Skip to main content

What are Insights?

Insights turn thousands of conversation minutes into objective, trackable data. You define the questions and metrics that matter to your organisation, and Harmony's AI extracts answers from every meeting in scope.

Instead of relying on subjective summaries, Insights create a single point of truth: every conversation is analysed against the same definitions, so a "Qualified Lead" or a "Successful Demo" means the same thing across teams.

Insights live in the Data Lake (the sidebar item with the database icon, route /org/{slug}/insights). Inside the Data Lake, work is organised into Projects.

The three building blocks​

Harmony's Insights system is built from three concepts:

Harmony’s Insights system is structured around three core building blocks that work together to organize, analyze, and interpret your organization’s conversational data.

The first building block is the Project. You can think of a project as the central container that holds everything related to a particular business initiative or thematic area—such as a product launch, sales methodology, or coaching effort. A project determines which conversations should be included or excluded by specifying assigned teams and, optionally, project criteria that further refine eligibility. Beyond simply delineating the data in scope, the project also controls what questions or analyses will be performed by housing the Insight Definitions. Furthermore, you select the language the AI should use when analyzing your data, tailoring results to your team’s working language. To ensure that analyses are meaningful, each project can be given detailed background context in the form of Project Knowledge, which shapes how the AI understands and interprets your conversations in line with your business reality.

The second building block is the Insight Definition. These are the specific analytical questions or metrics that Harmony’s AI runs against each conversation included in the project. An Insight Definition could ask something like, “Was a competitor mentioned?” or “Did the conversation include a discount discussion?” Each definition runs separately on every eligible conversation, producing one insight result per meeting. This approach ensures that every conversation is evaluated in the exact same way, creating objective, consistent measures for topics that matter to your organization.

The third building block, known as Project Criteria, serves as an optional but powerful filtering system. While a project normally includes every conversation from its assigned teams, Project Criteria enable you to be more selective—perhaps only including meetings of a certain type, with specific tags, or involving particular team members. By defining these criteria, you can make sure that only the most relevant conversations are analyzed for a given project, keeping your insights focused and accurate.

For a deeper understanding of how these components work together, see Setting up a project, Generating insights, and Project knowledge.

Once your projects and insights are configured, you can follow the journey of raw conversations as they are transformed into actionable data. Every answer produced by an Insight is traceable—each result links directly back to its source conversation. This means you can start with a high-level metric and always drill down to the specific meetings and evidence that support the outcome.

You have the ability to analyze your data at different levels of detail. If you want to focus on a single meeting, you can view the answers generated for just that conversation—such as whether the customer mentioned a budget. This granular detail is accessible on the individual conversation’s Data Lake tab. If you are interested in broader trends, you can step back to the project-level Reports tab, where you will see the aggregated results across potentially hundreds of conversations. Here, you might examine patterns like the average buyer engagement across all sales calls in a given month.

When it comes to the types of Insight Definitions you can add, Harmony provides a rich and flexible set of options organized into three categories.

Among the most popular types are those that address common analytic needs:

  • The Summary Insight produces a concise, structured overview of the conversation, highlighting the most important points and any identified action items.
  • The Competitors Analysis Insight detects which competitors are mentioned during a meeting and can highlight both their relative positioning and any stated advantages. You can further customize this by specifying a list of competitors to track or names to exclude from detection.
  • The Discount Insight identifies when a discount is discussed within a conversation and, where possible, captures details such as the specific rate or amount mentioned.
  • The Buyer Engagement Insight evaluates the participant’s engagement level on a standardized scale from 0 to 10. This isn’t just a number; every band of this scale is given an explicit interpretation, so you’ll know, for instance, that scores of 8–10 indicate strong engagement, while a score below 6 points to limited involvement.

For teams that want even more customization, Harmony offers insight types that you can tailor to your exact needs:

  • With Open Question, you define any free-text question you want the AI to answer, and it provides a unique response for every conversation.
  • The Yes/No Question Insight lets you specify a direct question with a clear affirmative or negative answer, and you can decide which outcome represents success or a positive result in your reporting.
  • Using Checklist, you create a series of items or steps for the AI to check in the conversation; each is evaluated and given a clear outcome, along with the reasoning behind it.
  • The Markdown type lets you supply a completely custom prompt, and the AI’s detailed response is returned as a rendered Markdown note.
  • Group allows you to categorize conversations by defining specific buckets with their own descriptions—the AI will automatically assign each conversation to the most appropriate group.
  • The Scale Rating gives you control to set up your own numeric scale—such as from 1 to 5, or any other range—plus the option to label each value for context.

Finally, Harmony includes some built-in insight types for specialized analysis:

  • The Funnel type helps track progression through predefined stages in your sales or engagement process, identifying where each conversation has reached in the funnel.
  • SPIN Improvement automatically offers suggestions to enhance the use of SPIN selling techniques based on the conversation’s content.
  • SPIN Distribution breaks down the conversation to show how much each part followed the four quadrants of the SPIN sales methodology: Situation, Problem, Implication, and Need-Payoff.

Each of these insight types is designed to answer important business questions in a way that’s structured, repeatable, and transparent—transforming unstructured conversations into consistent, actionable intelligence.

Both produce a numeric score, but they are different. Buyer Engagement is fixed at 0–10 with built-in interpretation per band. Scale Rating is fully configurable — you choose the range (default 1–5) and the labels per value.

For the full configuration reference, see Generating insights.

How Projects bring it together​

Insights are organised into Projects. A Project groups your definitions around a business initiative — like a product launch, a training programme, or a customer expansion campaign.

You assign one or more teams to the project. Eligible conversations are conversations whose conversation team is one of the project's teams. Project Criteria can narrow that further. Once a meeting is processed, Harmony runs every active definition in every project the meeting qualifies for and writes one result per definition per conversation. The Reports tab updates over WebSocket as new results land, so dashboards stay fresh without manual refresh.

Next steps​