Beyond the Happy Path: Josh Clark & Veronika Kindred on “Sentient Design”
What happens when a 30-year UX veteran teams up with a Gen Z designer to rethink interfaces in the age of AI? In this Commit & Push episode, host Damien Filiatrault talks with Josh Clark and Veronika Kindred of Big Medium about their forthcoming book, Sentient Design—a framework for building adaptive, mindful interfaces that collaborate with users instead of dictating to them.
Listen to the episode:
Cross-Generational Collaboration, On Purpose
Josh brings decades of product and UX practice; Veronika brings a native AI mindset. That tension is the point. She questions “best practices” that were forged for static, reactive UIs; he brings context about cycles of AI hype and winter. The result: a practical way to design with machine intelligence—not just for it.
What “Sentient Design” Really Means
No, it’s not about conscious machines. “Sentient” here means interfaces that are seemingly aware of context and intent, so they can adapt in real time. Designers share agency with machine intelligence to compose experiences “in the moment.”
Core principles
- Contextually aware & radically adaptive: UI responds to intent, not just clicks and forms.
- Collaborative (co-intelligence): The system acts like a helpful partner.
- Deferential: Delegate decisions; don’t abdicate them. Suggest rather than impose.
- Ambient: Steps forward when needed; fades back when not.
From Reactive to Proactive: New Interface Postures
- Chat (and beyond): Conversation is one posture—but intelligence can also rearrange screens, not just answers.
- Bespoke UI: Components from a trusted design system rearrange on demand (think Salesforce’s “generative canvas” concept pulling exactly what you need before a meeting).
- NPC pattern: The system participates as “another user” (e.g., a cursor that leaves comments in your canvas or doc), in line with your workflow—not in a separate chat pane.
Guardrails, Not Just “Happy Paths”
Adaptive systems blow up the old notion of a single “happy path.” Clark & Kindred argue for defensive design:
- Design for failure: Set broad rules, contingencies, and graceful degradation (the escalator should become stairs).
- Keep an analog default: Smart features shouldn’t break the basics.
- Constrain the playground: Let AI choose within a small, vetted set of components, data sources, and actions.
- Transparency & reversibility: Show what changed, why, and let users undo.
- Observability: Log UI state and choices so teams can debug unique, per-user configurations.
A Practical Pattern You Can Use Now
Treat the LLM as an intent translator, not an answer oracle:
- User input/context → LLM
- LLM returns a structured JSON spec describing the best presentation pattern (e.g., quick filters, a map, a card feed) and required parameters.
- Your UI layer renders from that JSON using your design system; your data layer fetches truth from trusted sources.
Why this works:
- LLMs excel at interpreting intent; they’re brittle as “answer machines.”
- Separating intent → presentation from data truth minimizes hallucination risk.
- Sandboxing patterns keeps the experience consistent, branded, and testable.
When Adaptive UI Helps (and When It Hurts)
- Great fit: Wayfinding through dense enterprise tools, dashboards that must surface “what matters now,” complex assistants that need “productive humility” (“I took a pass at this—confirm or adjust?”).
- Risky without guardrails: Fully generated interfaces, free-form database actions, or anything that moves critical controls unexpectedly. Always provide a stable route back to the familiar.
What Teams Should Do Next
- Audit tasks for intent ambiguity. Where do users fight filters, forms, and menus to express what they want?
- Codify a small “adaptive kit.” Define 5–10 approved layout patterns your LLM can request via JSON.
- Instrument everything. Capture the chosen pattern, inputs, and outcomes for support and iteration.
- Practice “suggest, don’t impose.” Default to reversible suggestions; escalate to auto-apply only when low risk.
- Build literacy. Try these ideas on a non-critical flow to develop intuition about where adaptive patterns shine.
Learn More
Clark and Kindred are wrapping Sentient Design with Rosenfeld Media and share ongoing experiments in their newsletter, A Little Big Medium (bigmedium.com). They also teach workshops and partner with teams to prototype adaptive interfaces grounded in defensive design.
Bottom line: Sentient Design isn’t sci-fi. It’s a disciplined way to let intelligence reshape presentation—safely—so users get from intent to outcome with less friction and more trust.