Spotify says AI helped lift monthly code changes to 17,000

The company is tightening safeguards and rollback plans as faster change rates expose more production risk.

Spotify says AI-assisted development helped push its monthly merged code changes from about 8,100 to 17,000 year over year in August, and it is now updating controls to cope with the higher change rate. The company’s review tied problems to content processing, fleet changes, compute capacity, and mobile app quality, but it did not find any material direct contribution from AI-written code. One June 24 incident delayed podcast and other content processing for hours after a batch workload, a quality-related increase in processing demand, and a scheduling bug combined to cut throughput by about 10%. Spotify also reported more than 2.5 million automated maintenance pull requests in Fleet Management, said it later added Honk as a coding agent for tougher backend changes, and described a Java migration across backend services that took three days. Separately, an automated dependency upgrade cleared existing checks but failed in production, so Spotify is strengthening safeguards, adding rollback capacity, and scheduling more automated changes during the owning teams’ working hours.