Hello Dynatrace Community,
We are currently evaluating the governance and cost-control options available for Agentic AI under a Dynatrace Platform Subscription.
Dynatrace Support informed us that it is currently not possible to configure a hard consumption or cost limit that automatically stops or restricts Agentic AI executions once a predefined threshold has been reached.
Our understanding is that:
- Budget Alerts and Cost Monitors provide notifications only.
- Exceeding a configured budget does not stop Agentic AI usage.
- Hard limits cannot currently be applied at account, environment, user-group, or individual-user level.
- Access can be restricted through the ai:operator:execute IAM permission, but this controls whether a user can access Agentic AI and does not impose a usage or cost quota.
Our customer operates with a predefined annual Dynatrace budget and needs to prevent Agentic AI consumption from causing the annual DPS commitment to be exceeded.
We would therefore appreciate feedback from other customers and Dynatrace experts on the following:
- Has anyone implemented a practical workaround to stop or restrict Agentic AI usage when a specific consumption threshold is reached?
- Is there currently a way to identify the Agentic AI consumption generated by each individual user or user group?
- Can Agentic AI executions be distinguished from manual actions performed through Notebooks, Dashboards, DQL, or other Dynatrace applications?
- Are there audit records, APIs, billing exports, Account Management reports, or DQL-accessible data that show:
- the user who initiated the execution,
- the operation that was executed,
- the Dynatrace capability that was consumed,
- and the resulting DPS consumption or cost?
- Can a Budget Alert or Cost Monitor notification be connected to a Workflow that automatically removes the ai:operator:execute permission from a user or group? If technically possible, would this be considered a supported or recommended approach?
- How are other enterprise customers governing Agentic AI access and consumption to avoid unexpected on-demand charges?
We are particularly interested in real-world implementations, dashboards, APIs, automation approaches, or governance practices currently being used.
Thank you in advance
