It seems obvious now AI is as advanced as it is that voice interaction should feature more strongly. Scanning and reading are quicker for long-form content, but where the result I want is a short answer, it would be as quick for AI to tell me.
The other obvious idea is that it would be great if screens could exist via Augmented Reality. There are ways of doing this now, but they are still cumbersome. Maybe that is where the breakthrough is needed.