Designing agents that work requires "ontological redesign" — intent capture is the new UX¶
Insight: Building real AI agents requires "ontological redesign" — defining entities, relationships, and intents around customer outcomes, not internal structures. Intent capture is the new UX: products that succeed will anticipate user context, detect discontent, and adapt autonomously. Most organizations are trying to graft AI onto brittle systems built for predictability, which fundamentally doesn't work for probabilistic, adaptive agentic systems.
Detail: Key distinctions: building bots (easy, just API wiring) vs. building agents (hard — requires long-term memory, adaptive interfaces, failure recovery). "If your agent never fails visibly, it's not thinking deeply enough." Real agents need distributed cognition with reasoning and retrying within boundaries. The shift from "pixel-perfect to outcome-obsessed" — designers and product teams must move from delivery output obsession to customer impact obsession. References Kwame Nyanning's "Agentics" book as the field guide.