AI accelerates thinking, it doesn't replace it
A practical, honest look at where AI actually helps in my design work, and where my own judgment still has to do the job.
Minhaz Ahmad Didat
Senior Product Designer
I use AI every day now. Claude, Cursor, Lovable, a few others depending on the task. I am not shy about it and I do not think designers should be. But I have also watched people expect it to do the part that is actually mine to do, and that is where it falls apart.
The way I think about it is simple. AI is very good at speed. It is not good at deciding what is worth doing. The speed is a gift. The deciding is still the job.
Where it genuinely helps
Prototypes are the biggest one. I can turn an idea into something real and clickable in hours instead of weeks. That changes the conversation with stakeholders completely, because they react to a working thing instead of a static picture. Half the debates I used to have in meetings just disappear when people can click.
Research synthesis is another. After a round of interviews I have a pile of notes, and AI helps me cluster and surface patterns faster than doing it all by hand. But I read the raw notes myself first. The clustering is a starting point, not a conclusion, and I have caught it smoothing over the exact quote that mattered most.
For design systems, it helps me scaffold. Draft the tokens, rough out a component, generate the boring first pass of documentation. Then I go in and do the real work, which is the taste and the edge cases.
Where it does not
AI does not know what problem is worth solving. It does not sit with a confused user and feel the moment they give up. It does not carry the context of six years of watching how people actually behave in a product. It will happily give you a confident answer to the wrong question.
It also has no taste of its own. It has an average of everyone's taste, which is a different thing. If I let it drive the visual decisions, the work comes out competent and forgettable. The interesting choices, the ones that make a product feel like it has a point of view, still come from a person.
Prompting is a craft, not a trick
I treat prompts the way I treat any interface. I care about the input, the output, and the reliability in between. Getting consistent, on brand results out of a model is real work, and it is a skill worth building. But it is a skill in service of the design, not a replacement for it.
My honest position
AI made me faster and let me explore more options than I ever could before. That is a real advantage and I lean into it. What it did not do is make the thinking optional. If anything it raised the bar, because when everyone can produce a decent first draft in minutes, the judgment about what to keep is worth more, not less.
So I use it hard, and I stay responsible for the outcome. That is the whole philosophy. Accelerate the craft, own the decisions.