What “AI-assisted design” actually means in practice

I get asked what AI-assisted design actually means more than almost anything else, usually by designers who are equal parts curious and worried. And I understand why. The phrase sounds like it should mean one clear thing, and instead it means about six, depending on who is selling you something.

So here is the honest version, from someone doing it daily. AI-assisted design is not a tool you bolt on at the end to make pictures faster. It is a shift in where your time goes. The generating part, the bit everyone fixates on, turns out to be the smallest part of the change.

Let me show you what it actually replaced in my week.

What does AI-assisted design actually mean?

AI-assisted design means using AI across the whole design process, not just the visual output: research synthesis, briefing, exploring options, writing the connective tissue between design and engineering, and pressure-testing decisions. The design judgement stays with you. The volume work moves to the machine.

The mental model most people carry is that AI does the creative part and you supervise. It is almost exactly backwards. The AI is brilliant at the parts that were never creative in the first place, the summarising, the reformatting, the first pass, the third variation. Your judgement is what it cannot do, and that judgement is now the thing you spend most of your day on.

I noticed this properly when I looked at where my hours actually went after a few months of building an agent-based Design Ops system at Drova. I was not designing fewer screens. I was making more decisions per hour, because the gaps between decisions had collapsed. The waiting was gone. What was left was almost entirely the thinking.

Where the assistance actually shows up

If you shadowed me for a week, you would see AI turn up in places that have nothing to do with generating a layout.

It shows up in research, where I hand it forty pages of user interview notes and ask what patterns I am not seeing, then argue with its answer. It shows up in briefing, where I turn a vague Slack request into a structured brief before a single pixel exists. It shows up in exploration, where I ask for six directions I would never have drawn myself, not because they are good, but because two of them shake something loose. It shows up in the unglamorous middle, writing the component documentation, drafting the release note, translating a design decision into language an engineer can build from.

The output is the least assisted part of my process, not the most.

That is the reframe I wish someone had handed me earlier. When people say AI is coming for the creative work, they are looking at the one place it is weakest. A model fills every unspecified gap with the most statistically typical answer, which is a fancy way of saying the most average one. Average is fatal to design. So the creative decisions, the taste, the “no, not that, this” stayed exactly where they always were. With me.

The part nobody warns you about

Here is what the enthusiastic version leaves out. AI-assisted design is more demanding, not less.

When a first draft costs minutes, you produce ten times as many first drafts, and every one of them needs judging. The bottleneck moves from making to deciding, and deciding is tiring in a way that pushing pixels never was. I have finished days where I made almost nothing with my hands and felt more spent than after a week of production, because I spent eight hours choosing.

This is the tax nobody mentions. The scarce resource is no longer your ability to produce. It is your ability to know what good looks like, quickly, over and over, without your standards eroding as the volume climbs. Nielsen Norman Group’s survey of more than 800 UX professionals found that 92% already use generative AI tools, most of them weekly or daily. The production capability is already everywhere. Taste does not scale the same way. Taste is the moat.

Which is why I keep saying the same thing to worried designers: the skill to protect is not your ability to push pixels. It is your ability to look at ten options and know, fast and for real reasons, which one is right.

So what should you actually change?

Start by moving one non-visual task to AI and keeping the visual ones for yourself. Not the other way around.

Most people try to make AI design the screen and are disappointed, which is the wrong experiment. Give it the research synthesis, the meeting-notes-to-brief translation, the first draft of the documentation, the “explain this decision to a stakeholder” email. Watch how much time comes back. Then spend that time on the decisions, because the decisions are what your name is actually on.

This is the same idea I keep circling in designers who resist AI are asking the wrong question: the craft is not disappearing, it is moving upstream, into the judgement we always claimed was the real job. AI-assisted design is just what that move looks like on a Tuesday. Less making. More deciding. A brief that matters more than the file it produces.

The designers who thrive in this are not the ones with the best prompts. They are the ones who got clearer about what they actually believe, because now they have to say it out loud, in a brief, to a machine that will take them literally.

So when someone asks me what AI-assisted design means in practice, the shortest true answer is this: it means your judgement finally became the whole job. Which is either the best news of your career or the scariest, depending on how confident you are in it.

—

---

Frequently asked questions

Want the workflows without the guesswork? The Claude Starter Guide comes with the full skills kit on GitHub, including the non-visual workflows that give a designer’s week back.