GenAI Analytics

intent insights overview intent insights is a genai powered analytics feature that uses genai to help you understand what users are asking your chatbot , how often those queries occur , and how effectively your bot is handling them it enables you to identify the top user intents track answered vs unanswered intents understand query frequency trends measure the quality of bot responses over time intent insights is available as an early access (beta) feature and will continue to improve as we refine the analytics models where to find intent insights in the console ui bot → analytics → genai analytics → intent insights if the genai analytics section does not exist for your version or bot, it will be created automatically when the feature is enabled what you’ll see in intent insights top intent insights this section provides an intent level view of user queries handled by your bot for each query, you can see user intent grouped representation of similar user queries frequency number of times the intent occurred answered / unanswered the user's query mapped to the answered or unanswered section based on how the bot handled it average satisfaction score (1 to 5) you can sort intents by frequency identify high impact unanswered intents track performance trends over time search for specific intent or questions reports intent insights supports two downloadable reports overall summary report contains all query intents (answered and unanswered) includes frequency and average satisfaction score reflects exactly what is shown in the table for the selected date range intent level report available for individual intents that are created in the analytics the report includes actual user queries bot responses timestamps ratings useful for deep analysis and training improvements sample report note the reports need to be downloaded separately for answered & unanswered intents, and the data depends on the data range filter key definitions answered query a query is marked as answered when the chatbot understands the user’s intent and handles it independently using its configured flows, knowledge, or reasoning without transferring the conversation to a live agent unanswered query a query is marked as unanswered when the chatbot cannot handle the user’s intent on its own this may happen due to missing training, incorrect or fallback responses, or when the conversation is transferred to a live agent an unanswered query does not always mean a poor experience, it indicates human or system dependency frequency frequency shows how many times a particular query topic was raised by users during the selected time period similar queries are grouped together to provide a clear view of common user needs average satisfaction score the average satisfaction score represents how well the chatbot handled a query intent, rated from 1 to 5 across sessions this score reflects response quality and relevance, regardless of whether the user's query was answered or unanswered how intent insights works (in simple terms) intent insights runs after a chat session ends (no real time processing) conversations are analyzed using genai to group similar queries into intents classify them as answered or unanswered assign a quality rating (1 to 5) this process does not impact live bot performance supported bot types intent insights works across all available chatbot configurations, depending on the enablement of the feature in the bot's setup feature availability & enablement intent insights is available as early access (beta) for selected customers it can be enabled at the bot level and not enabled by default data starts populating from the date of enablement onwards to enable this feature, please contact your account manager or product team note enabling intent insights may incur additional costs based on your setup known limitations (beta) insight's accuracy will improve over time as models and prompts are refined analytics are post session (not real time) the feature does not auto train the bot, manual improvements are required best practices review high frequency unanswered intents weekly by downloading the individual intent level report for the low rating intents understand the type and way the queries are asked by the users and train them on the bot see the cases where the bot does not have an answer for the frequently asked queries, and add them to the flows or faqs check for any anomaly in the patterns in a sudden bump in the spike, negative rating for a previously working flow, and fix them this may be due to unhandled edgecases on the bot flows api issues delays, timeouts, etc any changes done in the existing flows track satisfaction score trends to measure improvement on the bot frequently asked questions does this affect live bot performance? no all analysis happens after the session ends can short answers still receive high scores? yes ratings are based on correctness and relevance, not response length does this replace existing analytics? no intent insights complements your existing analytics