Circle K Cuts DevOps Incident Triage to Under 15 Minutes with Amazon Bedrock

Retail chain Circle K automated its cloud incident triage process using Amazon Bedrock, decreasing average investigation time from 45–60 minutes down to under 15 minutes. According to internal incident-response metrics reported by Circle K, this represents a 67% to 75% reduction in response time across its infrastructure. The system replaced manual diagnostics by deploying agentic workflows that investigate alerts and deliver root-cause analysis directly to Slack. Initially built with Amazon Bedrock Flows, the triage pipeline orchestrated parallel agents to query AWS resources across accounts and Amazon OpenSearch Service logs, cross-referencing findings with service product cards stored in Amazon Bedrock Knowledge Bases. A subsequent GitHub agent inspected repository code and commits to identify root causes. An analysis prompt node then aggregated the findings into an actionable report outlining probable causes, evidence, and remediation steps. To accommodate growing log volumes and longer-running multi-service investigations, Circle K migrated the system in April 2026 to Amazon Bedrock AgentCore. The deployment consolidates specialized tasks into a single agent running on AgentCore Runtime, utilizing AgentCore Gateway to manage tool access across logging, code repositories, and knowledge bases while supporting session persistence and OpenTelemetry observability.

Key takeaways

• Circle K reduced DevOps mean triage time from 45–60 minutes to under 15 minutes, achieving a 67–75 percent reduction based on internal metrics.
• The triage workflow automatically gathers context across AWS resources, OpenSearch traces, and GitHub repositories before sending root-cause reports to Slack.
• In April 2026, the team migrated the pipeline from Bedrock Flows to Amazon Bedrock AgentCore Runtime and Gateway to support longer sessions and persistent context.