> For the complete documentation index, see [llms.txt](https://docs.harmony.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.harmony.io/dashboard-and-analytics/analyzing-conversation-insights-and-ai-enrichments.md). # Analyzing Conversation Insights and AI Enrichments Harmony automatically analyzes every conversation through its evaluation pipeline and surfaces the results as AI enrichments - structured signals that help you understand what happened in a conversation and why. You can use these enrichments to filter, explore, and act on your conversation data directly in the conversations viewer, without any manual configuration or refresh steps. This article explains the available AI enrichments, how scoring and reasoning work, and how to use filters to find the conversations that matter most. {% hint style="info" %} The AI enrichment filters in the conversations viewer replace the previous conversations analytics in the management dashboard. All conversation insights are now consolidated in a single place. {% endhint %} ### How AI Enrichments Work Each conversation is automatically assessed by Harmony's evaluation pipeline after it ends. The pipeline evaluates the conversation across several dimensions and writes the results back to that conversation record in real time. Because enrichments are synced continuously, the filters you apply always reflect the latest assessed data - there is no need to manually trigger a refresh or export. Every enrichment produces two pieces of information: * **Score** - a structured value (such as a label or numeric rating) that categorizes the conversation for that dimension * **Reasoning** - a plain-language explanation of why the conversation received that score, giving you the context behind the assessment ### Available AI Enrichments #### Sentiment Sentiment reflects the overall emotional tone of the conversation as assessed by the AI. Use sentiment to quickly identify frustrated or dissatisfied users who may need follow-up, as well as highly positive interactions worth highlighting. The score indicates the detected tone (for example, positive, neutral, or negative), and the reasoning describes the specific signals in the conversation that led to that assessment. #### Interest Level Interest level captures how engaged or motivated the user appeared during the conversation. A high interest score typically indicates a user who is actively exploring your product or service, while a low score may suggest passive browsing or uncertainty. The reasoning for interest level can help you understand what topics or moments drove engagement, which can inform content and product decisions. #### Resolution Resolution indicates whether the conversation reached a successful outcome - meaning the user's question or goal was addressed. This enrichment helps you identify conversations where the AI or your team fully resolved an issue versus those that ended without a clear answer. Use resolution scores to spot patterns in unresolved conversations and find areas where your knowledge base or bot configuration may need improvement. ### Filtering Conversations by AI Enrichments AI enrichment filters are available directly in the conversations viewer. 1. Open the **Conversations** viewer from the main navigation. 2. Select **Filters** to open the filter panel. 3. Choose any of the available AI enrichment dimensions - **Sentiment**, **Interest Level**, or **Resolution**. 4. Select the score values you want to include and apply the filter. 5. Click on any conversation in the results to read the full enrichment details, including the reasoning explanation for each scored dimension. You can combine multiple enrichment filters with other conversation filters (such as date range, channel, or agent) to narrow results further. ### Using Reasoning to Understand Scores The reasoning field is particularly useful when a score is unexpected or when you want to validate the AI assessment before taking action. For example, if a conversation is scored as unresolved but you expected it to be complete, the reasoning will point to the specific turn or gap that caused the assessment. Reasoning explanations are written in plain language and reference the content of the conversation, so no additional interpretation is required. ### Tips for Getting Value from AI Enrichments * **Identify at-risk users** - Filter for negative sentiment combined with low resolution to find conversations where users may have left unsatisfied. * **Spot high-value leads** - Filter for high interest level to surface users who showed strong engagement and may benefit from proactive outreach. * **Improve bot coverage** - Regularly review unresolved conversations to find recurring questions your bot is not answering well. * **Validate quality** - Use sentiment and resolution together as a quick quality check across a time period or a specific channel. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.harmony.io/dashboard-and-analytics/analyzing-conversation-insights-and-ai-enrichments.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.