What changed
Circle K deployed an automated incident triage system using Amazon Bedrock, reducing mean triage time to under 15 minutes. The system uses an agentic AI architecture to investigate incidents autonomously and deliver structured summaries to Slack within minutes of an alert firing.
What this means in practice
The system, built using Amazon Bedrock, helps on-call engineers by providing actionable context instead of raw alerts. It automates repetitive context gathering and provides structured findings, allowing engineers to act immediately rather than spending an hour figuring out what happened. The solution uses Amazon Bedrock Flows as the orchestration layer, coordinating specialized agents that handle different parts of the analysis. The pipeline starts with a Grafana alert published to Amazon SNS, which forwards it to an Amazon SQS queue. An AWS Lambda function polls the queue, extracts the alert payload, and invokes the Flow. Inside the Flow, two agents are invoked in parallel: the AWSResourceExplorerAgent and the OpenSearchExplorerAgent. These agents query infrastructure and application-level logs, respectively. The GitHubExplorerAgent then runs, searching for affected service repositories and proposing fixes. All outputs feed into an AnalysisPrompt node, which produces a structured summary for on-call engineers.
Limitations and availability
The system uses Amazon Bedrock, a fully managed service that makes foundation models available through a single API. For model availability by AWS Region, refer to Supported models by AWS Region in Amazon Bedrock. The solution is available to AWS customers and requires integration with existing AWS services such as Amazon SNS, Amazon SQS, and AWS Lambda. The system is designed for cloud-native environments and requires access to AWS Identity and Access Management (IAM) policies, Amazon Virtual Private Cloud (Amazon VPC) configurations, and logging pipelines. The team noted that data privacy was a key consideration, as Amazon Bedrock does not share customer data with model providers or use it to train foundation models.
