Analyze product usage with Leo (beta)

Leo is an AI-powered conversational assistant designed to help you explore product data through a natural language interface. By asking questions about usage trends, behavioral patterns, and feedback themes, you can gain structured insights from Pendo objects such as accounts, segments, Pages, Features, and associated metadata.

This article explains what Leo is, when to use it, and how it works.

Prerequisites

To use Leo:

• A subscription admin must enable Leo in Settings > Subscription settings > AI access.
• You must access Pendo through either the US (app.pendo.io) or EU (app.eu.pendo.io) data center environments. Support for other regions is planned.

Use cases

Leo is Pendo's built-in AI experience, available inside the Pendo product. If you want to access Pendo data from an external AI tool, like Claude, Cursor, or ChatGPT, see Connect to the Pendo MCP server.

You can ask Leo questions about adoption, retention, and health, like “Which Features are most used by new enterprise visitors this month?”, and get a written response based on your Pages, Features, Track Events, and metadata.

The conversational interface interprets your intent and provides responses you can review, refine, and use in your workflows. Leo is most helpful when you want to:

• Explore usage patterns without building manual reports.
• Summarize behavior by visitor or account.
• See how different cohorts interact with your product.
• Identify Pages and Features with low or high engagement.
• Analyze conversion and drop-off through any sequence of Pages, Features, and Track Events (funnel analysis).
• Analyze product feedback.
• Learn more about Pendo and how to use it.
• Use skills to follow a guided workflow for consistent results, such as preparing for a customer call or analyzing product health.

How Leo processes your query

Leo runs on large language models hosted by Pendo and uses an internal orchestration layer to route your prompts to the right data sources securely. Models from OpenAI and Google Gemini interpret your prompts and generate natural-language responses based on your Pendo data.

Data sources include:

• Product usage data, including Pages, Features, and Track Events.
• Structural data, including subscriptions, applications, and Product Areas.
• User and account data, including visitors, accounts, segments, and metadata.
• Qualitative data, including feedback topics and themes.
• Guide data, including guide views, guide elements clicked, and poll and survey responses.
• Pendo Help Center, including procedural, reference, and conceptual articles on all areas of the Pendo product. Leo cites the articles it draws from and provides links to them.

Note: We don’t use your data to train our AI models. For more information, see our Trust Center.

Automation and control

Leo doesn't modify your Pendo data.

Conversations you have with Leo are available under Previous chats, where each conversation contains any visualizations you’ve created. You can download those visualizations, but you can't currently save them as a report or dashboard widget. 

You can:

• View results and refine queries iteratively.
• Stop a session at any time.
• Use responses to support other workflows, like tagging or segmentation.

Ask Leo a question

After Leo is activated for your subscription, you can access it:

1. At the top of Pendo, select Ask Leo. This opens the full-page interface.

   

2. Alternatively, you can access Leo in-app by selecting the Ask Leo button, which is visible on various pages and clickable elements across the platform. This opens the right-hand panel and is context-aware: it knows which page you're in and which data point you're viewing. This is indicated through the Context pill at the top of your prompt box. Access Leo through the following points:

   • Individual Page details.
   • Individual Feature details.
   • Track Event details.
   • Visitor and Account list pages.
   • Individual visitor and account pages.
   • Key metrics views.
   • Agent Analytics use cases, issues, and conversations.

   

3. Enter a question about your product data. You can ask about adoption, retention, feedback, or health, and get responses based on Pages, Features, Track Events, qualitative data, and metadata.

Tip: For best results, start with a focused question about a specific account, segment, or set of Pages or Features, then use follow‑up prompts to explore what you see.

4. Press Enter or return on your keyboard, or select the send icon, to submit your question.
5. Continue chatting to refine your question and ask follow-up questions. You can also reference specific Pendo items in your question to help Leo respond more accurately.

Reference Pendo items

Referencing specific Pendo items in your question makes your intent clearer and reduces the chance of ambiguity. This helps avoid unnecessary back-and-forth and ensures Leo uses the exact data you're asking about.

