What I actually use AI for every day
Let me be honest about what AI does in my actual design workflow. It is not generating full app designs from a prompt. That never works well enough for production. What it does extremely well is small, focused tasks that used to eat up my time.
Content generation is the biggest one. When I need placeholder text for a prototype, I do not use lorem ipsum anymore — I ask AI to generate realistic user names, transaction descriptions, or product titles based on the domain. When I am writing microcopy for error messages or tooltips, I generate 10 variations and pick from them. It is not replacing my writing — it is giving me a broader palette to choose from.
The second big category is research synthesis. I feed transcripts of user interviews into an AI tool and ask it to extract themes, quote relevant passages, and flag contradictions. This used to take me two full days. Now it takes a few hours of reviewing and refining. I still read every quote — the AI is a first pass, not a final judgment.
What AI still cannot do (and probably will not)
AI is terrible at taste. It can generate a competent design, but it cannot generate a memorable one. It can suggest colour palettes that work, but it cannot pick one that feels right for a specific brand in a specific culture at a specific moment. That kind of judgment — the 'I do not know why, but this feels better' instinct — comes from years of looking at things, and no language model has that.
AI also cannot handle ambiguity. It needs clear instructions. But design problems are messy. Stakeholders give vague feedback like 'make it pop.' User needs are contradictory. The best solutions emerge from sitting in discomfort, exploring multiple directions, and trusting your gut to pick the right one. That process cannot be automated because it is not algorithmic.
Finally, AI cannot build relationships. It cannot read the room in a stakeholder meeting. It cannot sense that the engineer is frustrated because a component is hard to implement, and adjust the design accordingly. Design is a social activity, not just a production activity, and AI is not social.
My three-rule checklist
I have three rules I follow whenever I use AI in my design process. These keep me honest and prevent me from outsourcing my thinking.
Rule one: AI output is a suggestion, never a deliverable. I never show AI-generated work to a client or stakeholder without heavy editing. The AI is a brainstorming partner, not a replacement for my judgment.
Rule two: If I cannot explain why I chose a design decision, I do not use it — even if AI suggested it. Every pixel I ship needs a reason behind it. AI can propose. I must defend.
Rule three: Use AI for speed, not for quality. Let it handle the fast, repetitive work so I can spend more time on the hard, creative work. The goal is not to work less. The goal is to spend my best hours on the parts that only I can do.
An honest take
AI is going to change design — but not the way people on Twitter say it will. It will not replace designers. It will make average designers more dangerous by letting them produce more bad work faster. It will make good designers better by removing the drudgery and freeing up their mental energy.
The designers who thrive will be the ones who treat AI like an intern — useful for heavy lifting, useless without direction. The ones who struggle will be the ones who expect AI to do the thinking for them.
I am optimistic about AI because I have seen how much faster I can explore ideas, how much richer my first drafts look, how much easier it is to communicate design intent without spending hours on slide decks. But I am also cautious. The core skills — empathy, taste, judgment, communication — those are still 100 percent human. And I think they always will be.