A 1980s photocopier study and a grocery-store robot point to the same lesson: people interact with automation based on what they think it can do, not what it was designed to do. In the agentic era, managing expectations and making system intent visible may matter more than adding another interface

AI prototyping makes it possible to test complex interactions — filters, dashboards, conversational AI — much earlier with realistic working prototypes. In one Ramp case, the team found a confusing edit-tracking issue that static screens simply couldn’t reveal

Coupons, vouchers, gift cards, and credit may look interchangeable in an interface, but they follow different rules and user expectations. Treating them as one pattern can create surprisingly confusing checkout experiences

AI makes UX testing less deterministic: the same input can produce different outputs, errors, and paths. So instead of checking one “correct” flow, researchers increasingly need to test ranges of behaviour, edge cases, and how gracefully the system fails

A case study on designing philanthropy around distrust rather than trying to “fix” it with better UI. The strongest insight: money may flow upward to institutions, while trust flows sideways through local communities — so the service had to follow the trust, not the org chart