Issues in Agent Analytics helps you identify and track potential problems in AI agent conversations. To find them, go to Product > Agent Analytics, select an AI agent or saved report, then select the Issues tab. This tab includes two types:
- Emergent issues are detected automatically and update dynamically based on recent conversation data.
- Tracked issues are persistent and let you monitor a specific problem over time.
Issues help you:
- Understand where visitors are struggling or getting frustrated with your AI agent.
- Identify the underlying causes of elevated rage prompt rates.
- Prioritize improvements to agent prompts, workflows, or guardrails.
Note: Issues are available for agents configured to capture full conversations. Prompt-only agents aren't supported.
The Issues tab has two views: emergent and tracked issues. The Emergent tab opens by default. Select the Tracked tab to view issues you've created to monitor over time.
Understand issue types
Each detected issue falls into one of three types, shown as a percentage of total conversations. Use these percentages to identify where your agent is struggling and prioritize improvements.
- Rage prompts. The percentage of conversations that contain at least one rage prompt, which is a visitor message expressing frustration, anger, or hostility. Signals include profanity directed at the agent, hostile all-caps clauses, or explicit frustration markers such as "try again" or "that's not what I asked," combined with evidence that the agent failed.
- Unsupported requests. The percentage of conversations where the agent could not fulfill the visitor's request. For example, the agent refuses to help or directs the visitor to another channel.
- Errors. The percentage of conversations that encountered an error, such as a system failure, processing limit, or failed tool or API call.
Understand emergent issues
Pendo automatically detects emergent issues by analyzing conversation content for patterns that indicate confusion, frustration, or failure to complete tasks. These groupings update as new conversation data comes in and reflect your selected filters at the top of the page. No additional configuration is required.
Each row in the Emergent issues table includes:
- Issue. A summary describing the detected problem.
- Instances. The number of times the issue occurred across all conversations.
- Conversations. The number of conversations associated with the issue.
- Visitors. The number of unique visitors affected.
- Accounts. The number of accounts associated with those conversations.
- Last occurred. The most recent occurrence of the issue.
Track an issue
Tracking an issue lets you save a specific problem and monitor how it performs over time. Emergent issues change based on filters and recent activity, but tracked issues persist as named problems in your agent's analytics so you can measure ongoing impact.
You can create a tracked issue if you want to:
- Measure how often a known problem occurs across conversations.
- Monitor the effects of changes to your agent prompts, workflows, or guardrails.
- Share consistent insights with engineering, support, or other teams.
You have two ways to create a tracked issue:
- From the Tracked tab, select + Create tracked issue above the table.
- From the Emergent tab, select + Track issue in the table row of the emergent issue you want to track.
Then:
- Enter a name for the issue. If you started from an emergent issue, the field is prefilled and editable.
- Describe the problem the agent is having. Use plain language to describe the problem. For example, "Agent says there's no data even though data exists for this account."
- Select Next. Pendo uses your description and existing conversation data to automatically generate detection rules.
- Review the auto-generated rules, and edit them or add your own. Up to 10 rules are supported.
- To edit a rule, select the rule's text box and make your changes.
- To add a rule, select + Add rule, then enter the rule.
- Select Create issue.
After creation, the tracked issue shows as Processing in the Tracked issues table. Processing starts at the beginning of every hour and can take up to 15 minutes.
Edit or delete a tracked issue
- Go to the Tracked issues table.
- Hover over the issue, then on the far-right of the table row:
- Select Edit to update the name, description, or rules. Edits apply to the last 90 days of conversations, and re-evaluation can add or remove conversations from the issue based on your updated name, description, or rules.
- Select Delete to remove the issue.
Note: When you add or edit a tracked issue or its rules, Pendo reprocesses your conversation data. Results can change as reprocessing completes.
View and manage tracked issues
Tracked issues are always listed in the Tracked issues table, but the data shown for each issue reflects your current filters at the top of the page.
The Tracked issues table contains the following columns:
- Issue name. The name and description entered when the issue was created.
- Instances. The number of times the issue occurred across all conversations.
- Conversations. The number of conversations associated with the issue.
- Visitors. The number of unique visitors affected.
- Accounts. The number of accounts associated with those conversations.
- Created by. The Pendo user who created the tracked issue.
- Created time. The date and time the tracked issue was initially created.
Manage detection rules for a tracked issue
Detection rules define what Pendo looks for when identifying conversations that match a tracked issue. You can edit, add, or delete rules at any time, up to 10 rules for each tracked issue.
Tip: Rules work best when they're specific. Describe a concrete behavior or pattern rather than a general topic. For example, "Agent responds with 'no data available' when account data exists" rather than "data issues".
- Hover over the tracked issue you want to update, then select the Edit icon in the far-right of the table row.
- Adjust the issue name and description as needed, then select Next.
- To edit a rule, select the rule's text box and make your changes.
- To add a rule, select + Add rule, then enter the rule.
- To delete a rule, hover over the rule and select the X that appears in the far-right of the text box.
- Select Save to apply your changes.
Investigate an issue
You can investigate any emergent or tracked issue by selecting one of the metric values in its row:
- Ask Leo. If Ask Leo is turned on, select the Ask Leo icon next to the issue name to launch a guided prompt and ask questions to investigate the issue further. The Ask Leo option becomes available when an issue has one or more prompts or conversations.
- Instances or Conversations. Opens a side panel that lists related conversations. From the panel, select View conversation to review its full context and see where the issue occurred. You can also select additional Actions from the issue instance:
- Create ticket copies the issue details to your clipboard so you can then paste them in a project tracking tool like Jira or Linear. You can review and edit the prefilled details, including issue context, evidence, and metadata, before copying.
- Watch replay shows visual replays of visitors interacting with your agent if you have Session Replay added to your subscription and turned on in the relevant application.
- Visitors or Accounts. Opens a list of affected Visitor or Account IDs. From here, you can search for specific records or go to the corresponding visitor or account details page.
When you open a conversation linked to an issue, Pendo opens the relevant part of the interaction. This helps you understand what caused the issue and how the agent responded. From there, you can share specific examples with engineering or support teams, or update agent prompts, workflows, or guardrails to address the issue.