Recording sessions

When ambient knowledge is enabled for an agent, we record a small sample of sessions to improve our service. Our system automatically selects and analyzes some of these sessions to generate knowledge articles, ensuring your knowledge base evolves directly from real interactions.

---

Knowledge generation

With the sessions that we record, not all are equally valuable. This guide explains how we identify the most useful ones for knowledge generation.

What We Look For

Active Users

We prioritize sessions where users are engaged - clicking through pages, filling out forms, and exploring features. More interaction reveals more about how people use the product.

Sufficient Duration

Sessions need enough time to show a meaningful journey.

Varied Activity

The most valuable sessions include a mix of actions - navigation, form input, and scrolling. This variety typically indicates genuine product use rather than someone stuck in one place.

What Gets Deprioritized

Frustrated Users

Sessions where users struggle - repeatedly clicking the same element or encountering errors - are ranked lower. While useful for troubleshooting, they don't represent typical usage.

Minimal Interaction

Sessions with almost no activity don't provide meaningful insights.

Why This Approach

We aim to surface sessions that represent normal user behavior. This means:

• Focusing on typical journeys, not outliers
• Prioritizing quality over volume
• Showing how users naturally interact with the product

Summary

Sessions are selected based on three factors:

1. Engagement - How actively did the user interact?
2. Duration - Was the session long enough to be meaningful?
3. Experience - Did the session represent typical usage?

This ensures the sessions you review reflect real user behavior.