Make component design decisions

  • 5 minutes

Design decisions rarely arrive one at a time. In this five-minute scenario, decide which component or tool fits each requirement before you read the answers.

Scenario: Extend the equipment-provisioning solution

The equipment-provisioning team brings you four new requirements:

  1. Reject any request that exceeds an employee's allowance, no matter which app, flow, or integration creates it, and reuse the same eligibility check from a manager's pre-check screen.
  2. Let coordinators ask an agent open-ended questions about request data, such as "how many laptop requests are pending this month?", without building a connector action for every possible question.
  3. Reserve a device at the depot system whenever a coordinator approves a request in the app, with no conversation involved.
  4. Keep a searchable history of eligibility-check failures in production for longer than a day, so support can investigate a failure reported last week.

Check your decisions:

  1. A custom API backed by a plug-in, with a synchronous check on the Request table. Server-side logic runs no matter which surface creates the record, and a custom API gives the same eligibility logic a reusable contract that the pre-check screen can call. A form script or a flow alone would miss requests created elsewhere.
  2. The Dataverse MCP server. It lets an agent query Dataverse data conversationally instead of needing one connector action per question. Weigh its cost: calls from agents built outside Copilot Studio are billed.
  3. A Power Automate cloud flow. The trigger is a data event, not a conversational turn, so a cloud flow fits. A Copilot Studio workflow (agent flow) fits when a conversation invokes the multistep action.
  4. Application Insights. Plug-in trace logs hold limited text and age out after 24 hours, which suits active debugging, not investigating last week's failure.

If any answer surprised you, reread the unit it comes from before the knowledge check.