Traversal Raises $120M at a $1B Valuation to Put Causal AI Agents on Enterprise On-Call

Traversal, the New York company building AI agents for enterprise site reliability engineering, has raised $120 million at a $1 billion valuation—its largest check yet as customers put the product on real on-call. The November 3, 2026 company announcement says Kleiner Perkins, Redpoint Ventures, and Khosla Ventures co-led the round, with Sequoia Capital, Battery Ventures, and Amex Ventures participating. Kleiner Perkins, which led the Series A, increased its stake; Sequoia, which led the seed, put growth capital behind the company beyond its existing position. Total funding is now $173 million after a June 2025 stealth exit with $48 million and a December 2025 Amex Ventures strategic investment.

CEO and co-founder Anish Agarwal frames the problem as applied AI under hard constraints: every company’s production graph is different, and downtime costs remain extreme. The announcement cites a 2026 Splunk and Oxford Economics finding that downtime costs rose 50% in two years and that engineers lose seven hours a week to troubleshooting, alongside Gartner’s projection that 85% of enterprises will adopt AI SRE tooling by 2029, up from under 5% in 2025. Traversal’s pitch is that coding agents accelerated the inner loop of software creation; the outer loop—keeping systems up—still burns human nights reconstructing what already happened.

Why frontier chat agents are not enough on-call

Traversal argues that general frontier agents are not ready to own production incidents alone. On ORCA-bench, a production-fidelity benchmark the company built with Cornell Tech and Columbia University, the best of five frontier agents reached 25.3% accuracy on medium-difficulty incidents and 10% when incident reports were vague. The suite includes 1,079 root-cause tasks; hallucination rates ran as high as 40% in the company’s telling. Those numbers are Traversal’s own benchmark results and should be read as self-reported motivation for a specialized stack, not as an external audit.

The product story centers on a Production World Model and Causal Search Engine that keep an updated picture of services, dependencies, and incident causes. Founders come from causal machine learning research: Agarwal and Raj Agrawal hold MIT PhDs; Raaz Dwivedi earned his at UC Berkeley and, with Agarwal, has maintained faculty appointments at Cornell Tech and Columbia while building the company; Ahmed Lone joined from Citadel Securities. The research framing is deliberate—Dwivedi describes “superintelligent AI SRE” as making the outer SDLC loop resilient so autonomous coding’s inner loop does not outrun operations.

Traction claims customers will recognize

Traversal says ARR grew 9× year-to-date, product usage nearly 100×, and monthly active users nearly 20× over the same period. Customers ran more than 150,000 investigations last quarter; the product supported over 4,000 engineers across more than 1,000 major incidents, 50,000 minor incidents, and 5,000 significant production changes. Named references in the announcement include Nubank (founder and CEO David Vélez on early on-call rollout), Harvey (CTO Siva Gurumurthy on AI-native lifecycle extension into production), American Express (CTO Matthew Liste on production insight), plus PepsiCo and DigitalOcean. Investor quotes from Khosla, Kleiner Perkins, and Redpoint emphasize MTTR compression and the difficulty of building equivalent internal AI SRE.

What the money is for

Traversal says the round funds extending AI across the software lifecycle by making production more resilient, and that it is hiring across research, engineering, and go-to-market. The company does not disclose revenue figures beyond the ARR growth multiple, does not publish ORCA-bench as an open leaderboard in the announcement, and does not detail model vendors inside the agent stack. The same announcement concedes that frontier generalist models still fail most medium-difficulty ORCA tasks, the reliability gap Traversal says its agents are built to close. The new money takes the company to $173 million raised since it came out of stealth in June 2025, and the hiring plan spans research as well as sales.