Hi Robert,
I ran additional tests and confirmed that both the Interaction transcript and Copilot transcript show the same behavior, so this appears to be a transcription-layer issue rather than Copilot summarisation.
"Okta" is sometimes correctly recognised, but we are also seeing variations such as "octa," "opta," and even "october very fine," which suggests the engine is applying phonetic matching with varying confidence.
Since dictionary management biases recognition rather than forcing exact replacement, this may be a model confidence limitation rather than a configuration issue. We are using en-AU native transcription.
Has anyone had success improving recognition accuracy for short brand names like this using phrase hints or multi-word dictionary entries?
------------------------------
Phaneendra
Technical Solutions Consultant
Monash University
Australia
------------------------------
Original Message:
Sent: 02-27-2026 03:40
From: Robert Niblock
Subject: Transcription accuracy issue with Copilot (Dictionary term not resolving)
Hey Phaneendra,
I saw in the Dictionary Management article, they refer to this "Dictionary management is only supported for Genesys Cloud native transcription dialects." Understand dictionary management View summary - Genesys Cloud Resource Center
I'm wondering that it might not be resolving because of the transcription setting for Copilot?
I guess the easiest way to test this would be have a user without Copilot say Okta a bunch in an interaction and then check the transcription and see if it uses Okta instead of Octa.
------------------------------
Robert Niblock
Contact Centre Technology Analyst
Original Message:
Sent: 02-26-2026 21:03
From: Phaneendra Avatapalli
Subject: Transcription accuracy issue with Copilot (Dictionary term not resolving)
Hi everyone,
We are currently using Agent Copilot and have identified an issue with transcription accuracy.
For example, the word "Okta" is being transcribed as "Octa" in the interaction transcript, and this then carries through to the Copilot-generated summary.
We have already added "Okta" in Dictionary Management, however it is still not being picked up correctly in the transcript or summary.
Has anyone experienced a similar issue?
Is there a way to improve transcription accuracy for specific terms?
Appreciate any guidance.
Thanks!
#AICopilot(Agent,SupervisorAdmin)
------------------------------
Phaneendra
Technical Solutions Consultant
Monash University
Australia
------------------------------