Many AI products project confidence, jumping from what a user says to an answer without showing how they got there. Assuming the AI might have misunderstood, and letting users correct it, builds trust. Understanding becomes something shaped together, not something the AI achieves alone.
Before leaving my last role, I fed two years of work into an AI tool, creating a chat interface that lets people keep asking questions long after I've gone - which, depending on your perspective, is either considerate or deeply unsettling.
Experiments aren't about shipping features. They're about showing what's possible, making mistakes and celebrating the odd agility that only a non-profit can pull off. Not every idea is good, and that's the whole point.
Whilst monitoring traffic to popular open-access legal content, it looked like our audience was shrinking. But a deeper dive into six months of analytics revealed a surprising truth - AI was amplifying the legal content beyond our website. So whilst visitors are declining, influence could be increasing.
Traveling through Oman, I noticed how regulated bilingual shop signs achieve a rare thing: perfect clarity. Back home, our signs trade meaning for mystery - a reminder that design reflects culture as much as language.
A reminder that being “good with technology” often means navigating the irrational logic of our digital age — and learning how to design better systems from its flaws.
Recruiting participants exposes the hidden complexity of user research, where logistics, trust and timing often determine success more than methodology.