What it does

Analytics shows how an agent is actually performing: session volumes, resolution and escalation rates, which topics fire, and where conversations fall over. It's how you tell whether the thing is earning its keep.

Key facts

• Core measures include engagement/resolution rate, escalation rate and abandonment.
• Topic-level data shows what's triggering and what's failing to match.
• Generative-answer usage and outcomes are reported alongside scripted topics.
• Data can be exported for deeper analysis rather than living only in the built-in dashboards.

When to use / skip

Use analytics from the day an agent goes live — it's how you prove value and find what to fix. There's no real "skip"; an agent you don't measure is one you can't improve or defend at budget time.

Configuration decisions

• Which metrics define success for this agent, agreed with the business.
• Whether you export to Power BI or another tool for richer reporting.
• Who owns the numbers and acts on them after launch.

Gotchas

• "Resolution" is inferred, not certain — a user giving up can look like a resolved session. Read the metrics sceptically.
• Vanity volume metrics flatter; escalation and abandonment tell you the real story.

Consultant notes

• Agree success measures before build, not after — otherwise analytics becomes a post-hoc justification exercise.
• Plan a review cadence with an owner; dashboards no one reads change nothing.
• Tie analytics back to the business case so value is demonstrable at renewal and capacity reviews.

---

Recheck available metrics and exports each release.