A few years ago, taking an idea from a conversation to something you could actually click, test or put in front of a client took time. There were wireframes to work through, interfaces to design, assets to create and usually a developer somewhere in the process before the idea became real.

That gap has become incredibly small.

I’ve noticed it most in the way I work. An idea that might once have stayed as a sketch or a Figma concept can now become a working prototype surprisingly quickly. I can explore the interface, test how an interaction feels and sometimes build enough of the product to understand whether the original idea was even worth pursuing.

That has made me think a lot about what AI actually changes for designers.

There’s understandably a lot of conversation around whether AI makes design less valuable. My experience has been almost the opposite. When creating something becomes easier, deciding what should be created becomes much more important.

AI is very good at giving you options. It can generate layouts, imagery, interfaces, copy and code at a speed no designer could realistically compete with. But having more options doesn’t necessarily mean you’re closer to the right answer. In some cases, you’re simply creating more things to make decisions about.

That is where I’ve found experience starts to matter more.

A large part of design has always happened before anything appears on the screen. You’re trying to understand what the business needs, what the user needs, where those things conflict and which parts of the problem are actually worth solving. Then, once you start designing, you’re constantly making smaller decisions about hierarchy, interaction, typography, behaviour and what can be removed.

AI accelerates the execution of those decisions, but it doesn’t remove the need to make them.

From designing outputs to designing systems

The part of AI that interests me most isn’t image generation or getting a machine to produce another interface. It’s the possibility of designing systems that can continue producing good work.

I recently started exploring this through branded design tools. Instead of creating a visual treatment for a client and then handing over instructions explaining how to recreate it, I began looking at whether the treatment itself could become a tool.

The designer still establishes the visual language: the typography, colours, image treatment, proportions, constraints and behaviour. But instead of expecting someone else to interpret those decisions correctly every time, they can be built directly into the experience.

That changes the relationship between a brand and its guidelines.

If a brand only uses four colour combinations, why give someone an unrestricted colour picker? If imagery always follows a particular treatment, why make someone rebuild that treatment manually? If spacing follows a system, why allow arbitrary values?

Suddenly the designer isn’t just producing the final asset. They’re defining the environment in which future assets get made.

This is where I see AI becoming genuinely useful for design. Not because it gives us unlimited possibilities, but because we can decide which possibilities should exist in the first place.

Constraints become more valuable, not less

There’s something slightly contradictory about working with generative technology. AI offers almost infinite variation, while good design often comes from deliberately reducing variation.

A strong identity doesn’t need twenty typefaces and fifty colours. A good interface doesn’t need every possible component. A useful product doesn’t need every feature someone can imagine.

It needs enough.

That makes constraints incredibly important. The challenge is deciding which parts of a system should remain flexible and which should never change.

This is already familiar territory for designers. Brand guidelines, component libraries and design systems have always been ways of creating consistency without designing every possible outcome in advance. AI simply gives those systems a new role because the person using them may increasingly be a machine as well as another designer.

It means decisions that used to live implicitly in someone’s head need to become much clearer. Why does this component behave this way? Which colour combinations are allowed? How should the brand respond when the format changes? What makes something recognisably ours?

Those are design questions, not AI questions.

Taste is becoming harder to automate

There is another side to all of this that I find reassuring.

As these tools become widely available, access to them becomes less interesting. Eventually, almost everyone will be able to generate a polished image, produce a reasonable interface or turn an idea into a basic working product.

What doesn’t become equally distributed is judgement.

Two people can sit in front of exactly the same AI model and produce completely different work. The difference comes from what they notice, what they reference, what they reject and how far they’re willing to refine something before considering it finished.

Designers sometimes call that taste, but taste isn’t something mystical. It comes from years of looking at things, making things, getting things wrong, understanding why they’re wrong and slowly developing an instinct for what feels resolved.

AI can make that judgement more powerful because it gives us greater leverage over what we already know. It can help us explore ten directions instead of two, prototype something instead of merely describing it and test an idea before committing significant resources to it.

But it still needs someone to recognise which of those ten directions is worth pursuing.

The designer’s role is getting bigger

The most exciting change for me is that the boundary around what a designer can make is expanding.

I don’t need every idea to end as a static Figma file anymore. A brand identity could lead to a tool that helps the client create future assets. A product concept can become an interactive prototype. A repetitive design process can potentially become a small piece of software.

That doesn’t mean designers suddenly need to become developers. It means the distance between having an idea and being able to test that idea is shrinking, and that gives designers much more agency.

For me, that is a much more useful way to think about the future of AI and design.

The parts of design that are becoming easier are largely the mechanical ones. The harder parts — understanding people, identifying the real problem, establishing constraints, creating coherent systems and knowing when something is actually good — haven’t disappeared.

If anything, when making becomes abundant, those decisions become easier to see.

AI can help us make more. The designer still has to decide what is worth making.