AI Use Case Prioritization Agent

A Copilot Studio agent that helps makers and reviewers capture, assess, prioritize and route AI agent use cases using approved guidance and rubrics. Assessments are recorded in Dataverse and returned as a downloadable HTML report.

Built by Microsoft (Customer Acceleration Team) and approved through the Microsoft IT (MSIT) security and Responsible AI review.

What it does

• Guides intake of a candidate AI use case: business objective, users, data, outcomes and constraints.
• Grounds scoring in approved rubrics: impact, delivery complexity, risk and agentic fit.
• Resolves category and subcategory from approved guidance, and asks rather than guesses when unsure.
• Records the use case and assessment in Dataverse only after explicit user confirmation.
• Produces a downloadable HTML assessment report in the conversation.

Contents of this solution: 1 agent, 8 skills, 4 knowledge sources, a Dataverse MCP tool and 2 Dataverse tables (cadpbm_ai_usecase, cadpbm_aiassessment). This gallery download is the no-app, unmanaged edition.

Choose your version

The same solution ships in four versions, all version 1.0.0.2. The gallery download is version 1; the others are in the GitHub repository and its releases.

The app-included editions add a Power Apps code app. They need the Power Apps Code Apps environment feature enabled and a Power Apps licence for every user who opens the app. See Dependencies.

The reviewer app (app-included editions)

The app turns recorded assessments into a browsable, filterable and comparable portfolio.

Use Case Catalog: browse and filter every recorded use case by impact, complexity, reach and agentic fit.	Compare: plot use cases on an Impact vs Complexity matrix to spot quick wins.
Use Case detail: problem, solution, benefits, risks, personas and data sources.	Assessment detail: scorecards, positioning matrix and per-factor breakdowns.
Analytics: agents per customer, category spectrum, top submitters, agent-type mix and industry coverage.

Prerequisites

• A Power Platform environment with Copilot Studio and Dataverse.
• Approved guidance and rubric documents as knowledge. They are included in the solution.
• The Dataverse MCP server enabled for the environment.

Full steps: Deployment and Dependencies.

After import: assign the security roles (Users: own records; Admin/Intake: all records), set content moderation to High, disable the code interpreter, and run the smoke tests.

Security and Responsible AI

This solution went through the Microsoft IT (MSIT) approval process. The documentation is published with the source:

Key controls:

• Human oversight: nothing is written to Dataverse without explicit confirmation; delete operations are disabled.
• Grounding: scoring uses approved rubrics. Missing evidence is marked NOT ENOUGH INFO and the agent asks a targeted question instead of guessing.
• Prompt-injection resistance: instructions inside uploaded documents are treated as content, not behaviour.
• Least privilege: Microsoft Entra ID group access and Dataverse roles; connections run under the invoker’s context.
• Content moderation set to High, code interpreter disabled.
• Data minimisation: the customer or context name is captured only to tell use cases apart.

Source, release notes and support

Source, documentation and all four packages: https://github.com/alicja-gilderdale/ai-use-case-prioritization-agent. Raise issues in that repository’s issue tracker.