AI UX Patterns — User Transparency

Kathryn Grayson Nanz argues that transparency is the price of AI adoption: users won't integrate tools they can't see and understand, so the more transparent a feature is — about what it does, what data it can reach, and what it costs — the less hesitation users will have adopting it.

The piece breaks transparency into four practical UX patterns. Permissions must be more than yes/no: users need to see which applications the tool can access, in which ways, and approve each capability (read vs. send email, reference vs. delete tables, run scripts, search files) individually. Permissions are not one-and-done: a user may allow an action once under supervision but not permanently, grant one folder but not all, or want a notification each time an action fires.

Data collection has to be symmetric: a system that can "remember" must also be able to "forget," and the user needs final say — plus a history of every grant and the ability to revoke it at any time. Cost transparency means telling users, before they approve a plan, roughly how much time and money/tokens it will consume, so they can revise a request that exceeds their comfort. Finally, a visual signal for autonomous action — a banner, sidebar, or outline — should mark whenever the AI is driving independently, so users understand system state and don't unintentionally interrupt a long-running process.