Traversal Raises $120M at $1B to Automate Incident Response

By Grace Kim

Traversal announced a $120 million Series B at a $1 billion valuation on November 3, 2026, lifting the AI startup’s disclosed funding to $173 million.

The Series B was co-led by Kleiner Perkins, Redpoint Ventures and Khosla Ventures, with significant participation from Sequoia Capital and Battery Ventures, according to Traversal’s announcement. Pulse2 and The SaaS News also listed Amex Ventures as a participant.

Traversal said Sequoia and Kleiner Perkins, which led the company’s $48 million Series A in June 2025, re-upped in the round. Its blog said both had “doubled down on earlier investments.”

What bottleneck is Traversal targeting?

Traversal cited a survey in which 42% of developers said they now use AI to write most of their code, up from 12% a year earlier. Another 52% said they spend more time debugging after that code reaches production.

The company also said unplanned outages cost Global 2000 companies a combined $600 billion in 2026, 50% more than two years earlier.

“Enterprises bought coding agents expecting faster delivery and leaner engineering costs,” CEO Anish Agarwal wrote in Traversal’s announcement. “But more code has meant more complexity to manage, and more opportunities for software reliability issues and incidents.”

The bottleneck Traversal targets is keeping AI-written software running reliably in production.

How does Traversal test causes?

Traversal says its two core components emerged from nearly two years of research before their formal introduction in March 2026.

• Production World Model: A continuously updated, AI-readable representation of an organization’s software environment. It ingests logs, metrics, deployment histories, source code and operational knowledge.
• Causal Search Engine: A system that tests thousands of hypotheses in parallel to separate causation from correlation when an incident fires dozens of signals at once.

Traversal’s four founders are CEO Anish Agarwal, a Columbia professor; Raaz Dwivedi, a Cornell Tech professor; Raj Agrawal, an MIT PhD; and Ahmed Lone, a former Citadel Securities employee. The company says they bring backgrounds in causal machine learning and reinforcement learning from MIT, Berkeley, Columbia and Cornell.

On ORCA-bench, developed by Traversal with Cornell Tech and Columbia University, the best of five frontier agents recorded:

• 25.3% accuracy on medium-difficulty incidents.
• 10% accuracy when incident descriptions were vague.
• Hallucination rates reaching 40%.

Traversal uses the results to argue that conventional frontier models are insufficient for production incident response. It developed the benchmark, and neither its announcement nor Pulse2’s reporting cites an independent validation.

The technology aims to replace guesswork with tested causes, but the benchmark evidence still comes from Traversal.

How quickly are customers adopting Traversal?

Since the March 2026 launch, monthly investigations rose from 513 to nearly 73,000 during 2026, according to Traversal and Pulse2, a more than 100-fold increase.

Pulse2 reported that annual recurring revenue had increased 9x year to date. A Forbes profile cited by American Bazaar described 9x growth over the preceding six months.

Pulse2 also reported the following activity in Traversal’s most recent quarter:

• More than 150,000 investigations
• Activity across 4,000+ engineers
• More than 1,000 major incidents
• 50,000 minor incidents
• 5,000 significant production changes

Customer lists differ across reports. Pulse2 cited Nubank, Harvey, PepsiCo and DigitalOcean. A Forbes profile cited by American Bazaar named American Express, Pepsi and Blackstone.

Nubank founder David Vélez said five weeks into rollout, Traversal was “already becoming part of how our on-call engineers work incidents.”

Roughly two-thirds of platform usage now extends beyond root-cause analysis, according to Traversal, into:

• Alert triage
• Release-risk assessment
• CMDB validation
• Security investigations
• Customer-experience monitoring
• Production support

The usage data show expansion into adjacent workflows, not the size of the revenue base.

What do the new products do?

Backed by the Series B, Traversal is extending beyond reactive investigation with two products:

• Traversal Workers: Long-running AI agents that start gathering information as soon as an anomaly triggers. They identify likely root causes and help coordinate remediation before an engineer opens an investigation.
• Traversal MCP: A feed of production intelligence for development tools, allowing coding agents to access operational context when writing or updating code.

The long-term goal is “self-driving production”: systems that autonomously detect issues, understand causes, take corrective action, verify recovery and prevent recurrence, closing the outer loop of an AI-native software development lifecycle.

The new tools shift Traversal from diagnosing failures toward directing production systems.

What remains unproven?

• Revenue: Traversal has disclosed only ARR growth multiples, not absolute ARR.
• Validation: ORCA-bench and the causal-reasoning claims come from Traversal; neither the company announcement nor Pulse2’s reporting cites a third-party evaluation.
• Competition: The SaaS News describes established observability and AIOps platforms alongside newer entrants. Traversal identifies causal AI as its differentiator, while Pulse2 and The SaaS News provide no head-to-head comparison.
• Hiring: Traversal says it will expand research, engineering and go-to-market teams without disclosing targets or timelines.

The $1 billion valuation on $173 million in disclosed funding gives Traversal unicorn status before it has published absolute ARR or independent benchmark validation. Investors are pricing reliability tooling as a potential category-defining market for agentic AI. The next test is whether ninefold ARR growth and nearly 73,000 monthly investigations translate into durable recurring revenue while the causal engine earns independent support for self-driving production.