Course

Designing AI UX (for uncertainty)

Traditional software is deterministic: one input, one right answer. AI isn't — it's probabilistic, occasionally wrong, and sounds equally sure either way. This short course is the UX playbook for that reality: how to design for output with no single right answer, communicate confidence and error honestly, replace the intimidating blank box with guided prompts, keep a human in the loop as autonomy grows, and give agents the approve / observe / undo controls users need to trust them. Six short lessons for PMs, designers, and founders — every screen drawn, no jargon.

6 lessonsIntermediate36 min

Start the course Get the app

Course lessons

Designing for probabilistic output

There's no single right answer — so offer variants, a Regenerate button, and editing, not one verdict.

›

Communicating confidence & error

AI sounds equally sure when right and wrong — so your UI must show sources, flag uncertainty, and make verifying easy.

›

Guided prompts vs the blank box

An empty input intimidates users into freezing — chips, examples, and templates give a running start.

›

Human-in-the-loop (Crawl-Walk-Run)

Turn autonomy up slowly — most AI features are a trustworthy workflow, not a free-roaming agent.

›

Agent UX (approve / observe / undo)

When an agent takes real actions, gate the irreversible, show its work, and make things undoable.

›

The AI UX playbook

Five moves for uncertain output — and the one principle that ties them together.

›

Take the full course

Learn it properly — tracked & personalized

Follow the whole course as a guided journey — sequential lessons, progress, streaks and a personalized path, all in the app.

Get the app