AI didn’t replace my design process. It changed where the thinking happens.
I remember a time when every design thinking sessions ended the same way: someone taking photos of the wall before we peeled the sticky notes off. User quotes in one colour. Pain points in another. Random observations that seemed irrelevant until someone connected them to something across the room.
Someone would step back, cross their arms, and say: “Wait. I think we’re solving the wrong problem.” And just like that, the room had its aha moment.
As someone trained in design, I’m very familiar with the phases of design thinking — and I love seeing how they connect with AI.
Different models label the phases differently, but the rhythm is familiar: understand, explore, and materialize. In practice, my process moves through empathizing, defining, ideating, prototyping, testing, and, when the work continues, implementation. That process lived on walls, whiteboards, spreadsheets, and research repositories.
Then AI entered the workflow. At first, I treated it like a useful tool. Summarize this. Rephrase that. Give me a cleaner version of this paragraph. AI has been helpful but it wasn’t transformational — at least not at first.
I tried feeding my research into AI instead of only spreading it across sticky walls. Here’s what happened.
The Philosophy: AI Changes the Speed, Not the Responsibility
The biggest misconception about AI in design is that it does the thinking for you. It doesn’t. At least, it shouldn’t.
AI can process information faster than I can. It can summarize research, cluster feedback, generate alternatives, and reframe a problem in multiple ways.
However, AI cannot tell me what matters.
It doesn’t know the politics inside an organization. It wasn’t in the room when a user hesitated before answering a question. It doesn’t understand why a small workflow issue can damage trust in an entire product. That part is still human. So my philosophy is simple:
AI can help me explore the problem space, but I’m still responsible for choosing the direction.
Design thinking was never about the colorful and fun-shaped sticky notes, workshops, or frameworks. Those are just tools. The real work is moving from ambiguity to clarity without losing sight of the people you’re designing for. AI didn’t remove that work. It changed how quickly I can move through it.
1. Empathize: From Sorting Notes to Spotting Signals
Before AI, gathering research insights and patterns looked like this:
Interview users. Review notes & highlight quotes. Write observations. Group the sticky notes. Step back. Rearrange everything. Step back again.
Eventually, patterns would emerge.
It worked, but it took time. Sometimes too much time. And when you’re deep in the material, it’s easy to miss patterns hiding in plain sight.
Now, AI helps me create the first pass.
I can feed interview summaries, survey responses, or research notes into AI and ask:
- What recurring needs appear across these interviews?
- Where do users contradict themselves?
- What emotions show up repeatedly?
- What are users trying to accomplish beyond the task itself?
- What assumptions might we be making?
- What patterns could I be missing?
The value isn’t just speed. It’s perspective.
AI can sometimes notice that several users describe the same frustration using completely different words. Or something might actually be a trust issue. Or that people aren’t struggling with the feature itself — they’re struggling with the decision around it.
However, I don’t accept the summary blindly. If AI says users are “overwhelmed,” I go back to the transcripts and look for evidence. If it identifies a pattern, I ask where that pattern came from. If something sounds too neat, I ask for contradictions.
Because empathy isn’t quote sorting. Empathy is attention.
AI helps me get through the volume faster so I can spend more time understanding what people actually meant.
2. Define: Better Problem Frames, Without Losing the Human Insight
The Define phase used to feel like a high-stakes writing exercise.
You take everything you learned and compress it into one clear problem statement. Too broad, and the team doesn’t know where to start, I’ve been in those rooms. Too narrow, and you risk solving a symptom instead of the real problem. And this is where AI has become surprisingly useful.
I can give it the research context and ask it to reframe the problem from different perspectives:
- What does this problem look like from the user’s point of view?
- What does it look like from the business side?
- What if this is a trust problem instead of a usability problem?
- What if this is an onboarding issue instead of a feature issue?
- What assumptions are hidden inside this problem statement?
For example, a team might start with: “Users aren’t completing the onboarding flow.” That sounds actionable, but it doesn’t explain why. AI can help generate alternative frames:
- Users don’t understand the value of completing onboarding.
- Users don’t trust the product enough to invest more time.
- The flow asks for too much before proving its usefulness.
- Users complete the task, but don’t reach the outcome they expected.
These aren’t answers. They’re lenses.
The danger is that AI can make weak thinking sound polished. It can produce a clean problem statement that feels convincing but removes the tension, emotion, and context that made the insight important.
That’s why I treat every AI-generated definition as a draft. The human part of the process is deciding which frame creates the most useful conversation.
3. Ideate: AI as the Silent Third Teammate
Brainstorming with a team is energizing, but it has limits.
People get attached to their own ideas. The loudest person can shape the direction. The first few suggestions can anchor the room. Sometimes everyone tries to be practical before the team has explored enough possibilities.
On the other hand, AI doesn’t care if its idea gets rejected, it will generate another twenty options. That’s why I think of AI as a silent third teammate, not the creative director. Not the decision-maker. More like the person in the room who keeps asking: “What else could this be?”
One thing I’ve learned: I try not to start ideation with AI. I prefer letting humans go first. Let the team express its instincts, experience, and strange ideas. Then bring AI in to expand the field.
