I inherited this Genesys Cloud org six months ago and I am currently auditing our Dialog Engine bot flows. I need to export the utterance history to identify where our intent recognition is failing. I have utilized the ‘Export’ feature in the Bot performance view, but the resulting CSV file is completely missing the ‘Confidence Score’ for each utterance. This makes it impossible to distinguish between a strong match and a lucky guess. Is there an API endpoint to retrieve the full utterance metadata including confidence scores, or is this data intentionally omitted from the export?
I encountered a very similar reporting gap when I was integrating Bold360 (Genesys DX) with the platform. The standard ‘Performance’ export in the UI is highly summarized for supervisor consumption and typically strips out those granular developer metrics like confidence scores. To get the full picture, you must utilize the /api/v2/analytics/botflows/sessions endpoint. This provides a deep-dive JSON stream of every bot turn. Within each session object, you will find an array of intent matches, and each one includes the raw confidence score as a floating-point value. It is much more data-intensive to parse, but it is the only way to perform a proper intent tuning audit.
I am the security officer and I am currently performing a SOC2 compliance review of our bot data retention. I am very concerned that these ‘Confidence Scores’ and raw utterances are being stored in the analytics data lake! If a customer accidentally provides PII during a bot turn, that PII is now captured in your ‘Utterance History’. If you are exporting this data for offline analysis, you must ensure that your export script includes a PII scrubbing layer! I do not want to see raw utterances being saved in an unencrypted CSV on a developer’s laptop! Genesys should really provide an ‘Auto-Scrub’ feature for these bot exports before someone triggers a PCI violation!