Glad that helped!
We saw the same behaviour - a single topic can work well for broad detection (e.g. "Authentication Issue" with MFA/password phrases), but it does become quite broad when trying to identify specific scenarios.
What worked for us was keeping a broader topic and then using workflow logic to differentiate scenarios, rather than creating too many separate topics.
This gave us a bit more flexibility without having to maintain multiple overlapping topics.
------------------------------
Phaneendra
Technical Solutions Consultant
------------------------------
Original Message:
Sent: 07-08-2026 09:23
From: Steven Casper
Subject: Creating a phrase in within topics based on caller (ANI) not the agent (or vise versa)
Thank you, I am testing now. I do like the ideal of within the phrase you can make these same adjustments. It seems like an all or nothing option. (having to have multiple topics to be more specific). Once again, thank you for providing the community support, I was hoping for.
------------------------------
Steven Casper
W.W. Grainger, Inc.
steven.casper@grainger.com
------------------------------
Original Message:
Sent: 07-08-2026 08:46
From: Phaneendra Avatapalli
Subject: Creating a phrase in within topics based on caller (ANI) not the agent (or vise versa)
Hi Steven,
We've worked on a similar use case using STA topics.
You can achieve this by leveraging the participant field we restrict detection to external (customer) participants, which helps avoid agent-driven matches or repeated prompts.
Topics are also semantic rather than exact keyword-based, so having good phrase variations is important. From what we've seen, regex isn't supported in the phrase widget.
We started with a small set of high-confidence customer phrases and refined them using real transcripts. We also combined topics with workflow logic to detect specific scenarios rather than relying purely on phrases.
------------------------------
Phaneendra
Technical Solutions Consultant
------------------------------