The information on the dashboard is updated hourly.
Some of the information in the Insights dashboard is also available in Explore. See Analyzing your autoreplies with articles and Analyzing your bot builder activity (Legacy).
This article contains the following sections:
Opening the Insights dashboard
You can open the Insights dashboard to view information about your AI agent's performance.
To view the dashboard
- In Admin Center, click Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage AI agents for messaging.
- Click the AI agent you want to monitor.
- Click the Insights tab to display the dashboard.
Understanding the dashboard metrics
The dashboard provides a comprehensive overview of AI agent performance metrics and response performance. The information on the dashboard is updated hourly.
This section describes the following areas of the dashboard:
Performance metrics
The dashboard performance metrics offer you a quick glance at the AI agent's overall performance.
These metrics include:
- Active users: Total number of unique users who actively participated in the conversation, meaning they were sent messages or selected options presented by the AI agent. This metric is updated hourly and includes all interactions, resolved and unresolved, from the last 30 days.
- Transferred to agent: Percentage of unique users who engaged with the AI agent and were handed over to a live agent, resulting in a ticket. These interactions are considered unresolved.
- Automated resolutions: Percentage of AI agent conversations resolved by the AI agent.
For detailed information about how a conversation is resolved by the AI agent, see About automated resolutions for AI agents.
AI agent responses (Legacy)
Response types include:
- Recommended an article: The AI agent found a relevant help center article and presented a link to it.
- Generated a reply: The AI agent found one or more relevant help center articles that addressed the question and shared a summary with the customer in the conversation. Generative replies include a link to the most relevant article.
- Showed an answer: The AI agent presented a custom answer flow.
- Couldn't answer: The AI agent couldn't provide an answer.
The numbers displayed in this chart represent all responses given to users over the previous 30 days. However, responses to the Present options step—either the end user selecting an option or entering free text matching a presented option—are not counted.
Next steps to improve performance
The Next steps to improve performance section includes information you can use to update your AI agent to boost its effectiveness.
Review conversations
- Unresolved conversations: AI agent conversations that were not resolved by the AI agent, including those that remained unresolved and those that were solved by a live agent.
- Automated resolutions: AI agent conversations that were resolved by the AI agent.
For more information, see Reviewing AI agent conversation transcripts.
Improve your help center
This section links you to your help center, where you can create or edit articles to help the AI agent address topics identified in unresolved conversations.
Resolve intents without answers (Legacy)
You can view at a glance the top topics customers ask about during AI agent conversations that do not have existing AI agent answers. Zendesk AI uses past AI agent conversations to identify these top customer topics, known as intent suggestions, based on the following:
- The intent is not linked to an existing answer.
- The intent has been matched to a customer question or comment at least three times over the last 30 days.
- The intent is related specifically to the conversation AI agent the questions came through.
This helps you understand which actions your should take to improve AI agent performance. You can review your top intents without answers, then click through to the full list of intents to create an answer for an intent or to link an intent to an existing answer. Doing so might increase the number of customer support requests resolved and, in turn, lower the percentage of no AI agent responses.
For more information, see Reviewing and assigning intents for common questions without AI agent answers (Legacy).
9 comments
Denise Sehlmeyer
Would like to see this available either in Explore or by non admins (but only the Insights, not rest of Bot) as our learning team needs this information so they can make updates to our help center articles.
8
Jee
Can you share if Explore can cater a customizable field that enables historical based reporting for the bot performance metrics apart from the 7-day performance?
4
Ele Hoare
Is the ‘bot responses’ pie chart viewable in explore? It would be good to have an easy to access way of seeing the unanswerable questions without having to manually read transcripts or review each intent
3
Rachael Norris
Why can we only report on this by week? Would love to be able to aggregate the data and also report on queries.
4
Rachel Martin
I would like to be able to find the associated tickets for the unanswered conversations that were transferred to an agent. Currently the only information you give is the conversation ID and customer ID which I can't search for. What use is that? Can you please add the ticket ID?
1
Filippo Menegatti
So is it possible to have this data customizable in the Explore section?
3
Jakub Raduszewski
How to access this Bot Responses chart in Zendesk Explore?
2
Steven Hampson
Hi there. We are new to Messaging Bots and I'm trying to understand the numbers. There seems to be a contradiction I wonder if someone could help clarify. On the Insights Dashboard, it defines Active Users as ‘users who sent at least one message to the AI agent’; in this article it states ‘Total number of unique users who actively participated in the conversation, meaning they were sent messages or selected options presented by the bot.’
It is users who sent or received at least one message? (I think it makes more sense if the word “were” is removed from the article definition, but not sure if that's correct.)
Thanks!
0
Joshua Bentley
We would also like to be able to report on this via Explore. Ticket deflection has been touted as one of the benefits of the AI agent conversation bots, but we can't see enough data. The Answer Bot Explore dashboard doesn't reflect this data, I'm not finding any pre-made reports, and the existing datasets either don't have the information or I'm missing something.
2