Best practices for writing prompts in Leo

Last updated:

This article explains best practices for writing prompts when using Leo, Pendo's AI assistant. The clearer your intent, scope, and desired output, the better Leo can understand your request and return useful results.

Prompt guidelines

  • Be specific about scope. Name the feature, segment, or time range. "Users on Pro plan in the last 30 days" beats "users recently".
  • State your intent. Start with what decision you're making. Leo routes better when it knows the why, not just the what.
  • Request a format. Ask for a table, chart, or ranked list if you want structured output — Leo will match your format signal.
  • Chain your question. Use "then" to stack steps: pull the metric, then explain what's driving it, then suggest next actions.
  • Anchor to an outcome. End with the decision or action: "…so I can decide whether to invest more in this feature".

How to structure your prompt

Write your prompts in a way that is readable by AI. Below is an example of a clear structure with the relevant information Leo requires to fulfil your request:

[intent] + [scope/filter] + [entity] + [output format] + [decision context]

Example prompts

Quick-fire prompts to get started

Use these prompts for fast, targeted answers to specific questions. They're ideal for preparing for a standup, checking in on a metric before a customer conversation, or quickly confirming what the data shows before making a call.

Feature usage snapshot

Use when you need a quick usage count and trend line for a specific feature.

How many unique visitors used [Feature Name] in the last 30 days? Break it down by weekly trend and show as a bar chart.

Release adoption check

Use when you want to measure the impact of a release by comparing behaviour before and after launch.

Show me adoption of [Feature X] since we released it on [date]. Compare the 4 weeks before and after. Visualise as a before/after chart.

Account NPS lookup

Use when you want to check how a specific account feels about the product, and whether they've left any feedback.

What NPS score did [Account name] submit most recently, and when? Include any verbatim feedback if available.

Funnel drop-off

Use when you want to find where users are dropping out before completing a key action.

Show me the conversion funnel from [Page] to [Feature] in the last 90 days — what's the drop-off rate?

Prompts for deep analysis

Use these prompts when you need to go beyond the numbers and understand the story behind a metric. They're best for investigating unexpected changes, identifying friction in the user journey, or gathering evidence before a product decision.

Explain a metric drop

Use when a metric has changed and you need to understand what's driving it.

Our weekly active users for [Feature] dropped. What's driving it? Check for any frustration signals, relevant feedback themes, and suggest 3 hypotheses with supporting data.

Frustration analysis

Use when you want to identify which areas of the product are causing the most user friction.

Pull rage clicks and error clicks for [Product Area] over the last 14 days. What pages are getting the most frustration signals?

Feedback theme extraction

Use when you want to understand what users are saying about a feature at scale.

Analyse feedback submitted in the last 60 days mentioning [Feature]. Cluster into topics. Show as a ranked table.

Session replay sample

Use when you want to watch real user behaviour in a specific area of the product.

Show me 5 replays from users who visited Leo homepage but not Leo chat/conversation in the last 7 days.

Prompts to generate first drafts quickly

Leo can produce working first drafts as well as surface data. Use these prompts to skip the blank page and get a starting point for in-app content, product documents, or development specs.

Segment creation

Use when you need to identify a specific group of users based on their behaviour.

Create a segment of users who have visited [Feature] at least 3 times in the last 30 days but have not completed [Key Action].

In-app guide draft

Use when you want to create onboarding or guidance content for a specific user workflow.

Draft a 3-step tooltip guide to help new users complete [Workflow] for the first time. Tone should be concise and action-oriented.

One-page PRD

Use when you need a structured starting point for a new feature idea.

Write a one-page PRD for [Feature Idea]. Include: problem statement, target user, proposed solution, success metrics, and open questions. Use the standard Pendo PRD format.

Draft instructions for your coding agent

Use when you want to turn a Leo-generated concept into a brief for your coding agent.

Based on the above, I want to build [Feature Idea]. Generate a prompt for [Coding Agent] to build a draft prototype.

Was this article helpful?
0 out of 0 found this helpful