AI Insights

Overview

AI Agent Metrics is a report inside your Reports area that shows how your AI Agent is performing on real customer tickets. In one place it tells you how many conversations the AI handled, how many it resolved without a human, how many it passed to your agents, where it got stuck, how customers felt about those conversations, and which topics deserve your attention first.

It is built for two kinds of readers. If you manage a support team, it tells you where the AI is helping and where it needs better setup. If you use the product day to day, it helps you see which customer topics the AI handles well and which ones still reach your agents.

Before you start

• Your AI Agent must be turned on. AI Agent Metrics only appears when the AI Agent is enabled for your workspace. If you do not see the tab, your AI Agent is not active yet — enable it first, or ask your account admin.
• You need reporting access. Viewing the Reports area requires analytics permission on your account. If the Reports icon or the report is missing, ask your workspace admin to grant you reporting access.
• Give it some conversations. The report reflects tickets your AI Agent has handled over the selected time period. A brand-new setup, or a period with no AI-handled conversations, will show empty states (“No Reports Found”) until there is activity.
• Availability. During the beta, AI Agent Metrics is being rolled out gradually. If your workspace has the AI Agent enabled but the tab is not yet visible, it may not be switched on for your account yet — your account manager can confirm.

Step-by-step

1. Read the Overview

The Overview tab answers “how is my AI Agent doing overall?”.

Ticket Flow Summary — the large number, Closed tickets started with AI, counts closed tickets that were first assigned to the AI Agent (whether the AI finished them or a human did). Below it, three figures break down what happened to those tickets:

1. Resolved by AI Agent — the share of tickets that were assigned to the AI and got resolved by the AI, by a user or automatically during the selected date period.
2. Reassigned to Human Agent — the share the AI handed over to a human agent.
3. Stuck Queries — the share where the AI couldn’t resolve the request because of missing knowledge, unclear instructions, a system issue, or an unexpected input.

On the right, two more cards:

1. Total Sales by AI Agent — the total value of orders placed by customers while they were interacting with the AI Agent, over the selected period. The arrow on this card takes you to the Revenue Metrics report for the full breakdown.
2. Net Sentiment Score — an AI-based score of how positive, negative, or neutral customer conversations were, calculated as positive minus negative sentiment on a scale from −100 to +100 (shown as Negative (-100) · Neutral (0) · Positive (+100)). The arrow opens the sentiment view under Topic Breakdown.

Ticket Flow Overview — below the cards, this flow diagram shows where tickets start, how the AI handles them, and where the work ends up, reading left to right across four stages: Channels → AI Handling → Topics → AI Agent Score. Hover over any point to highlight the paths that pass through it. Use the Export button (top right of this section) to receive the underlying data by email (see step 6).

2. Explore Topic Breakdown

The Topic Breakdown tab (heading Topic Volume and Outcomes) groups AI-handled tickets by topic. Choose one of the three views at the top right:

1. Reassignment Topics — topics most often handed from the AI to a human agent.
2. Negative Sentiment Topics — topics where customer sentiment skewed negative.
3. Stuck Queries — topics where the AI most often got stuck.

Within each view:

1. On the left, the topic list shows each topic with its ticket count and its share of total tickets. Use the Sort by control to order topics by the highest or lowest number of tickets.
2. Select a topic to open its Topic Overview on the right — a detailed breakdown including its outcome split, a trend over time, an AI Agent Helpfulness Score, a Recommended Next Step, and a preview of Related Tickets.
3. Select View All Tickets to open the full list of tickets for that topic.

3. Prioritise with Focus Areas

The Focus Areas tab (heading Impact vs. Volume Matrix) helps you decide what to fix first. Each dot is a topic; its position combines how many tickets it has (volume) with how urgently it needs action (impact). As the on-screen note explains: bigger dots and top-right placement mean higher priority, and urgency is calculated from how often tickets get escalated to a human, get stuck, or carry negative customer sentiment. Hover a dot to see that topic’s volume, reassignment rate, negative-sentiment rate, stuck rate, and overall impact score.

4. Export the data

1. Ticket Flow Overview export — on the Overview tab, select Export in the Ticket Flow Overview section to export the full flow data for the current period and channels.
2. Topic tickets export — from a topic’s View All Tickets page, use Export to export that topic’s ticket list.

In both cases the export is prepared in the background and sent to you by email — you will see the confirmation “Data export has started. You will receive an email containing the exported data shortly.” There is no direct download in the browser; check your inbox for the file.

Limits & good to know

• Beta feature. Figures are still being validated during the beta; use them as a guide, not as a final source for formal reporting.
• Requires the AI Agent to be active. With no AI Agent enabled, the report will not be available.
• Empty states are normal at first. If a tab shows No Reports Found, there is simply no data for the chosen channels or time period — widen the date range or adjust the channel selection and check again.
• Comparisons use the previous equal period. A change such as “from 0 of previous 8 days” compares your selected window against the window immediately before it of the same length.
• A ticket can appear under more than one topic. Because a single conversation can touch several topics, topic-level counts can add up to more than the total number of tickets — this is expected.
• Sales and sentiment are AI-derived. The sales figure reflects orders placed during AI interactions, and the sentiment score is an AI estimate of conversation tone — both are indicative.