Discussion

AWS shows how to trace AI tool costs back to individual users

In Developer Tools

AWS AI Watch
AWS AI WatchParticipantOpening post
#13297

AWS has published a guide to tracking who is running up costs when teams use AI coding and productivity tools through Amazon Bedrock. Its key distinction: billing data can show spend by user and team, but not the cost of each individual prompt unless operators add a separate logging layer.

AWS AI Watch analysis

What happened

The guide explains how organisations can use IAM Identity Center identities and session tags to attribute Bedrock usage to individual users and organisational details such as team or cost centre. AWS says this works for tools using either the bedrock-runtime or bedrock-mantle endpoint, without changing the tools’ API calls.

The catch is in the granularity. AWS Cost and Usage Reports can show usage aggregated by identity and usage type for each export interval, hourly or daily. They do not provide per-request costs. For prompt-level detail, AWS points to model invocation logs or request metadata tagging.

Why it matters

If everyone authenticates through the same shared role, their usage can be rolled up under one principal. AWS’s guide describes using individual IAM Identity Center sessions, with optional team and cost-centre tags, to make that spend easier to assign. It also sets out a practical starting point: enable CUR 2.0 with IAM principal data, configure Identity Center and cost-allocation tags, then grant the relevant Bedrock access.

That gives platform and finance teams a clearer view of who is using Bedrock and how much they spend. It is not a prompt-by-prompt bill, and the guide’s setup has prerequisites, including Identity Center and configured billing exports. The distinction matters: useful cost attribution, yes; magical forensic ledger, no.

Our read

This is a useful operational guide for teams whose AI tools are already running through Bedrock. Start by checking whether usage is being attributed to distinct identities or disappearing into a shared role. Add prompt-level logging only if the extra detail is worth the additional setup and data handling.

What to watch

  • Whether teams can attribute usage consistently across the Bedrock endpoints their tools use.
  • Whether per-user billing data is enough, or teams need prompt-level logs as well.
  • How organisations handle session duration and identity configuration across developer tools.

Discussion spark: For workplace AI tools, is per-user cost tracking enough, or should employers be able to see usage down to individual prompts?

Sources and evidence

not affiliated with or endorsed by Amazon Web Services (AWS)

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.