My actual process (with AI)
Junior designers ask me what my design process looks like. i get why — school teaches you a clean sequence: research, define, ideate, prototype, test. It feels like something you can hold onto.
The honest answer is: it depends. And to that end, maybe there is none.
The shape of the work changes based on what’s ambiguous, what’s already decided, who’s available, what’s feasible, how much time you have, etc. Some projects i go deep on research first. Others i jump straight to a rough prototype because the fastest way to get alignment is to put something in front of people. Sometimes I pull in an engineer early to determine (or push) the constraints. It gets more tangled when you’re balancing several projects at once.
The double diamond is a decent map, but right now, the field is changing fast, because we’re actively figuring out where AI fits. It’s added new places to stand in the middle of the work:
- In divergent phases, I can use AI to flood the space with twenty directions, half of them bad, and react. Reacting is faster, and a wrong idea sharpens my sense of the right one.
- When narrowing down, I use it as a thought partner — lay out two options, have it argue both sides. It catches the thing I was conveniently ignoring.
- I can have it roleplay a proxy stakeholder to find obvious holes before spending someone’s real time. A cheap rehearsal, never a replacement for actual research.
- I can use claude code as a window into the codebase — what exists, where a new thing fits, how much friction a direction causes — without grabbing an engineer for every small question.
None of these fit neatly into a stage; they happen whenever they’re useful. I don’t have it all figured out — the proxy stakeholder can lull you into thinking you’ve done research when you haven’t and the idea-flood can drown your own judgment if you lean on it too early.
So what do you follow, if not a process? Instinct. The ability to read a situation and know what it calls for. When to talk to someone versus just decide and move. When you’ve been diverging too long. When a prototype will unlock things faster than another round of research.
That instinct doesn’t come from the frameworks. It comes from doing the work, making the wrong call, noticing, and adjusting — over and over.
So the actual advice: Play. Experiment more than feels comfortable. Try the stakeholder chat earlier than you want to. Prototype something rough when your gut says do more research. Use AI somewhere and see what breaks or improves. You won’t find the right sequence or process by reading about it. You’ll find it by building enough reps that you start to feel when something’s off — and right now, while none of this is settled, those reps are worth more than they’ll ever be.