Skip to main content

Guided Walkthroughs

Create hands-on software training with a voice or text coach, shared-screen progress detection, narrated screen-recording generation, verifiable steps, reference clips, and a floating learner panel.

Written by Justin

Guided Walkthroughs turn real software workflows into hands-on practice. The learner does the work in the actual application while a voice or text coach guides them and checks each step against their shared screen.

Walkthrough is a mode inside AI Conversations. Use it for software onboarding, standard operating procedures, and system training.

  • Guided Walkthrough: The learner works in the real application with a floating coach panel.

  • Video Simulation: The learner practices on clips and screenshots inside Mindsmith. Use this when learners can't access the real system.

Get access

Guided Walkthroughs are in closed beta. To turn Walkthrough mode on for your workspace, message Mindsmith support. Once a Walkthrough exists, any collaborator can edit it and every learner can take it.

Record screen is available to every author. Without Walkthrough access, a recording is attached as a regular video.

Create a Walkthrough from a screen recording

The fastest way to build one is to do the task once and narrate as you go. The Mindsmith Agent turns your narration and the on-screen changes into the project, coach, and steps.

  1. Create a standalone Conversation, or open the Mindsmith Agent in a lesson.

  2. Open the attachment menu and select Record screen.

  3. Choose your microphone, then the tab, window, or screen to record.

  4. Complete the task, explaining what each action is for and what success looks like.

  5. Stop, review, and select Use recording.

  6. Ask the Agent to turn the recording into a Guided Walkthrough, then review it in the Conversation editor.

Silent recordings give it much less to work with.

To start without a recording, open the Conversation editor, choose Walkthrough as the mode, and pick a template or build the steps yourself.

Configure the Walkthrough

  • Project: What the learner will complete and the outcome you want.

  • Location: The name and URL of the application. Learners get a Go to button that opens it, so link to the best starting page.

  • Introduction: What the learner will do, what access they need, and roughly how long it takes.

  • Modality: Voice & Text, Voice Only, or Text Only.

  • Coach: Display name, subtitle, voice, and how the coach behaves.

Walkthroughs are coached to completion, not scored. They have no scorecard or passing percentage.

Build the steps

Five to ten steps usually covers a workflow. Each step has:

  • Step title: A short action or outcome, such as "Create the return."

  • Success looks like: What the learner should see when they're done. Shown to the learner.

  • Coaching notes: Private. Where the controls are and how to help if the learner gets stuck.

  • Grading notes: Private. The visible evidence that proves the step is done.

  • Verified by: Screen checks the learner's shared screen. Transcript relies on what the learner says or types. Use Transcript for decisions, explanations, downloads, or work outside the shared screen.

  • Step resources: An image, link, copy action, or practice file the learner can open from the panel.

For screen-verified steps, describe the finished state, not the click. "The project name appears in the header and the unsaved indicator is gone" can be verified. "Click Save" can't. Don't depend on the exact values you typed during the recording, because learners will enter their own. A spinner or a closed dialog doesn't prove completion. If no reliable visual exists, use Transcript.

Test before sharing

Preview it as a learner, with the same starting URL, permissions, and data your learners will have. Check that:

  • The Go to link opens the right starting page.

  • Each screen check advances only when the finished state is visible.

  • Coaching notes help without narrating every click.

If a correct action doesn't advance a step, make its grading notes more concrete, or switch it to Transcript.

What learners experience

  1. The learner reviews the introduction and steps, selects Set up conversation, and chooses the tab or window to share.

  2. They select Open the coach panel. The floating panel stays on top while they work.

  3. In the panel, they select Go to to open the application, then Begin conversation.

  4. Screen-verified steps advance when the result appears. Transcript-verified steps advance through the conversation.

  5. The learner can pause, change what they're sharing, switch from voice to text, or end the Walkthrough.

  6. At the end they see Walkthrough complete with their total time, time per step, and a transcript.

When a learner asks where something is or says they're stuck, the coach shows a zoomed snapshot of their screen with the control highlighted.

The first start can take up to about two minutes while the coach starts up. Leaving and starting again begins a new attempt from step 1.

Requirements and privacy

  • Browser: Desktop Chrome, Edge, or Firefox. Safari and mobile browsers aren't supported.

  • Launch: The Walkthrough needs its own browser tab. In an LTI or web embed, learners are prompted to open it in a new tab. Walkthroughs don't work in SCORM or xAPI packages or static exports.

  • Access: Learners need their own access to the application and any practice data.

  • Recording: Sessions are recorded, including the shared screen, microphone, and coach, and authors can review them in Analytics. Ask learners to share only the tab the task needs.

Usage

Walkthroughs use increased Conversation credits from your learner Conversation allowance. Testing in the editor doesn't count, though it has daily limits. See AI Conversations for billing.

Did this answer your question?