Henry Modisett, Head of Design at Perplexity
- “technological innovation is outpacing our ability to apply it”.
- We’re no longer constrained by AI model capabilities at the current moment: there are so many unsolved product/design problems with existing capabilities, and that delta will just keep getting bigger.
- The next big challenge in AI-driven design will be walking the tradeoff between being opinionated/vertical and being adaptable/horizontal.
- More traditional UI design centers around well-crafted visual languages: designers use their expertise and research to determine the best way to deliver cohesive experiences to users.
- Future agentic AI applications will enable a lot more just-in-time UI that gives what users want to see, and adapts the experience to whatever they ask for. (This is a core guiding principle in Henry’s current work at Perplexity.)
- Horizontal applications can start as verticals, then expand as people unlock more capabilities. For example, designers at Perplexity may craft a specific fantasy-football interface for a sports fan; that sports fan might then try using Perplexity for other things as well, like searching for cooking recipes.
- Open question: will this manifest as extensions of what already exists in text-based chats (like how Gemini can personalize the first half of a Google search page), or will layout-based adaptation start happening soon (like changing your more frequently-clicked buttons to be bigger)?
- Giving AI personality can be a good way to connect to users, but becomes obnoxious/creepy when it goes too far.
- Distinguishing AI applications as tools versus companions: people don’t want to talk to their car or toothbrush, but see chatbots as friends/therapists/conversationalists/coworkers.
- A single application like Perplexity or Claude can be one or the other at any given time for any given person, swapping between research mode and therapist mode at will.