To insert items inline, including accounts, applications, Features, Pages, Product Areas, segments, Track Events, and visitors:

1. Select Insert in the bottom-left of the text box, or enter @.
2. Begin typing to search the available items.
   
   
    
3. Use the arrow keys to navigate the options, then press Enter or return on your keyboard to insert the item in your text box.
4. Press Enter or return, or select the send icon, to submit your question. Chat back-and-forth to clarify intent and ask additional questions.

Use visualizations

When Leo returns a chart or table, you can customize the view, copy the result, or export it to use in other tools. Use the icons in the top-right corner to:

• Modify the visualization. Switch chart type or customize table columns.
• Export. Download tables as CSVs, charts as PNGs, and documents as PDFs.
• Split your view. Open a side-by-side panel with the visualization on one side and the chat on the other.

Stop a query

You can cancel a query in progress if it's taking too long or if you want to change direction.

To cancel a request while it's processing, select Stop in the bottom-right corner of the text box.

Manage conversations

Leo automatically saves each conversation. In the top-right corner of the conversation, you can:

• Select New conversation to start a new thread. We recommend starting a new conversation for each topic.
• Select Previous chats to search and reopen past conversations.

Skills

Skills are guided prompts that invoke a specific workflow. Each skill helps you accomplish a defined task, such as preparing for a customer call or analyzing product health.

You can invoke a skill in three ways:

• Select a skill tile from the homepage. Some tiles open Leo with a pre-written prompt.
• Call the skill using the slash command in any Leo conversation. Type / into the prompt box to open a filterable list of skills.
• Request what you want in natural language in any Leo conversation.

After you start a skill, Leo prompts you to specify the applications, accounts, Pages, or Features you want it to use.

The following skills and corresponding slash commands are available:

• Customer call prep
• Account review
• Product health
• Recent activity
• Recent pages
• Power users
• Feature adoption
• Drive adoption
• Feature announcement guide
• Onboarding guide
• Opportunity validation
• Product discovery
• Activity decrease

For more information about what each skill accomplishes, see the descriptions on the tiles in the homepage.

Skill stability

Customer call prep and Account review are the most reliable skills. They follow a predefined sequence of steps to produce an account health report.

All other skills are labeled Experimental in the UI. Experimental skills are still being refined, and their workflows or outputs may change over time.

Feedback analysis

You can use Leo to gather insights from feedback sources created within Pendo or connected from external tools. This helps you identify trends, themes, and opportunities from customer feedback.

What you can do with feedback depends on whether you have a paid Listen subscription. For more information about Listen, see Overview of Pendo Listen.

With a Listen subscription, you can:

• Analyze feedback from all connected sources without any date restrictions.
• Generate feedback reports and export them.

Without a Listen subscription, you can access feedback analysis in Leo with the following limitations:

• You can only analyze feedback data from the last 30 days.
• You can only interact with feedback data through Leo conversations. Listen features such as feedback views, ideas, idea tests, and portals aren't available.

Set up feedback analysis without a Listen subscription

If you don't have a paid Listen subscription, you can configure feedback analysis with Leo by following these steps:

1. Go to Listen (Preview) in the main navigation menu.
2. Follow the onboarding instructions to set up your company profile and connect your feedback sources.
   • Company profile. Provide and verify your organization's product offerings, along with any additional business context, to help AI better understand your feedback.
   • Feedback sources. Connect Pendo-built sources, such as surveys and guide poll responses, or external sources, such as Zendesk, Salesforce, and call transcription tools.
3. Open Leo to analyze your connected feedback sources.

Best practices and troubleshooting

Leo's responses depend on how your Pendo data is structured and how you frame your questions. For the most reliable results:

• Start with focused questions. Ask about one account, segment, or set of Pages or Features at a time, then use follow-up prompts to explore related patterns.
• Use shorter date ranges. Long date ranges or complex queries (for example, combining usage, feedback, and metadata) can take longer to process or return incomplete answers.
• Review results carefully. Because Leo uses generative AI, always verify responses before relying on them for decisions.

Troubleshoot errors and timeouts

You might occasionally see an error message when you submit a question. These errors don’t affect your underlying analytics data, but they can interrupt a response, especially for complex questions or long date ranges.

If this happens:

1. Try rephrasing your question to focus on a smaller scope (for example, one segment or a shorter date range).
2. If the problem continues, share a screenshot of the error message and the conversation URL with your Pendo account representative or our support team so we can investigate in more detail.