Hi Community,
I’m currently designing a Virtual Agent topic where users can describe their issue in free text, and the bot will either create an Incident or a Service Request based on the description. Before creating any ticket, I’d also like the Virtual Agent to suggest relevant Knowledge articles for possible self-resolution.
I have a couple of questions and would appreciate any best practices or real project experience:
What is the recommended approach for presenting relevant Knowledge Base articles to users first, based on their issue description, before proceeding with Incident creation?
What is the best way to dynamically determine whether the user’s request should result in an Incident or a Service Request purely from their description?
Do you typically rely on LLM-based intent classification, NLU intents, keyword mapping, or a hybrid approach?
My goal is to keep the experience natural for users while maintaining accuracy and control on the backend process.
Thanks in advance for any insights or examples you can share!
