Plurai Raises $13M Seed to Simulate AI Agents Like Self-Driving Cars

Plurai has raised $13 million in seed funding to apply self-driving car simulation techniques to enterprise AI agents, the startup announced Wednesday. Team8 led the round, with participation from U&I, Velocity and Mercer.

The company targets a documented reliability gap. LiveBench 2.0 shows Claude Fable 5.2 and ChatGPT-6 Astra succeeding on 82.9 and 75.6 out of 100 tasks respectively. In a separate test, AI infrastructure company EON placed agents in a simulated corporate environment and found the strongest model tested, Claude Opus 4.8, performed tasks at 76.8% accuracy.

Co-founders Ilan Kadar and Elad Levi previously worked at Nexar, which develops autonomous driving technology. Kadar worked with Waymo to simulate and evaluate its end-to-end deep learning driving software.

"The moment that we saw the wave of AI agents, we saw the same pattern: That it is very easy to build a demo, but [hard] to get something that will be reliable and cost effective," Kadar told TechCrunch. "Since we saw the same pattern, the same pitch was that we want to build this simulation layer."

What goes wrong with agents today?

AI agents share a structural weakness with end-to-end driving models. Both are probabilistic systems operating in unpredictable environments. The failure modes are concrete:

• Travel-booking agents can ignore customer spending policies under pressure.
• Agents querying a single database answer reliably, but combining multiple data sets tends to fail.
• Long customer service conversations can overwhelm an agent with context and produce wrong turns.

How does the simulation layer work?

Plurai builds a digital twin of a company's electronic systems, including the knowledge base and tools an agent will access in production. It then maps all possible agent flows, and uses that graph to prepare the model with evaluations and fixes for edge cases. The platform runs alongside existing agentic tools such as Claude Code, Codex, Copilot and Cursor.

Kadar says the larger promise lies elsewhere. Plurai wants to use the gathered data to post-train small, open-weight models tailored to individual customers. Tests run jointly with Nvidia and Nebius showed Plurai's models outperforming mid-tier frontier models at lower cost.

"We can take a large language model, we can take the task, decompose it into a smaller task using the simulation, train [an open-weight] model, and using the simulation, verify that the quality is higher or comparable to the frontier model, and an order or two orders of magnitude cost reduction," he said.

Why cost, not capability, may decide the market

Reah Miyara, a Google DeepMind researcher who backed Plurai as an angel investor, said the company's edge comes from understanding failure modes in production and generating synthetic data to tune models. He framed the commercial pressure in blunt terms, noting IT leaders are seeing that "frontier-capable models are out of distribution with my specific data, and the cost of these things isn't sustainable in the long term."

Plurai draws its thesis from robotics. Powerful general AI models give robots impressive capabilities, but those robots remain too unreliable for wide deployment. Robots powered by task-specific AI models are already in the field, serving customers and collecting real-world data.

"My view is that no matter how smart OpenAI and Anthropic model will be — and they will be much smarter — they will not know your business and your policy," Kadar said.

With $13 million in hand, Plurai will now have to prove that simulation-trained, customer-specific small models can hold their ground as frontier models improve and prices fall.