Plurai Raises $13M for Agent Simulation as Databricks Updates Genie

Enterprise adoption of artificial intelligence agents accelerated on Wednesday as startup Plurai secured $13 million in seed capital to apply autonomous-driving simulation techniques to business software, Databricks unveiled an expanded multi-step version of its Genie agent, and new financial disclosures highlighted surging commercial revenue across leading model providers.

Plurai Raises $13 Million to Simulate and Post-Train AI Agents

Addressing reliability gaps reflected in recent industry benchmarks—where LiveBench 2.0 showed Claude Fable 5.2 and ChatGPT-6 Astra succeeding on 82.9 and 75.6 out of 100 tasks, respectively, and AI infrastructure firm EON found Claude Opus 4.8 performing simulated corporate tasks at 76.8% accuracy—startup Plurai announced $13 million in seed funding led by Team8 with participation from U&I, Velocity, and Mercer. Co-founders Ilan Kadar and Elad Levi previously worked at autonomous-driving technology firm Nexar, and Kadar also worked with Waymo simulating and evaluating end-to-end deep learning driving software.

Plurai builds a digital twin of an enterprise’s electronic systems, knowledge bases, and production tools—integrating alongside Claude Code, Codex, Copilot, and Cursor—to map possible agent flows and generate synthetic data for edge cases. Backed by angel investor and Google DeepMind researcher Reah Miyara, Plurai also uses its simulation data to post-train smaller open-weight models; in tests conducted with Nvidia and Nebius, the company reported matching or exceeding mid-tier frontier quality while reducing costs by one to two orders of magnitude.

Databricks Unveils Multi-Step Genie Agent for Executive Insights

At enterprise data platform Databricks, chief executive Ali Ghodsi unveiled an updated version of the company’s Genie AI agent powered by an internal enterprise ontology that enables unsupervised multi-step workflows. Demonstrating his own daily “dogfooding” of the tool rather than a canned demo, Ghodsi showed how Genie analyzed low-scoring sections of an employee satisfaction survey, evaluated executive remediation proposals, summarized a difficult sales pitch to a prospective customer, and prepared him for a podcast interview.

Citing British anthropologist Robin Dunbar’s theoretical limit of 150 stable human relationships, Ghodsi noted that leading a 12,000-employee organization makes it difficult to hear candid negative feedback directly, positioning Genie as an enterprise alternative to consumer tools such as OpenClaw, Instinct, or Muse while using Databricks’ Unity Gateway to cap usage and swap in lower-cost open-weight models.

OpenAI Quarterly Revenue Crosses $10 Billion as Enterprise Race Intensifies

Meanwhile, Bloomberg reported that OpenAI’s third-quarter revenue crossed $10 billion for the first time, representing a nearly 50% jump from $6.7 billion in the second quarter following an earlier 18% sequential increase on $5.7 billion in quarterly sales. OpenAI’s adjusted operating margin narrowed from -150% in the second quarter to -109% in the third quarter, lifting its annual revenue run rate to $50 billion across its user base of more than 1 billion users, even as archrival Anthropic—which counts cloud partner sales differently and is targeting a $1.4 trillion IPO valuation—reported $17 billion in third-quarter revenue and an $81 billion annual run rate after adding $10 billion in a single month through October.