A consultant is enhancing the Sales Qualification Agent for a SaaS company. The company keeps a set of disqualification criteria (e.g., 'no budget confirmed within 90 days,' 'competitor already under contract') in a living SharePoint document that sales ops updates weekly. The consultant wants the agent to weigh new leads against these criteria without manually re-encoding rules every time the document changes. What should the consultant do?

The correct approach is to add the SharePoint document as a knowledge source for the Sales Qualification Agent in Copilot Studio. Copilot Studio agents such as the Sales Qualification Agent support knowledge sources, and pointing the agent at the live SharePoint document lets it reason against the current version of the disqualification criteria every time the document changes, with no re-encoding step. The comma-separated export imported as a static Dataverse reference table that the agent's scoring logic reads once during setup is incorrect because a one-time import creates a static snapshot that goes stale as soon as sales ops edits the SharePoint document again, defeating the goal of keeping the agent current. The weekly email and manual system-prompt update approach is incorrect because manually rewriting the system prompt every week is exactly the recurring maintenance burden the consultant is trying to eliminate, and it introduces delay and transcription-error risk. The Power Automate flow that copies the SharePoint document's text into the lead record's description field for reps to read manually is incorrect because it only gives reps something to read themselves; it does nothing to inform the agent's own scoring logic, which is the actual requirement. Wiring a knowledge source is also more resilient to formatting changes in the source document than any of the manual or one-time alternatives, since the agent retrieves from the live source at query time rather than from a cached copy.