The cost of downtime is already vast and poised to explode as AI generated code leads to systems that are less well understood when things go wrong. Financial losses from software outages grew to more than $400 billion in 2024, and the productivity burden is even greater, with the world’s 40 million developers estimated to spend 30% of their time reacting to bugs and errors.

Phoebe was founded by Matt Henderson and James Summerfield, formerly CEO and CIO of Stripe Europe, who sold their first startup, Rangespan, to Google in 2014. Henderson explains that their motivation for building Phoebe was experiencing the frustrations of software failures first-hand in past roles: “High-severity incidents can make or break big customer relationships, and numerous smaller problems drain engineering productivity. Software monitoring tools exist, but they aren’t very intelligent and require people to spend a lot of time working out what’s wrong and what to do about it.”

Phoebe’s breakthrough comes from ‘swarms’ of AI agents that search for evidence among vast siloed data to evaluate many different potential causes and solutions for a problem. “With Phoebe supporting the diagnosis, we’ve found that time-to-resolution can be up to 90% faster” says Henderson.

What’s even better than a faster resolution is preventing the incident before it occurs. Phoebe can already predict some incidents from leading indicators, then generate preemptive fixes before disruption occurs. The company’s vision is for these agents to serve as an “immune system for software”, eventually making it possible for the overwhelming majority of emergent problems to be prevented before becoming customer-impacting failures.

Roni Hiranand at GV said: “AI has transformed how code is written, but software reliability has not kept pace. Phoebe is building a missing layer of contextual intelligence that can help both human and AI engineers avoid software failures. We love the boldness of the team’s vision for a software immune system that preemptively fixes problems.”

Jay Davies, Head of Engineering for Reliability and Operations at Trainline, said: “Phoebe has already had a real impact on how we investigate and remediate incidents at Trainline. Work that used to take us hours to piece together can now take minutes and that matters when you’re running critical services at our scale. We’re excited to see Phoebe progress and help us operate efficiently.”

For more startup news, check out the other articles on the website, and subscribe to the magazine for free. Listen to The Cereal Entrepreneur podcast for more interviews with entrepreneurs and big-hitters in the startup ecosystem.