Follow Along: in-app guidance versus static documentation

Most documentation tools stop at the document. For many tasks that is exactly right: a clear page with screenshots, read once, done. But there is a moment every support and onboarding team knows. The reader has the guide in one window and the application in another, alt-tabs between them, loses their place, and asks a colleague anyway.

Follow Along is our answer to that moment. This article explains the three levels of help a published Recordable guide can offer, what each one requires, and how to choose between them. The short version: it is one guide, and you fix a step once for all three.

Level 1: Read it

The reader opens the shared link and works through the steps. Reading is free and does not require a seat.

This fits tasks that people do occasionally, readers who already know the application, and anything used as a reference rather than a script. It is also the level that exports and translations serve: a PDF or a translated copy is still a document to read.

Writing one that holds up is its own craft; we set out a method in How to write an SOP from a screen recording.

The limit of this level is the alt-tab problem above. A long procedure in an unfamiliar application is hard to follow from a separate window, however well it is written.

Level 2: Follow Along, in the browser

With Follow Along, the browser extension keeps the guide beside the work. The current step stays in view and the next target on the page is highlighted. The person does the task themselves; for supported actions, completing the step advances the guide automatically, so they are not clicking Next between every action.

This fits an unfamiliar task, a long procedure, a new hire's first weeks, or a client's staff using a system an implementation consultant set up for them. It works where a document alone leaves people guessing which of three similar buttons is meant.

What it requires:

• An eligible browser workflow. Follow Along does not cover native desktop applications.
• An account and the extension for the person following.
• Follow Along enabled for that person. It is included for Creator seats; for readers it is a separate per-viewer option. The current terms are on the pricing page.

What it does not promise: if the screen has changed since the recording, the target may not be found. The reader still has the instructions and can flag the step, which brings us to the loop at the end of this article.

Level 3: A supervised AI run

A recorded workflow can also serve as the procedure for an AI run. The agent works from the same reviewed steps, and the run is supervised: you review what it plans to do, and it pauses before actions you could not undo and asks. This is the level for repetitive, well-bounded tasks where a human has already reviewed the guide and does not need to click through it again.

Two things it is not. It is not "autonomous"; supervision is the point, and our acceptable use policy reflects that. And it is not a promise that every application works or that runs never fail. Which plans include AI runs, and how much, is on the pricing page.

How to choose

Three questions usually settle it:

1. How often is the task done, and by whom? Rare tasks by experienced people: read it. Frequent tasks by new people: Follow Along. Frequent, bounded tasks by the same person over and over: consider a supervised run.
2. Where does the task live? Follow Along is for the browser. A desktop application can be recorded and read as a guide, but the in-app overlay does not apply there.
3. What happens if a step goes wrong? If the answer is "someone notices and fixes it later", any level works. If the answer is "money moves" or "a record is deleted", keep a human in the loop: read or follow, and let an AI run only pause and ask rather than proceed.

What stays the same across all three

There is one guide. When a screen changes, the author re-records or edits that one step and everyone with the link, at every level, has the current version. People following the guide can report an unclear or outdated step, the author sees it in a review inbox and resolves it. Guides do not update themselves, but a step is fixed once, not once per format.

That is the practical difference from static documentation. Not that the document goes away, but that the same reviewed steps can meet people where they are: on a page, beside the work, or running under their supervision.