Anticipatory design is designing products that predict what a user needs and act on it before they ask, cutting the choices and steps a person has to make. The term came from Aaron Shapiro, who framed it as making decisions on the user's behalf to ease decision fatigue. Think predictive search, smart replies, auto-filled forms, and "your weekly mix." It goes a step past personalization and leans heavily on AI experience design to get the prediction right.
Use cases
Done well, it removes friction. Done badly it takes control away and unsettles people.
The helpful default: A booking form pre-fills your details and suggests the flight that fits your calendar, so a ten-field task becomes one tap. Friction gone, trust intact.
The trapped user: The system books, buys, or hides something based on a wrong prediction, with no obvious way to undo it. The lack of escape is the problem.
How it's used in practice
Anticipate the reversible stuff: Pre-fill fields, set smart defaults, send reminders. Save costly or hard-to-undo actions for an explicit confirm.
Always offer an override: Make undo and "not this" one tap away, so a wrong guess never traps the user.
Explain the why: Show the reason ("because you watched X"). Blind prediction reads as creepy; a stated reason reads as helpful.
Earn trust before automating: Start with suggestions and defaults. Move to acting alone only once the predictions prove reliable.
🪄
Pro-tip: The line between helpful and creepy is control and reversibility. Anticipate the boring, reversible things and ask before anything expensive or permanent. The failure mode is a wrong prediction the user can't escape. And if you can't explain why the system did something, don't let it act silently.
Challenges & limitations
It can erase agency: Over-automation makes people passive and can box them into a filter bubble, removing useful serendipity and choice.
It depends on data: Cold starts and sparse data make early predictions weak, and the tracking it relies on raises real privacy and consent stakes.
