Launching the generation

Before generating a taxonomy, you define the analysis scope:

• The period: by default the last 30 days, but you can freely adjust the time window according to your needs — for example going back 6 months for a more representative view, or targeting a specific period around an event
• The sources: if you have multiple data sources connected, you can choose which ones the AI should use to generate the taxonomy

Once the scope is defined, you tell the AI what you want to understand — for example the contact reasons of your customers, the friction points on your online journey, or the topics raised in your inbound calls. The AI then analyzes all the data over the selected period and proposes a structured taxonomy.

What the AI proposes

The generated taxonomy is organized into three hierarchical levels:

• Categories — the major themes identified in your data, for example After-Sales Service, Delivery, Online Experience
• Subcategories — more precise topics within each category, for example Warranty management, Refund delays
• Findings — the most granular observations, extracted directly from verbatims, for example Transparency on repair timelines, Clarity of warranty coverage

This structure allows you to move from a macro view to a very fine reading of the topics expressed by your customers, without ever losing the thread.

Managing and refining the taxonomy

The AI-generated taxonomy is a working base, not a fixed result. You can freely:

• Modify the name or description of a category, subcategory, or finding
• Delete themes that are not relevant to your use case
• Manually add categories or topics you want to monitor even if they didn't emerge spontaneously from the analysis

This is an important step: a taxonomy well-calibrated to your business needs will produce much more actionable dashboards.

Validating the taxonomy

Once you're satisfied with the structure, you validate the taxonomy. It is then ready to be attached to an Insight Explorer dashboard to move on to visualization.

‍