You need to configure the generative AI reusability requirements.

Calling the prompt from a Power Automate cloud flow provides a reusable service-level entry point that can be invoked by different Power Platform experiences. The flow can accept parameters from each caller, pass those values into the prompt, and then route the generated result to an app, Dataverse, email, or another process. This addresses reuse more effectively than binding the prompt directly to one canvas app. An input variable is necessary for dynamic data, but by itself it does not establish cross-service orchestration; the flow exposes and manages those inputs for multiple callers. Returning JSON can standardize output handling but does not make the prompt reusable by itself. Direct canvas-app invocation is valid for an app-only requirement, not for the stated need to span Power Platform services. The flow should define explicit input types and validate required values so the case’s inconsistent-input problem is not repeated. The AB-410 Study Guide lists both “add inputs to a prompt” and “consume a prompt in cloud flows.” Option B provides the reusable execution layer while allowing dynamic values to be supplied at runtime.

Study Guide reference/topic: “Create prompts and models in AI Hub — Consume a prompt in cloud flows.” AB-410 Study Guide | Microsoft Learn technical reference