Salesforce Decode
Salesforcedecode
Back to questions
AgentforceIntermediatetopicstaxonomynlpdiscovery

Design topic taxonomy and utterance coverage for a new Agentforce deployment

Real World Scenario

You are launching Agentforce for an insurance carrier with 200+ common call reasons. The project team wants to know how many topics to create and how to organize them.

Expected Answer

• Analyze historical case and chat transcripts to cluster intents—aim for 15–40 topics at launch, not 200 • Group topics by customer journey stage and back-office system boundary • Define clear topic descriptions and example utterances—overlap causes misclassification • Map each topic to allowed actions, knowledge collections, and escalation rules • Plan a "general inquiry" fallback topic with safe responses and easy human escalation • Establish a monthly topic gap review using unrecognized utterance logs • Version topics with change logs; avoid renaming slugs that break analytics continuity

Follow-Up Questions & Answers

Click to expand — each follow-up includes a direct, interview-ready answer

Direct answer: Analyze historical case and chat transcripts to cluster intents—aim for 15–40 topics at launch, not 200 Also consider: Group topics by customer journey stage and back-office system boundary In practice: Define clear topic descriptions and example utterances—overlap causes misclassification Balance speed of delivery with maintainability.

Architect Perspective

Topic sprawl degrades classification accuracy. Start with journey-based groupings validated against real transcripts. A smaller, well-maintained taxonomy outperforms exhaustive intent lists that nobody updates.