A few years ago, screens were designed, prototyped, edited. Now we’re moving into design systems that design themselves, think about Google’s components, detailed and probably took ages, now with AI we can made those in seconds. We’re getting to the point where we can create designs faster. AI is changing how, what and who can design. Steps are becoming more efficient and easier to do. Non designers can now generate UI’s.
Dan Saffer a professor of practice at CMU HCI Institute said this about AI:
https://findingourway.design/2026/03/14/68-ai-and-design-fundamentals-and-the-future-ft-dan-saffer/
He talks about these 4 layers:
Imagine an AI that monitors and covers other AI’s. think of it as a chief of staff watching the other staff, the interface isn’t about doing tasks, it’s about helping users decide what to delegate versus what to keep. Stuff like this will become key and be more mainstream with technology AI is going to help us to enhance our systems.
Your devices configure themselves around what you’re doing, AI will learn your patterns, anticipate them instead of react to them. It would be great if AI reminded you of Tabs you had open or ask if you want to update around 4.23am as your always not using the device for at least an hour. Automation is great, but there should be caution for it as it shouldn’t control what you do.
The current smart home landscape is downloading a million apps on your phone, And you know in five years that will break and fall over, and whenever you have to change your phone it becomes a real chore to add them all back. The promise of smart homes is seamless but there is a bit of friction to get there. We need intent over commands, maybe instead of saying to turn the light on, the light should know your in the room and turn on. It may be smart enough to know those things, but they need to be reversible, you need control.
Invisible AI needs Trust to learn, help us and innovate. And right now that is hard to do due to ongoing development, the media spreading fear mongering stories, etc. This is something to consider, who is listening to us? It’s the same with Alexa, it’s not listening until you say it’s name, but how does it hear it’s name if it’s not listening? (It probably blocks out everything until they hear the vowels 🤷♂️)
Temporal AI, ephemeral interfaces, and ambient interfaces represent a shift toward AI that is proactive, reliable, context-aware, and disappearing. It ensures AI agents can run long-lasting tasks (days, months) reliably, even if systems crash, by treating AI interactions as durable workflows.
AI-generated UI elements that appear instantly on-demand to fulfil a specific task and vanish when finished, reducing screen clutter and cognitive load. Ambient Interfaces are AI interfaces that are embedded passively into the environment, analysing data and performing actions without active, constant user input. Together, these concepts move computing away from static applications towards fluid, invisible assistance that anticipates user needs and works reliably in the background.
AI that serves us the right document, suggests a phrase is helpful, this is Just in Time assistance. The process of figuring out is really important. The current timeline of AI is you can ask for something to be done and it’ll do it, no thought. There is value in thinking of the problem. If you just use AI generators to make interfaces you’re not really learning anything. Automation is useful until it isn’t.
Interfaces that will change per user, a type of artificial intelligence that continuously learns and updates its models in real-time, allowing it to adapt to new data, changing environments, and user behaviours without needing to be retrained. Unlike traditional static AI, it uses feedback loops to evolve its performance, making it highly effective for dynamic, unpredictable scenarios