I might ask:
- Give us ideas that solve this without adding a new feature.
- What would the simplest version look like?
- What would a premium version look like?
- What ideas directly challenge our assumptions?
- What are ten approaches we probably haven’t considered?
The first batch is usually predictable. That’s fine.
The useful ideas often appear after I push back, add constraints, or ask AI to combine two unrelated directions. AI expands the option space. The design team still has to judge desirability, feasibility, and viability. It can help us diverge further, but humans still have to converge.
4. Prototype and Test: Validate Before the Work Becomes Precious
The longer you work on something, the harder it becomes to change. That’s one of the oldest truths in design.
By the time a concept becomes a high-fidelity prototype, the team has usually invested time, opinions, and emotional energy. Feedback starts to feel expensive. Sometimes people defend the work simply because they’ve already made it.
AI helps me test ideas before they become precious. Before committing to high-fidelity work, I can use it to:
- Draft user flows
- Generate different value propositions
- Write onboarding copy variations
- Create usability test scenarios
- Explore edge cases
- Anticipate objections
- Turn a rough concept into something people can react to
This means I can validate the thinking earlier.
Instead of asking, “Do you like this design?” I can ask better questions:
- Does this concept make sense?
- Would you trust this?
- What would you expect to happen next?
- Where would you hesitate?
- What feels unnecessary?
- What problem do you think this is solving?
AI can also help prepare discussion guides and usability tasks, making the testing process more structured.
AI cannot replace real user validation.
It can simulate possible concerns. It can help us prepare. It can help us analyze results afterward. But it cannot tell us with certainty how real people will behave. That still requires putting the idea in front of users.
Noticing Patterns with AI
The biggest change wasn’t that my process became automated. It became more intentional.
Before AI, working through research materials — user responses, research notes, and usability findings — took a lot of manual effort. I would read through everything, highlight recurring ideas, and group patterns piece by piece.
Now, AI has become a sense-making partner in that process. It helps me surface patterns I may not have fully noticed and look at familiar material from another angle. I still return to the source material to validate what it finds, but it gives me a different starting point.
I stopped asking: “What can AI generate?”
And started asking: “What can AI help me notice?”
What Actually Changed
Finding research insights became continuous
I no longer have to wait until the end of discovery to start making sense of what I’m hearing. I can connect the dots as I go, then adjust the next interview or research question.
Problem statements became less precious
I used to spend too much time trying to write the perfect problem statement. Now I generate several frames, challenge them, and choose the one that gives the team the most useful direction.
Ideation became less dependent on the room
AI doesn’t replace the team, but it expands the team. It gives us more starting points, stranger combinations, and fewer blank-page moments.
Concepts became testable earlier
I can turn rough thinking into something people can react to before committing to high-fidelity work.
My role became more editorial
I spend less time organizing information and more time understanding and deciding what matters. That’s probably the biggest change. AI gives me more raw material to work with, but raw material is not the product. Human judgment is still the job.
What Hasn’t Changed
For all the speed AI gives me, a few things remain stubbornly human.
AI doesn’t understand context the way people do
It can process what I give it, but it doesn’t know everything I’ve left out. It doesn’t know the client’s internal politics, the team’s technical constraints, or the history behind a decision.
AI doesn’t know which trade-off is worth making
Product and design work is full of trade-offs. Speed versus quality. Simplicity versus flexibility. Business goals versus user needs. AI can list the options, but it cannot take responsibility for the decision.
AI can sound confident when it’s wrong
This is especially dangerous during research insight gathering and spotting patterns. A polished summary can feel authoritative even when it’s built on a weak interpretation. That’s why every insight still needs to be traced back to evidence.
AI cannot care about the user
It can help me understand people. It cannot care on my behalf. That part is still the job.
Principles I Try to Follow
As AI becomes more embedded in my design process, I’ve started following a few principles.
1. Human context before machine input
AI is only as useful as the context I give it. If I haven’t done the research or understood the problem, it will only help me become wrong faster.
2. Use AI to widen options, not make the final call
AI is strongest when I use it to explore possibilities. The final decision still needs human judgment.
3. Trace every insight back to evidence
If AI finds a pattern, I return to the transcript. A polished summary is not proof.
4. Validate with real people
AI can help prepare the test, but real users still determine whether the experience works.
The Sticky Notes Are Still There
I don’t think AI makes design thinking will obsolete.
If anything, AI has exposed which parts of the process were actual thinking and which parts were just manual labour.
Sticky notes were never the magic. The magic was making invisible thinking visible. Creating shared understanding. Giving teams a way to externalize assumptions and move them around.
Prompts do something similar. A prompt is not an answer. It’s an invitation to think differently.
So yes, my process has changed. I now move between sticky notes and prompts, between walls and language models, between human intuition and machine-generated possibilities.
But the responsibility hasn’t changed.
The product designer still has to understand people, frame the right problem, and decide what deserves to be built.
The tools changed from Post-its to prompts. The fundamentals of product-thinking stayed the same.
Thanks for reading! 💗
