on applications where company-specific knowledge, internal policies, and consistent execution are important.
Featured image credits: Magnific.com
For more stories like it, click the +Follow button at the top of this page to follow us.