Expedia Group has introduced STAR, an internal AI-assisted observability system designed to help engineers investigate production incidents faster. InfoQ reports that the platform uses service telemetry and large language models to analyze incidents, suggest root causes, and support response workflows.
STAR is built with FastAPI, Datadog, Celery, Redis, and Langfuse. Rather than asking a model to improvise freely, Expedia uses structured workflows that collect telemetry, guide analysis, and produce assessments that engineers can review. That design keeps the system closer to an investigation assistant than an autonomous incident commander.
The update reflects a broader shift in enterprise AI adoption. Companies are moving from chat interfaces toward systems that sit inside operational workflows, where outputs need traceability and where wrong answers can delay recovery or send teams down the wrong path.
The useful limit is also clear: STAR helps accelerate investigation, but the final judgment remains with engineers. In production reliability work, AI speed only matters if teams can verify why the system reached its conclusion.