HYPE OR FACT? · CHECKED 4 NOVEMBER 2026
Traversal raises $120M for its 'AI SRE': can AI really find why software broke?
OUR VERDICT
Partly factOur original verdict: “Fact or hype: partly fact. The $120M round and the ORCA-bench benchmark are real and public, and the benchmark itself shows general agents usually miss the true cause. The accuracy and time-saving figures for Traversal's own product are the company's and its customers' claims, not independent results.”
THE CLAIM
“Traversal, valued at $1B after a $120M Series B, says its AI SRE keeps code running by finding the root cause of outages; its new benchmark shows the best AI agents find the real cause only one time in four.”
What’s actually true
- Traversal announced a $120M Series B at a $1B valuation on 3 Nov, co-led by Kleiner Perkins, Redpoint Ventures and Khosla Ventures (Business Wire).
- Its ORCA-bench paper, with Cornell Tech and Columbia, has 1,079 on-call tasks built on real-looking telemetry and code. The best general agent found the true root cause 25.3% of the time on medium tasks and about 10% on hard tasks with vague reports, and agents often named a cause that wasn't there.
- Traversal says its own specialised agents do much better, and its website quotes customers with high root-cause accuracy and large time savings. Those are company and customer figures, not independent tests.
- It's an enterprise product with no published pricing. We haven't tested it.
Sources
Cost
Enterprise pricing only; not published.
USE IT IF
Larger engineering teams with frequent production incidents and good logs, metrics and traces.
SKIP IT IF
Small teams hoping a general chatbot can run incident response; the benchmark shows it usually guesses wrong.
How we knowChecked on 4 November 2026. Verdicts: Fact = the claim holds up; Partly fact = real, but with caveats or overstated parts; Hype = the headline claim isn't supported. Products change quickly, so check the maker's site before you buy. Our methods & sources ↗