You are here:
Set Up the Find Expert Action for Agentforce
Enable and configure the Find Expert action to help support reps resolve complex cases.
Required Editions
- From the Agentforce Setup page, activate the Find Expert action.
- Make sure that the action is configured to analyze the case subject, description, reason, product, and language.
- Enable Data Cloud integration so the reasoning engine identifies similar historical cases.
- Install and Deploy the Service Cloud Data Kit.
- Create a search index (create a separate help doc for this). Search Index should be created on default data space.
-
Fine-tune how the reasoning engine prioritizes experts. The standard configuration weights
are:
- Expertise Relevance (60%): Based on keyword extraction and profile skills.
- Performance History (30%): Based on historical resolution speed and CSAT.
- Availability (10%): Based on current workload and online status in Salesforce or Slack.
-
Configure Slack and collaboration tools.
- Slack Integration: Enable one-click messaging so support reps can message experts directly with full case context.
- Team Configuration: Make sure that the system is permitted to add experts to the case team for ongoing consultation.
Understand Expert Ranking and Scoring
The Find Expert action uses an intelligent matching system to identify the best human resource for any given issue. The system learns expertise by monitoring outcomes rather than just static profiles.
- Resolution Patterns: Agentforce tracks how quickly and effectively a support rep resolves specific types of issues.
- CSAT Integration: High customer satisfaction scores for specific case types increase a support rep's Expertise Score for that category.
- Outcome Comparison: If Support Rep A resolves an issue in 2 days with a 4.8 CSAT, they are ranked higher for that topic than Support Rep B, who takes 5 days with a 3.2 CSAT.
Beyond performance, the reasoning engine analyzes:
- Assigned Skills: Profile skills and endorsements.
- Historical Ownership: Users previously assigned to similar cases or added to relevant case teams.
Did this article solve your issue?
Let us know so we can improve!

