Copilot Real-Time Assist - Knowledge Article Relevance Score unexpectedly low

General Discussion
1

Is anyone else seeing consistently low relevance scores from Copilot’s real-time assist? We’re on Genesys Cloud, API v2. The expectation is the platform should surface articles exceeding a score of 0.7, but it’s consistently below 0.5 - basically doing jack all.

The knowledge base connector syncs fine - no 503s this time, thankfully. Article metadata is populated correctly, and the intent mapping in the Copilot UI looks sound. I’ve checked the relevance scores returned by the platform; they’re coming back like this.

{
 "articleId": "kb-1234",
 "relevanceScore": 0.32,
 "matchedIntents": ["billing", "payment"],
 "queryText": "payment issue"
}

The documentation states the relevance score considers semantic similarity and the article’s popularity (view count, upvotes). Low view counts could be the culprit. It’s worth noting, new articles consistently score lower.

1 Like