A repeatable 5-step process for shipping faster without thinking less. AI handles the scaffolding. You make the decisions.
%
less time on manual scaffolding
tools, one clear sequence
design decisions outsourced to AI
The Process
1.
Industry research
Before touching any design tool, I research the industry standard manually. Not to copy it — to understand what users in this space already expect, and where the ceiling is.
I use Google and Gemini AI to map the landscape: existing patterns, user mental models, gaps competitors haven't filled. What I'm looking for: what every player in the space gets wrong.
Tools used: Gemini AI, Chat GPT
2.
Deep research & validation
Perplexity for sourced, verifiable research, not hallucinated summaries. Every claim gets a source I can actually read. ChatGPT is the proof-checker: does this hold up, or is it just one product's quirk?
Everything gets documented in Perplexity. Not in notes, not in a separate doc in the same tool doing the research, so context stays together.
3.
User flow & IA
I take everything from Steps 01 and 02, research, mental models, competitor gaps, add my own analysis, and write it into one structured prompt. Claude drafts the user flow and IA. I read it, challenge it, and re-prompt anything that doesn't hold up.
Usually 2-3 iterations. The structure has to feel right before anything moves forward.
Non-negotiable rule
AI generates the structure. I make the decisions. If Claude suggests a 5-step flow and I know users drop off after step 3,
I don't accept the output; I re-prompt with that constraint.
4.
Rapid wireframing
Once flow and IA are locked, I move to Google Stitch. Not for me, for the business and development team. The wireframe is a communication tool, not a design deliverable.
The goal: does everyone agree on what we're building before I spend time making it look like anything? Multiple iterations, keep prompting until the structure is signed off.
5.
Visual design
Wireframes approved. Now Figma. I use the existing design system and component library, not redesigning patterns that already work. The design system decisions were made earlier in the project. This step is applying them correctly.
The main judgment call: information hierarchy. What's in the first fold? What's the first thing a user's eye hits? That's not something AI decides. That's experience, research, and knowing the user.
Visual design
Visual design is the execution of decisions already made. By the time I'm in Figma, the hard thinking is done. This step should be the fastest in the process; if it isn't, something went wrong earlier.
Why This Works
Time saved
70% less time spent on manual scaffolding, building flows from scratch, formatting research, structuring IA from memory. That time goes back into decisions that actually require a designer.
What AI doesn't replace
Knowing when a flow has one step too many. Knowing why a particular layout will confuse a specific user. Knowing when the AI's suggestion is technically correct but experientially wrong.
You can't prompt your way to that. That's what you're hiring a designer for.
THE ONE RULE
Before accepting any AI output, at any step, ask yourself: do I actually agree with this, or does it just look convincing? If you can't answer confidently, you don't know the problem well enough yet. Go back to Step 01.