After a fresh run of AI/design engagements, I’m revisiting the principles I shipped with a decade ago. The surface area changed yet the job of design hasn’t: make decisions clearer, safer, and faster for real people. Here’s my updated, field-tested list and here’s to another decade of human centered design.
Use AI only when it unlocks value you can’t reach with heuristics or simple controls.
If a simpler pattern wins, ship the simpler pattern.
Right-size autonomy to the stakes to keep people in charge.
Let them choose, review, approve, undo, and take over when automation stumbles. Be proactive where errors are cheap and reversible; slow down where stakes are higher.
Be explicit about data & limits.
Treat data quality as product quality — plan it, monitor it, maintain it.
Close the feedback loop so users see how their input changes outcomes.
Don’t just say “thanks.” Show what will change, when, and how. People are far more likely to contribute again when feedback turns into visible progress.
Design recovery, not just success.
Your design is as strong as its weakest link. AI systems misfire. Build the product so people can revert to a safe default.
Lead with benefits, not buzzwords.
People pay for time saved, risk reduced, and confidence gained.
