AWS Bedrock Inference Profiles Give Organizations Detailed AI Usage and Cost Visibility

Organizations using AWS Bedrock to power agentic applications have historically lacked granular visibility into which teams or apps are consuming AI tokens and at what cost. AWS Bedrock inference profiles address this by assigning each application a named model profile tagged with app and team identifiers, which are then stamped onto every API log entry. Bedrock model invocation logging routes data to both Amazon S3 and CloudWatch, enabling durable storage and real-time alerting without requiring code changes in the calling applications. Once logs are flowing with identity tags, Amazon Athena can query the S3 bucket to break down token usage and estimated costs by application, team, or individual IAM caller. This setup allows engineering and finance teams to identify cost anomalies — such as a runaway loop or unnecessary use of a premium model — at a per-application level.