Power Platform Community Forum Thread Details

I've noticed that it behaves like a prompt, but sometimes it doesn’t trigger as expected. I changed my bot’s Generative AI behavior to Classical, added all the necessary trigger keywords, and then switched it back to Generative AI.

However, the issue is that if I use related phrases like "I want to talk to a supervisor" instead of exact keywords like "agent" or "live support", the bot doesn’t always trigger correctly.

Can anyone explain how I can achieve a setup where the bot understands related phrases and triggers the best-suited topic without needing exact keywords?

Categories:

General topics

Hi sohom-rik3,

Ideally in Generative AI mode, we should not be facing this if we have provided a clear description of the topic in the Trigger.

In Generative AI mode, bot relies more on Intent detection via semantic understanding. So, when we say “I want to talk to a supervisor,” it might not trigger the right topic (like "escalation to live agent") unless the bot understands the intent based on trained context.

Topic titles and descriptions are now also used by the LLM to figure out what the topic is about. So rename "Topic 1 – Agent escalation" as "Talk to a human agent or supervisor"

You can try to follow some best practices when describing,

1. Use Plain Language : Avoid technical jargon unless necessary. Describe it the way a user would think about the task. Example: "This topic helps users check the status of their order using their email or order ID."

2. Start with Purpose : Begin your description with what the topic is meant to solve or handle. “This topic handles customer requests for rescheduling appointments.”

3. Include Trigger Intent : Mention the common questions or phrases that lead to this topic. “Triggered when users ask to talk to a human agent or mention 'live support', 'supervisor', or 'representative'.”

Thanks.

Visit my Blog