Score and route leads with a lead targeting agent

Step 1: Learn what actually predicts conversion from your own dataExcel

Export historical leads with their known outcome (converted or not) into a Table, then use PivotTables to check conversion rate broken out by company size, engagement level, source channel, and response speed.

Find which of these actually correlates with conversion in your own data before assuming a generic lead-scoring signal (like "company size") applies to your specific business.

Step 2: Build the scoring agentCopilot Studio

In Copilot Studio, have the agent score each new inbound lead against those validated signals and return a score with the specific factors driving it, not just a number.

Prompt idea:

Score this inbound lead's conversion likelihood based on company size, source channel, and engagement signals from the past week. State the score and the top 2 factors driving it, using our validated scoring model.

Step 3: Route by score, with a human checkpoint on the top tierExcel

Auto-route high-scoring leads to the appropriate rep queue quickly, since speed matters for hot leads, but have a sales ops person spot-check top-tier routing weekly to catch scoring drift.

Step 4: Feed actual outcomes back into the modelExcel

Track whether scored leads actually convert and periodically re-validate the scoring signals against fresh outcome data — a model built on last year's patterns can go stale.

Check the work

• Compare scored leads against actual conversion outcomes on a regular cadence to check the model's accuracy hasn't drifted.
• Confirm the stated "top factors" for a score are things actually present in that lead's data, not a generic explanation.
• Watch for the model overweighting one easy-to-measure signal (like source channel) at the expense of ones that actually matter more.

Source: Microsoft Copilot Scenario Library — Sales (2026)