Question

A financial services company runs a customer-facing chatbot on Amazon Bedrock that invokes Anthropic Claude models. Finance needs to attribute Bedrock inference spend to individual business units (retail banking, wealth management, and insurance), each of which uses the same shared AWS account. The team must produce per-business-unit cost reports without deploying separate accounts or provisioned throughput. What is the MOST appropriate approach?

Accepted Answer

Create an application inference profile for each business unit with distinct cost allocation tags, route each unit's invocations through its profile, and analyze spend using Cost Explorer and Cost Allocation Tags.

Answer

Enable model invocation logging to CloudWatch and parse the log volume per business unit to estimate proportional cost.

Answer

Purchase separate Provisioned Throughput commitments for each business unit and use the commitment IDs to split the bill.

Answer

Configure a Bedrock Guardrail per business unit and use guardrail metrics to allocate inference costs.