Ep 201: When Trust Becomes the Interface | Heidi Chow

Every AI feature makes a promise it can't fully keep: trust me, even when I'm wrong. Users rarely see a confidence score. They never see a training set. All they get is the interface in front of them.

In this episode of UX Evolved, host Eric Ellis sits down with Heidi Chow — a product manager by way of UX design — to unpack what trust in AI actually looks like at the interface level. Heidi has spent her career designing for the moments users have to decide whether to believe what a system is telling them: from an AI-powered feature that scanned users' teeth to confirm treatment fit, to her current work building AI tools that ground answers in real source material instead of asking users to take them on faith.

Eric and Heidi get tactical about the patterns that build (or break) user confidence in AI-driven products: the copy choices that signal certainty versus possibility, when and how to bring a human into the loop, what to show a user versus what to keep behind the scenes, and why designing trust in after the fact rarely works as well as designing it in from day one.

Chapters:
00:00 – Intro: AI trust from an experience angle, not just a tech angle
01:08 – Welcome & background: from recruiter to anthropology to UX bootcamp
04:10 – Heuristics that still hold up in the AI era: why feedback and transparency matter
08:00 – Case study: the Byte AI aligner fit scan
09:20 – The "may" vs. "can" distinction — subtle copy, real trust implications
16:08 – Human-in-the-loop: opt-in checks, user agency, and avoiding "AI slop"
18:01 – Confidence scores: what clinicians saw vs. what users saw
21:17 – Legal, regulatory, and copy decisions in a health context
23:12 – Testing AI-driven copy: evaluative research and A/B reactions
24:53 – Building Prequel: surfacing compliance requirements with AI agents
26:58 – Founder life: judging AI output quality without an expert safety net
29:10 – Can trust be retrofitted into an "AI-first" product?
29:46 – The "layer cake" model of trust: UI, data, and system-prompt layers
32:18 – Where UX designers still add value as UI work gets commoditized
33:22 – Vibe-coding in Claude Design: a product-detail-page example
34:19 – Closing thought: AI gets you 80% there — humanity closes the gap
35:47 – Final question: what would you redesign in the universe?

UX Evolved is a show about how AI is reshaping the practice of user experience — hosted by Eric Ellis. Guests are practitioners who've lived the trade-offs firsthand — designers, PMs, and founders building real AI-powered products.