Brief Overview

AI Insights answers questions about the data in your connected WalkMe systems. How you ask shapes what you get back: a question with enough detail returns a sharper answer than a broad one, and usually saves you a round of follow-ups.

This article covers how to phrase your questions, how to check the answers you get, and what to do when a result looks wrong.

For an overview of AI Insights, how to access it, and the full list of limitations, refer to AI Insights.

Ask Better Questions

Most of what you can do to get a better answer happens before you send the question.

Include the details that matter

The more your question says, the closer the answer lands to what you're looking for.

Details worth naming:

• System and environment — which system you mean, and whether you want production or test data
• Timeframe — “in the last 30 days” or “since June 2026,” rather than leaving the range open
• Content type — Smart Walk-Thrus, Launchers, onboarding tasks, and so on
• Audience — a region, a department, or a user segment
• What you're measuring — completion, errors, engagement, or drop-off

Broad: “What content is broken?”

Specific: “Which Smart Walk-Thrus in my Salesforce production environment had the most errors in the last 30 days?”

Both work. The second one gets you to an answer you can act on in one step.

Say what you need the answer for

Telling AI Insights what you're trying to decide helps it pick the right shape for the answer — a ranked list you can work through, a comparison you can present, or a single number you want to track.

For example: “I'm preparing a quarterly review for our CIO. Summarize onboarding completion over the last three months and tell me what changed.”

Know what AI Insights can't answer

Some questions are outside what AI Insights can reach. Knowing which ones saves you a round trip. It can't:

• Create, change, or delete WalkMe content
• Access tracked events or saved filters
• Access changes made in the WalkMe Editor, such as the last publish date or folder location
• Read what's on your screen
• Compare your data with data from other customers or organizations
• Access systems or accounts you don't have permission to view

When a question is out of scope, AI Insights tells you directly instead of guessing, and usually suggests a related question it can answer.

Build on your question instead of starting over

AI Insights keeps context within a chat, so a follow-up doesn't need to repeat what you already said. Narrowing down over a few messages usually beats trying to write one perfect question:

1. Start broad to see what's there
2. Narrow to the system, segment, or timeframe worth a closer look
3. Ask for the answer in the form you need it, such as a table or a chart

Every answer comes with a suggested follow-up question. It's a shortcut, not a limit — ask your own question instead whenever you have one.

Check What You Get Back

An answer is a starting point for the next question, not the last word.

Read the assumptions before you act

Every answer states the assumptions behind it, such as the timeframe and which systems are included. Read those first — they're the fastest way to tell whether an answer is about the question you meant to ask.

If an assumption is wrong, say so in your next message. AI Insights answers again using your correction, so you don't need to start over.

Say so when a result looks wrong

If a number surprises you, tell AI Insights. Questioning an answer is a normal part of the conversation, and often the fastest way to a better one. Try:

• “That engagement rate is much lower than I expected — what's driving it?”
• “Are you including our test environment in this?”
• “Show me the breakdown behind that number.”

AI Insights flags anomalies it notices and phrases inferences as possibilities rather than facts, so the wording of an answer tells you how firm it is. And if you know about something the data doesn't explain — a release, a holiday, a migration — mention it and ask AI Insights to take it into account.

Keep What's Useful

Two things are worth doing once you have an answer you're happy with.

1. Save the results you want to keep

Chat history lets you go back to an earlier chat, but anything you'll reuse is better saved out of the chat:

• Save data results as a widget on a new or existing Insights dashboard, where anyone with dashboard access can see it
• Copy tables, charts, and text straight from the chat

A saved widget keeps updating with your data, so it's the better choice for anything you want to track over time rather than read once.

2. Rate the answers you get

Each answer can be rated. Use it — positive when an answer was useful, negative when it missed. Feedback is the most direct signal we have for improving how AI Insights answers.