AI as a design material: building intelligent product experiences

Executive overview

Most teams are using AI to speed up existing workflows — faster prototyping, faster research, faster delivery. The bigger opportunity is treating AI as a design material that enables entirely new kinds of experiences, not just faster versions of old ones.

Sentient design — embedding intelligence into the interface so it adapts to user intent in real time — is the next phase of product design.

The shift from tool to material

• ChatGPT's breakthrough wasn't generation — it was understanding intent without rigid syntax.
• Interaction design has always bent toward users: punch cards → CLI → GUI → touch → speech → AI.
• AI brings that arc to its logical end: interfaces that manifest the experience the user needs rather than making users search for it.
• "Sentient design" = intelligent interfaces that are context-aware, intent-aware, and adaptive.
• The authors have identified 14 new experience patterns that emerge when intelligence is woven into the interface layer.

AI accelerates the lean loop — it doesn't replace the human in it

• Customer research remains irreplaceable; AI simulations of users are not a substitute for real conversations.
• AI has compressed research timelines from weeks to days, removing the "research is too slow" objection.
• The result: fast and cheap are now a commodity — the only remaining differentiator is whether the product is good.
• Teams can now afford to take a beat, do the research, and build the right thing rather than racing to ship first.

Rethinking what LLMs are actually good at

• LLMs are like Leonardo DiCaprio in Catch Me If You Can: convincing performers who understand form and context, not facts.
• They understand what something should sound like and what the user wants — they don't know whether it's true.
• Practical implication: let LLMs drive the experience/interface layer; route to trusted systems for facts.
• Layering the LLM behind the interface (rather than exposing it directly) also reduces hallucination risk.
• "They are not answer machines or fact machines. They are dream machines."

Cross-disciplinary teams and the common problem space

• The system prompt now does triple duty: business requirements, design spec, and technical document.
• Engineers, designers, and product managers can all work in the same space and influence each other's disciplines.
• One client project stalled because only engineers owned the prompt; bringing in designers and product leads fixed interpretation problems immediately.
• AI is flattening technical barriers — marketers can now produce credible design prototypes — but expertise and perspective still matter.
• Google Gemini's design lead observed: "Maybe we're all becoming product managers" — not literally, but a shared product vision is now accessible to everyone.

Failure, psychological safety, and innovation culture

• Failing fast only works when there is psychological safety to fail at all; without it, small mistakes accumulate into large ones.
• Some organisations say they want to be industry-leading but actually want to be industry-standard — the two require very different risk tolerances.
• Clarity about where to innovate and where to consolidate is more useful than a vague aspiration to innovate everywhere.
• When AI doesn't work as expected, the right response is curiosity about what the material can actually do, not blame.
• Understanding AI's real capabilities requires hands-on experimentation; theory and hype are both unreliable guides.