Spotify has expanded its use of AI-assisted development, more than doubling monthly merged changes from approximately 8,100 to 17,000 year over year in August. The company is adapting its quality and reliability controls to handle the higher volume of software changes across its production systems.
Spotify operates nearly 3,000 production services supporting 777 million monthly active users, around 100 million concurrent clients, and 11 to 12 million backend requests per second. Its review identified issues involving content processing, fleet changes, compute capacity, and mobile application quality. The company has responded with changes to monitoring, testing, rollout safeguards, rollback capacity, capacity planning, and quality measurement.
One incident involved Spotify's content processing pipeline, which handles more than 500,000 new songs, videos, podcasts, and audiobooks each day. On June 24, a scheduled batch workload competed with new episodes, a quality improvement increased processing requirements, and a scheduling bug reduced throughput by about 10%. Episodes that normally appeared within minutes were delayed for hours. Spotify subsequently added end-to-end monitoring, fixed the scheduler, moved batch jobs to lower priority, reworked workload prioritization, and increased capacity.
Fleet automation created a different challenge. Spotify's Fleet Management system has merged more than 2.5 million automated maintenance pull requests, with most automatically merged after safety checks. The company later added Honk, a background coding agent that runs Claude using the Agent SDK in Kubernetes pods. Fleetshift handles target selection, scheduling, and progress tracking, while Honk performs code modifications and can run builds in CI. Spotify reports that a recent Java migration across its backend services took three days.
An automated dependency upgrade nevertheless passed existing checks and failed in production, according to Spotify. The company is strengthening safeguards, expanding rollback capacity, and scheduling automated changes during owning teams' working hours.
Amit Jha, a software engineer at JumpCloud, in a LinkedIn post discussing Spotify's findings, said: More changes meant greater pressure on testing, review and observability.
Alex Beck similarly focused on the verification challenge in a LinkedIn post, Writing code got easier. The hard part now is making sure the work is actually good before it goes live.
The findings provide context for Google's 2025 DORA research, which found that AI adoption was associated with higher software delivery throughput and product performance, while also reporting a negative relationship with delivery stability. The research identified automated testing, version control, and fast feedback loops as practices relevant to managing software delivery as AI adoption increases.
Spotify combines Fleet Management for standardized maintenance with Honk for complex code changes, while Fleetshift handles orchestration and CI provides validation. Quality and optimization work rose from 27% to 31% of merged changes, while maintenance and configuration fell from 31% to 25%. Spotify reports no corresponding increase in rework, although pull request size and code complexity are rising. Its incident review found no material direct contribution from AI-authored code, but change volume had outpaced some verification controls.