AI Usage by Agent & Feature

Overview

The AI Usage by Agent & Feature report gives you visibility into how your team is using built-in AI capabilities (Reply Polish and Ticket Summary). It breaks down usage two ways — by agent, to see who’s relying on AI most, and by feature, to see which AI capabilities are used most — along with the token volume each generates, so you can track adoption and gauge relative usage without needing cost data.

Where to Find

Go to Support → Reports → AI Usage

AI Usage by Agent & Feature
How It Works

The report displays a graph with the following types of data:

  • Usage by Agent 

    This section lists AI usage broken down by each agent. For every agent, it shows how many AI requests they made in the selected period, the total number of tokens consumed across those requests, and the average tokens per request (total tokens divided by request count, rounded up). This tells you which agents are relying on AI features most, and how “heavy” their typical AI usage is.

  • Usage by Feature

    This section shows the same usage data, but grouped by AI feature instead of by agent — currently either Reply Polish or Ticket Summary. For each feature, it shows how many times it was used in the selected period, the total tokens consumed by that feature, and the average tokens per use. This tells you which AI capabilities are being used most, and how much each one typically costs in tokens.

Together, Usage by Agent answers “who is using AI,” while Usage by Feature answers “what AI capabilities are being used.”

You can select a predefined duration or choose a custom date range. You can also apply custom filters to generate more specific results.

Filtering the Report
You can generate the report for different time periods, such as:

  • Last 30 days
  • This week
  • Last quarter
Notes / Tips
  • Only two features are currently tracked: Reply Polish and Ticket Summary. Any AI action outside these isn’t broken out separately yet — it’s bucketed under “Ticket Summary.”
  • This report measures tokens, not cost. There’s no dollar/currency figure — to estimate spend, apply your AI provider’s per-token pricing to the token totals yourself.
  • A rising average tokens per request is worth investigating, not necessarily alarming. It can mean longer tickets or prompts are being summarized or polished — check a few underlying logs before assuming inefficiency.
  • Usage by Agent reflects who triggered the AI action, not who owns the ticket. An agent who leans on AI to draft replies will show high usage here even if a different agent ultimately handles and closes the ticket.
What are your feelings