Spotify more than doubled its monthly merged changes, from about 8,100 to 17,000 year over year in August, as AI-assisted development spread across the company. Executives described the shift alongside the control changes needed to absorb it.
The scale involved is large. Spotify runs close to 3,000 production services for 777 million monthly active users, roughly 100 million concurrent clients, and 11 to 12 million backend requests per second. A higher volume of automated changes lands on all of it.
One incident came from the content pipeline, which ingests more than 500,000 new songs, videos, podcasts and audiobooks daily. On June 24 a scheduled batch workload collided with new episodes while a scheduling bug cut throughput about 10 percent, delaying releases that normally appear within minutes by several hours. Spotify added end-to-end monitoring, reprioritized batch jobs and increased capacity.
The company’s fleet automation has merged more than 2.5 million maintenance pull requests, most after automated safety checks. It now runs Honk, a background coding agent executing Claude through the Agent SDK in Kubernetes pods, with a separate system choosing targets and tracking progress. A recent Java migration across backend services took three days.
That scale also exposed gaps. An automated dependency upgrade passed every existing check and still failed in production. Spotify is expanding rollback capacity, tightening safeguards, and restricting automated changes to the working hours of the teams that own the affected services.
