A framework for matching AI interface modality to user intent and context — use a Task Audit (observe physical, social, cognitive constraints) and Input/Output Alignment Matrix to pick the right combination (voice for hands-busy, visual dashboards for analysis, alerts for monitoring). The key: AI fails if delivered through a lazy text interface; modality choices must be grounded in real-world observation, not convention
NNG: Stop Reporting UX Activity and Report Business Outcomes
An NN/g guide on reporting UX impact: stop reporting activity ("24 interviews") or UX metrics (SUS scores) — connect your work to business outcomes leaders care about: revenue, cost, risk, speed, retention. Bridge upstream UX metrics (task success, errors) to downstream business data (support volume, conversion, churn) to move UX from cost center to value driver
Prototyping: Your Interface Has a Tone. And Sometimes It Blames You
Interfaces often blame users through judgmental language ("invalid entry") — assuming a fictional ideal user who is patient and adaptable, causing real users to internalize failure as their own. The solution: clear, non-punitive language designed for people at the margins (curb-cut effect) works better for everyone, reducing friction and blame
Case Study: The Hidden Cost of Forcing Users to Decide
A case study on redesigning an e-commerce quiz (21 steps → 9): the core problem was forcing users to declare certainty (customization) instead of inferring intent (personalization) — ambiguity was treated as a failure state. The solution: conversational AI that treats uncertainty as usable input, asks targeted follow-ups only when needed, and shares the work of sensemaking
@uxdigest
