Skip to main content

How to Start Generating Insights

Insights are the data points Harmony extracts from your conversations. To begin analyzing meetings, you must first organize your work into Projects.

Projects act as the container for your analysis, grouping the specific questions you want to ask with the teams that are asking them.

1. Open or create a project​

Insights only run within the context of a project. To begin, click Insights in the navigation sidebar. Select an existing project that matches your current initiative. If you do not have any projects yet, follow our guide on How to set up a Project.

A project is essentially a "rulebook" for Harmony. It tells the AI which teams to listen to, what questions to ask, and what context to use when answering them.

2. Define your Insights​

Definitions tell the AI exactly what data to extract from your conversations.

  1. Navigate to the Configuration tab and define which conversations should be analyzed by this project. You can set the basic filters such as "only the calls longer than 5 minutes" or "Smart Criteria" written in the natural language, e.g. "run only if the customer asks about the discounts".
  2. Add Project know-how to provide details and context that will help the AI understand this project better.
  3. Then, to start analyzing meetings - Add Definition in the Definitions section. To do so:
    • choose the Insight Type (e.g., Yes/No, Checklist, or Scale).
    • Enter the Question you want the AI to answer (e.g., "Did the customer mention a budget?").
    • Set the desired outcome and add the additional context for better analysis results.
    • Link the previously created Project knowledge in the Data Source section to use for insight generation.
  4. Click Create.

Types of Insights you can create​

  • Summary extracts key points and action items from conversations. Use this to quickly understand what was discussed without reviewing full transcripts.
  • Competitors Analysis analyzes mentions of competitors and their advantages in conversations. The system automatically identifies which competitors come up and captures the context around those discussions.
  • Discount tracks whether any discounts were mentioned in the conversation. Use this to monitor pricing discussions and discount frequency across your deals.
  • Buyer Engagement analyzes buyer's engagement level on a scale of 1-10. Scoring adapts based on your team type (sales, support, recruiting, operations) to measure how involved the other party was during the conversation.
  • Open Question asks any custom question about the conversation. Use this for flexible analysis when you need answers to specific business questions unique to your workflow.
  • Yes/No Question asks a question that can be answered with Yes or No. Use these for qualification criteria, compliance checks, or any question with a clear binary answer.
  • Checklist creates a checklist of items to verify in conversations. Use this to track whether specific topics, requirements, or process steps were covered during meetings.
  • Markdown runs a custom prompt against conversation data and displays results in markdown format. This gives you full control over how insights are structured and presented for complex analysis needs.
  • Group analyzes conversations grouped by specific criteria. Use this to organize data by deal stage, customer segment, issue type, or any category you define for pattern comparison.
  • Scale Rating rates aspects of conversations on a custom scale with optional labels. Use this for scoring conversations across multiple dimensions like interest level, pain severity, or solution fit.
  • Funnel analyzes conversation stages and conversion through your sales funnel. See which milestones get covered and which get skipped to identify gaps in your process.

3. Rerun historical data​

If you add a new Insight to an existing project, it will not automatically analyze past meetings. You must trigger a rerun to process historical data.

  1. Navigate to the Rerun tab.
  2. Select a Date Range.
  3. Choose to Force rerun (includes completed conversations) or Keep existing insights.
  4. Click Preview to see the affected volume, then execute the queue.

4. Viewing the results​

Once configured, results appear in two places:

  • Meeting Level: Open any individual meeting to see the specific answers for that call.
  • Reports Tab: View aggregated charts, response distributions, and trends across all meetings in the project.

5. Best Practices​

Start Simple - Begin with basic questions like "How many times the competitior was mentionned?" or "What's our average engagement?".

Build Over Time - Expand as you learn. Observe what metrics matter, add complexity gradually, and refine your dashboards.

Regular Review - Check insights regularly to adjust as needs and knowledge change, remove unused widgets and add new questions.

Think of custom insights as asking questions. Start with "What do I want to know?" and phrase it naturally.