Plurai raises $13 million to test AI agents using simulations

In short: Plurai raised $13 million to help companies make AI agents more reliable by testing them in realistic simulations before they go live.

What happened

Plurai, a startup founded by Ilan Kadar and Elad Levi, announced a $13 million seed funding round. The round was led by Team8, with participation from U&I, Velocity and Mercer.

Plurai is focused on “AI agents,” which are AI tools that can take actions for you, like booking travel, answering customer questions, or pulling information from company databases. Benchmarks suggest today’s agents often fail too often to run without close supervision. For example, LiveBench 2.0 reported scores of 82.9 out of 100 tasks for Claude Fable 5.2 and 75.6 for ChatGPT-6 Astra, and a separate test by EON found Claude Opus 4.8 at 76.8% accuracy in a simulated office setting.

Plurai’s pitch is to use the same idea that helped self-driving car teams improve safety, which is lots of simulation. It builds a “digital twin” of a company’s systems (a realistic copy, like a flight simulator for software), including the tools and knowledge an agent would use at work. Plurai then maps the different paths an agent might take, tests edge cases (rare but important situations), and helps teams find and fix weak spots.

The company says it can also use the data from these tests to further train smaller “open-weight” models (models a company can run and adjust itself) tailored to a specific business.

Why it matters

Many people are starting to interact with AI agents through customer support, travel booking, and workplace tools. If these agents are unreliable, they can waste time, break company rules like spending limits, or give wrong answers. Tools that test agents before they are deployed could mean fewer mistakes for customers and less cost for businesses.

Source: TechCrunch AI