AI Agent Analytics gives you a workspace-level view of how your AI agents are performing and how their results compare with an earlier period.
Analytics data combines activity from all AI agents in your workspace. You cannot filter the page to view analytics for an individual AI agent.
Before you start
- To view data in AI Agent Analytics, your workspace must have at least one AI agent.If you haven't created an AI agent yet, follow the instructions in Create an AI agent with conversational setup to create your first AI agent.
- Anyone who can access SleekFlow AI can access the AI Agent Analytics page.
- Analytics data includes activity from all AI agents in the workspace. You cannot filter the page by individual AI agent.
Accessing AI Agent Analytics page
To access AI Agent Analytics:
- Click the
icon in the left navigation bar to go to the SleekFlow AI page - Click the “Analytics” tab in the top menu.
- You will be redirected to the AI agent
Understanding the metrics
AI Agent Analytics contains four main metric cards. Each card has an information icon next to its title that provides a definition of the metric.
Active contacts who interacted with AI
Active contacts who interacted with AI shows the number and percentage of active contacts whose conversations involved AI during the selected period.
A contact is counted as “With AI involvement” when an AI agent sends at least one message in their conversations. Otherwise, the contact is counted as “Without AI involvement”.
The card shows:
- The number of active contacts with AI involvement.
- The percentage of all active contacts they represent.
- The change in the number of contacts compared with the selected comparison period.
- Active contacts by AI involvement, split into With AI involvement and Without AI involvement.
- The total number of active contacts in the selected period.
Conversation volume and reply time
Conversation volume and reply time shows how conversations were handled during the selected date range. You can switch between two views:
- Conversation volume
- First reply time
Conversation volume is shown by default. Switching between the two views does not change the selected date range.
Conversation volume
The Conversation volume view shows conversations that received a reply, grouped by AI involvement:
- With AI involvement: An AI agent sent at least one message in the conversation.
- Without AI involvement: No AI agent sent a message in the conversation.
The chart shows conversation volume for each day or week, depending on the selected date range.
Hover over a point on the chart to see the number of conversations in each group and the Total conversations.
First reply time
The First reply time view shows the average time taken to send the first reply, split into:
- AI first reply
- Human first reply
If there are no replies of one type for a particular day or week, the corresponding line is interrupted and the tooltip shows No replies instead of a value of zero.
If there is no reply data for the selected period, the chart is replaced with No replies in this period.
Note: The Volume and First reply time views measure different things and their figures do not add up to each other. Volume counts conversations, while First reply time calculates average reply times. A conversation can also contain both an AI first reply and a staff user first reply.
No AI involvement on this card follows the same rule as Human-only on the CSAT card: no AI agent sent a message.
This does not necessarily mean a human handled the conversation. It can also include conversations handled only through automation, such as broadcasts or Flow Builder messages.
Exit conditions
The Exit conditions card shows how AI sessions ended during the selected date range, grouped by exit condition.
It includes both:
- Exit conditions configured for your AI agents.
- Exit conditions recorded by SleekFlow, such as timeouts.
For instructions on configuring exit conditions, see this Help Center article on configuring your AI agent.
The summary below the card title shows the number of AI sessions that ended, the number of exit conditions recorded, and, when applicable, the number of AI agents included.
Handover rate
Handover rate is the percentage of ended AI sessions where the AI agent met a configured exit condition and stopped handling the conversation.
Only AI sessions that have reached an end state are included. Sessions that are still open are not counted.
Below the handover rate, you can see:
- The change in percentage points compared with the selected comparison period, such as +2.1 pts.
- The number of AI sessions that ended and their share of all AI sessions during the selected period.
For example:
1,048 AI sessions ended, representing 62% of all 1,690 AI sessions.
Changes in handover rate are displayed in a neutral color. A higher or lower handover rate is not necessarily positive or negative on its own.
Exit condition table
The table provides a breakdown of ended AI sessions by exit condition. It includes:
- Exit condition: The name of the recorded exit condition.
- AI sessions: The number of AI sessions that ended with that condition, along with its share of all ended AI sessions.
- Agent: The AI agent associated with the exit condition. This column appears when data from multiple AI agents is included.
Rows are sorted by the number of AI sessions, from highest to lowest. The first 10 rows are shown by default.
Exit condition names are shown as recorded and are not merged. This means the same exit condition used by different AI agents, or conditions with different names or spellings, can appear as separate rows.
If an AI agent has been deleted, Deleted agent appears in the Agent column.
Note: This table counts AI sessions, not conversations. A conversation can contain more than one AI session, so these figures may not match the conversation counts in Conversation volume and reply time.
Setting the date range and comparison
The filters at the top of the page determine the data shown across AI agent analytics.
Date range
Use Date range to choose the period you want to analyze.
You can select:
- Last 7 days: The default option. Includes today and the previous six days.
- Last 30 days: Includes today and the previous 29 days.
- Last 90 days: Includes today and the previous 89 days.
- Custom range: Lets you choose a start and end date.
Dates are based on your personal timezone setting. If you do not have a timezone set, your company's timezone is used instead.
Select a custom date range
To select a custom range:
- Click Date range.
- Select Custom range
- Select a start date and an end date.
- Click Apply.
Click “Cancel” to close the calendar without changing the date range.
Custom ranges follow these rules:
Rule |
Message shown |
You must select both a start and end date. |
Select both a start and end date. |
The end date must be at least one day after the start date. |
End date must be at least 1 day after the start date. |
The end date cannot be later than today. |
End date must be before today. |
The date range can be up to six months long. |
Date range can't exceed 6 months. |
For date ranges of 45 days or less, charts group data by day.
For longer date ranges, including Last 90 days, charts group data by week. Weeks start on Monday, so the first week shown can begin before the selected start date.
Compare to
Use Compare to to choose the period you want to compare your analytics data against.
You can select:
- Previous period: The default option. Compares the selected date range with the same number of days immediately before it.
- Previous year: Compares the selected date range with a period of the same length starting one year earlier.
A comparison is always applied and cannot be turned off.
Comparison data is shown for:
- Active contacts who interacted with AI
- AI-assessed CSAT
- Exit conditions
Depending on your selection, changes are shown as vs previous period or vs previous year.
Conversation volume and reply time does not show comparison data.
Reset the filters
Click Reset to return the filters to:
- Last 7 days
- Previous period
Filter selections are not saved. Each time you open AI agent analytics, these default options are applied.
What AI agent analytics does not show
AI agent analytics does not currently provide:
- Filtering by individual AI agent
- Breakdowns by messaging channel
- Breakdowns by flow
- AI credit usage
- Knowledge source usage
- Lead scores