Bring customer evidence into Discovery

The fastest way to understand Discovery is to import one real piece of evidence manually.

1. Open Discovery in your workspace.
2. Use the text import path for a transcript, call notes, a support ticket, or another bounded body of text.
3. Paste the content, confirm the preview, and import it.
4. Open the resulting signal and review the proposed insights.

That first run gives you a concrete baseline for how Zentrik handles your evidence before you spend time on integration work.

When your evidence already lives in a supported tool, connect that source instead of moving data yourself. Open Settings → Integrations or start from the documentation hub to connect providers such as Zoom, Gong, Zendesk, or Jira.

After the connection is set up, import the records you want from Discovery. The same signal-processing pipeline runs after import, but your team does not need to maintain a custom script.

If your source is not listed in the hub, ask us about it first if you want a native integration. When a source already exposes transcripts, tickets, or exports through an API, we can usually help you get to a workable path quickly.

If your team wants to move on its own, use the technical import guides linked below. That path is a good fit when you already have access to the source API and want to load records into Discovery without waiting for a native connector.

You do not need the API to understand or use Discovery. Keep this guide product-focused and use the technical pages only when your team is ready to automate another source.

Use these links in order:

• Signals API quickstart for a first successful single-record import
• Signals import template for the full script with retries, polling, dry run support, and source mapping
• Signals API reference for the endpoint contract

After you confirm your selection (or after a connector delivers the text), Zentrik normalizes the content, proposes learnings, dedupes obvious repeats in that batch, and links proposals back to the signal so wording stays traceable.

Rich customer conversations and tickets often surface several candidates in one pass; shorter or internal notes may surface fewer—the signal remains the audit trail in both cases. Very long bodies may process asynchronously; watch status on the signal until work completes. If a run stalls, retry from the signal or split an oversized input.

Gong, Zendesk, Zoom, and the rest hand text into this pipeline after you choose what to import; paste and file paths feed the same steps directly.

After the first pass, most teams merge, edit, split, or link to existing insights so new quotes attach to learnings the org already approved—one shared map that stays easier to read as volume grows.

As more signals arrive, clusters tighten or shift. Expect the map to move when imports continue.

Semantic themes cluster what customers keep saying; taxonomy is the dimensions and values your team defines for stable rollups; trend views and the map show concentration, gaps, and change as new evidence lands.

For how those pieces fit together, read Taxonomy & themes: Taxonomy and themes (when to use each lens), then Maps and rollups for charts and working-session tips.

If your workspace requires review before items appear in rollups, complete that step so maps match your approval rules.

• Signals API quickstart: first successful API call.
• Signals import template: full import script for real sources.
• Signals API reference: request and response shapes for programmatic imports.
• Insights → opportunities: group learnings into problem and bet space.
• Taxonomy & themes: labels and map lenses.
• From customer evidence to initiatives: full flow to initiatives.
• Workspace context: workspace knowledge.
• Documentation hub: provider setup.

• Paste and uploads use the text your users provide; vendor credentials apply when they use a connector, configured under Settings → Integrations with normal workspace access rules.
• Users typically choose which records to import from a connected system unless a workflow automates selection.
• Processing is server-side; status and errors surface on the signal so people can retry or escalate.