J
member badge
Joanne Ang
Business Problem
When users ask Copilot to complete a task, they may not know which decisions or constraints they need to specify. Instead of surfacing these important gaps, Copilot may fill them with assumptions.
This can produce an outcome that appears complete but does not match what the user actually needs. For example, when asked to create a Support AI Agent, Copilot may assume what the Agent should handle or when it should hand off. The user may only discover these mismatches when reviewing or testing the proposed setup.
Desired Outcome
Help Copilot turn an incomplete request into an agreed plan before proceeding. Copilot should:
  • Ask clarifying questions with varying depth:
    Surface important gaps based on the request’s ambiguity, effort and potential impact.
  • Provide a recommendation:
    Recommend an approach based on the user’s answers and explain important decisions or trade-offs.
  • Provide a plan before execution:
    Summarize the intended outcome and proposed changes for the user to review and refine.
Copilot should provide useful direction quickly. Straightforward questions and bounded changes should proceed without unnecessary planning.
Use Cases
  1. A user asks Copilot to create a Support AI Agent. Copilot clarifies its responsibilities, Knowledge Sources and handoff rules before recommending and presenting the setup plan.
  2. A user asks how to create a Workflow for routing high-value leads. Copilot clarifies the qualification criteria, routing destination and exceptions before recommending a Workflow structure and explaining how to configure it.
Current Workaround
Users must anticipate which details matter and provide them upfront. Otherwise, they must identify unsuitable assumptions during review or testing and correct them through repeated messages.