The Playbook for Onboarding AI Features

AI interfaces are converging on very similar designs.

There's usually a single input field that "does everything", which unintentionally obscures their unique capabilities.

Many won't even remember that these features exist.

This is a step-by-step walkthrough of how to structure onboarding for AI features that require creative input from the user.

If you're considering implementing AI in the next few years, this could really help.

After adding a task, the empty state transitions into a list of your current tasks.

While this behaviour is incredibly common, it misses a trick.

No attempt has been made to encourage the discovery of new ideas—it's simply a list.

Which in this context is a problem, because the stickiness of this feature depends on Grok's ability to routinely deliver value.

Unlike a simple "bookmarking" feature (as an example, which would require no creative input), you may have created an uninteresting task as a test.

You shouldn't stop onboarding a creative feature like this, simply when the user has found one use for it.

Let's tie this up with a key event: the conversion to a paid plan.

After adding your second daily task, you'll see a paywall.

To add more tasks, you need to upgrade.

In other words, a paywall like this unintentionally deters experimentation.

It forces users to think smaller. It frames their usage as; "what are my most important 2 tasks".

And in that moment, if the user cannot think of 5 interesting examples, they won't upgrade—because there's no difficult trade-off being made.

Instead, consider how Grok could invert this 🖼 Framing—they could leverage this moment to inspire people with imaginative ways to use this feature.

The upsell would be far more effective if people were excited to try their ideas.