(Replace this first paragraph with a brief description of your new category. This guidance will appear in the category selection area, so try to keep it below 200 characters.)
Use the following paragraphs for a longer description, or to establish category guidelines or rules:
Why should people use this category? What is it for?
How exactly is this different than the other categories we already have?
What should topics in this category generally contain?
Do we need this category? Can we merge with another category, or subcategory?
Are you pointing the Agent Assist widget to the live transcription stream or the archived call recording? The knowledge surfacing engine only pulls suggestions when the real-time transcription payload matches the topic model tags. If the transcription is delayed or the language code mismatches, the ROI on those suggestions drops to zero because agents see nothing until the call ends. Check the webhook payload structure in the integration settings. Sometimes the payload gets truncated if the buffer is full, which kills the suggestion flow instantly.
Turn down the confidence threshold in the knowledge source config. The default 0.8, it’s too strict for noisy call centers. Set it to 0.6 and watch the suggestions populate. Make sure the ‘real-time transcription’ toggle isn’t disabled under the speech analytics profile, since some folks mix that up with the sentiment analysis flag. You might be blocking the text stream before it reaches the suggestion engine. Here’s the snippet for the threshold adjustment: {"knowledgeSourceId": "src_123", "confidenceThreshold": 0.6, "realTimeEnabled": true}. Don’t forget to re-index the knowledge base after changing the tags. The won’t pick up new keywords until the index refreshes. Also check the agent desktop permissions. If the group lacks ‘view knowledge’ rights, the suggestions hide even if the config is correct. Permissions override the config every time.