AI UX & Product Design
Interfaces people understand, trust, and keep using
Design AI features people actually adopt: choosing the right interface beyond the chat box, designing for uncertainty and errors, perceived speed, user control, onboarding, feedback, trust and disclosure, accessibility, and rapid prototyping.
- Lessons
- 9
- Reading time
- 88 min
- Level
- Intermediate
- Access
- Full course
Lessons in this module
- 8.110 min
Principles of AI UX
What changes when the software is probabilistic: set expectations, show the work, make correction cheap, and keep the human in charge of outcomes.
- 8.211 min
Beyond the Chat Box: Choosing the Right Interface
Chat is one interface among many. When to use inline actions, side panels, smart defaults, background automation, or no visible AI at all.
- 8.310 min
Designing for Uncertainty and Errors
Communicate confidence honestly, fail gracefully, show sources, and design the moments when the AI is wrong — because it will be.
- 8.410 min
Latency UX: Streaming, Progress, and Perceived Speed
Make slow models feel fast: streaming, skeletons, staged results, optimistic UI, and honest progress for tasks that take minutes.
- 8.59 min
Keeping Users in Control: Edit, Undo, Regenerate
Review-before-apply, granular accept and reject, version history, and undo — the controls that turn AI suggestions into something users rely on.
- 8.69 min
Onboarding, Empty States, and Suggested Prompts
Solve the blank-box problem: teach capabilities in context, suggest starting points, and help users form an accurate model of what the AI can do.
- 8.79 min
Collecting Feedback That Improves Your Product
Thumbs, edits, and implicit signals — what to collect, how to store it with the trace, and how to turn it into eval cases and fixes.
- 8.810 min
Accessible AI Interfaces
Streaming text and screen readers, keyboard flows, alt text generation, plain-language output, and making voice and visual features inclusive.
- 8.910 min
Prototyping AI Features Fast
Test whether an AI feature is worth building in days: prompt prototypes, Wizard-of-Oz tests, and a prototype-to-production checklist.
About this module
AI UX design is the work of shaping how people experience a feature built on a language model: where it appears, what it shows while it works, how it communicates what it knows and doesn't, and what users can do when it's wrong. It matters because AI output behaves differently from ordinary software. The same request can produce different answers, a wrong answer looks just as polished as a right one, and what the feature can do is often hidden behind a text box.
Many AI features fail on design rather than on the model. A capable model wrapped in a chat panel nobody opens, a summary with no way to check its sources, a draft that can only be accepted or thrown away, or a spinner that runs for twenty seconds will all lose users. The same model in a pre-filled field, with visible sources, granular controls, and honest progress, often becomes something people rely on every day.
This module covers the interface side of building AI products. You'll learn principles for designing around a probabilistic component, how to choose between chat, inline actions, smart defaults, and background automation, and how to design for uncertainty and errors without misleading confidence scores. You'll make slow models feel fast with streaming and staged results, build edit, undo, and regenerate controls, solve the blank-box problem with contextual suggested prompts, and collect feedback that turns into eval cases and fixes.
It also covers accessibility for AI interfaces, including streaming text with screen readers and AI-generated alt text, and how to prototype an AI feature in days with prompt prototypes and Wizard-of-Oz tests. Examples use TypeScript, React, and a running help-desk app, and every pattern is aimed at features you can ship.
Questions about AI UX & Product Design
What is AI UX design?
AI UX design is designing the experience of features built on AI models, such as chat assistants, drafted replies, summaries, and suggestions. It deals with problems ordinary UX doesn't face, like outputs that vary between runs and wrong answers that look confident. Good AI UX sets accurate expectations, shows evidence, and makes correcting the AI cheap.
Is a chatbot always the best interface for an AI feature?
No. Chat suits open-ended, exploratory tasks, but it asks users to know what to ask, type it out, and carry the result back to where they need it. Narrow, frequent tasks usually work better as inline actions, pre-filled fields, or background automation that leaves results waiting.
Should I show users an AI confidence score?
Usually not. Confidence numbers a model reports about itself are often poorly calibrated, and users can't act on a percentage. It works better to compute concrete signals in code, such as whether quoted sources actually exist, and show specific flags like which sentence isn't supported.
How do I make AI responses feel faster?
Stream text that people read, show skeletons shaped like the result, and send fast partial results such as labels before the slower full output. Pre-compute outputs on events when users usually view them, and for long jobs show progress in real units with an option to leave and be notified.
How should I collect user feedback on AI features?
Combine a simple rating with specific reason options, and capture implicit signals such as edits, regenerations, and discards for every output. Store all feedback linked to the exact generation, including the model and prompt version, then review it regularly and turn representative failures into eval cases.