METHODOLOGY · MMXXVI
The
Process.
Five phases. Human instinct and AI-accelerated at every step — from the first interview to the last A/B test.
ON AI & DESIGN
AI doesn't replace the designer. It removes the ceiling on how much ground the designer can cover.
I use AI as a thinking partner, not a shortcut. The research still has to be real. The synthesis still requires judgment. The concepts still have to be stress-tested with actual humans.
What AI changes is the leverage. Where I used to explore three concepts, I now explore ten. Where synthesis took a week, it takes a day. That speed goes back into the work — more iterations, tighter testing, better outcomes.
Phase by Phase
In Detail
01
discover
What's actually happening?
Before touching Figma, I need to understand the real landscape — what users do, what competitors do, and what the current product does (or fails to do). This is where assumptions get stress-tested against reality. User interviews, competitive teardowns, heuristic reviews, and analytics audits all feed into a single goal: uncover the truth about the problem before we commit to a solution.
Methods
User interviews
Competitive analysis
Heuristic evaluation
Analytics audit
Stakeholder alignment
AI Acceleration
Claude
interview script generation & transcript synthesis
ChatGPT
competitive landscape briefing
FullStory AI
behavioral pattern detection
Dovetail
automated research tagging
02
Define
What problem is worth solving?
A problem statement sharp enough to reject ideas is worth more than a hundred concepts.
Research is raw material. Definition is the work of turning it into direction. I synthesize interviews into personas that actually capture how people think — not just demographics. I map journeys to find where the experience breaks down. I write problem statements sharp enough that the team can test ideas against them. And I set success metrics before we design anything, so we're not inventing reasons it worked after the fact.
Methods
Persona development
Journey mapping
Problem framing
Success metrics
Design principles
AI Acceleration
Claude
qualitative insight synthesis from transcripts
Notion AI
synthesis documentation
Dovetail AI
cross-session pattern recognition
Miro AI
affinity mapping acceleration
03
Ideate
What could this look like?
The best idea usually isn't the first one. AI lets me find it faster.
This is the phase most designers rush and most stakeholders want to skip. I don't. Wide exploration before narrow commitment is what separates good solutions from obvious ones. I run design sprints to pressure-test ideas in days instead of months. I prototype interface metaphors — five or six competing directions — and stress-test them with real users before we decide which one to build. AI has fundamentally changed the ceiling here: I can explore ten times more territory in the same window.
Methods
Design sprints
Concept sketching
Interface metaphors
Lo-fi prototyping
Concept reviews
AI Acceleration
Midjourney
visual territory & mood exploration
Claude
concept stress-testing & devil's advocate
ChatGPT
divergent ideation partner
Dovetail
Runway ML
04
Build
Make it real, make it right.
Design moves in two tracks: UX enablers that run ahead of development so engineers have tested, scoped wireframes to write acceptance criteria against — and high-fidelity UI that follows one iteration behind. This rhythm eliminates the crunch where design and dev collide. Every component traces back to the design system. Every state is accounted for. Every spec is documented. I use AI to accelerate the mechanical parts so I can stay focused on the decisions that actually require taste.
Methods
UX enablers / wireframes
Hi-fidelity UI
Design systems
Component documentation
Dev handoff & QA
AI Acceleration
Cursor
component code generation from Figma
Figma AI
variant & state generation
Claude
UX copy generation for testing
GitHub Copilot
design token implementation
05
Define
Did it work? What's next?
Shipping is not the end — it's the beginning of learning. Every initiative needs a measurement plan written before it launches. I run A/B tests tied to specific conversion metrics, conduct moderated usability sessions, and synthesize analytics into the next round of bets. The loop between measure and discover is where compounding happens: each cycle makes the team smarter about what users actually need, not what we imagined they needed.
Methods
A/B testing
Usability studies
Analytics review
RICE re-prioritization
Iteration planning
AI Acceleration
Maze AI
automated usability test synthesis
Hotjar AI
session recording summaries
FullStory
anomaly & drop-off detection
Claude
insight synthesis & next-bet framing
Guiding Principles
how i work
Truth before solutions
Research isn't a phase you check off. It's the foundation everything else stands on. I won't design my way to a hypothesis I haven't validated.
Wide before narrow
The best idea is rarely the first one. Divergent exploration — especially with AI lifting the ceiling — produces better outcomes than premature commitment.
Process serves the work
Frameworks are tools, not rules. I adapt the process to the problem — a 0→1 startup needs different rhythms than a mature product team.