My Product Design Process with AI
Updated: 05/29/2026
I use AI tools to speed up routine tasks so I can spend more time on product strategy, complex data architecture, and UX problem-solving.
Here is how I design products using a mix of human judgment and AI efficiency.

01
Stakeholder Alignment
My Role
Leading the kickoff meeting: defining the scope with PM, Engineering Leads, and Business Stakeholders, and capturing business goals and the engineering limitations, system constraints, or infrastructure realities that dictate what can actually be built.
AI Leverage
Running a single phone-recorded kickoff transcript through multiple LLMs to compare outputs and extract key details.
Tools
Claude, Gemini, ChatGPT.
Deliverables
Core vision, engineering limitations, and success metrics, business assumptions regarding the user.

02
Competitive audit
My Role
By choosing direct and indirect competitors, I evaluate the market landscape to locate opportunities and understand how current products address the problem.
AI Laverage
Aggregating product feature data and sorting competitor traits into structured layouts.
Tools
Perplexity, Claude, Gemini
Deliverables
Market Positioning Map: A summary pinpointing where our product can offer unique value. Tracking feature parity, strengths, and weaknesses across competitors.

03
Personas, journey & flow
My Role
I gather user insights from whatever is available: support tickets, sales calls, existing research, or direct chats with internal users. From there, I build personas, map the journey, and find where users drop off or hit friction.
AI Laverage
Pulling common themes from these sources. Drafting first versions of personas and the journey map that I then refine.
Tools
Claude, FigJam, Gemini, ChatGPT.
Deliverables
Personas, user journey map, primary user flow, pain points with quotes.

04
Mid-fi wireframes
Activities
I work with back-end engineers to gather the available and required data, then build a basic layout that reflects real content. I skip lo-fi unless a product has no design direction yet.
AI Laverage
Generating layout drafts from a prompt to explore options before committing.
Tools
Figma, Uizard, Figma Make
Deliverables
Mid-fi wireframes, agreed layout direction, data requirements documented with engineering.

05
Usability testing
My Role
I run sessions with 3-5 users or internal teammates to check four things: does the flow have logic, is the copy clear for non-builders, are there gaps or broken interactions, does the value proposition land in 30 seconds.
AI Laverage
Pulling friction points and confusing copy from recorded sessions.
Tools
Claude, ChatGPT
Deliverables
Test summary with friction points, copy fixes, and flow changes for the next round.
06
Hi-fi mockups
My Role
For a new product with no design system, I explore multiple visual directions and pick one to build a starter component library. For an existing product, I align with the front-end team on the library they already use, then design within those components so handoff is faster.
AI Laverage
New product: Multiple hi-fi direction options from a brief, before committing to one.
Existing product: New screens and component variants that match the existing design system.
Tools
Figma, Figma Make, Stitch, Claude, ChatGPT
Deliverables
Hi-fi Figma file, component library, or extension of an existing one, finalized UI content.

07
Engineering handoff
My Role
I walk engineers through the Figma file in a live session, annotate spacing and states in dev mode, and write spec notes for anything Figma cannot show, and stay available through the build.
AI Laverage
Drafting interaction spec notes, edge-case documentation, and the engineering Q&A doc.
Tools
Figma dev mode, Claude, ChatGPT
Deliverables
Annotated Figma file, interaction spec notes, engineering Q&A doc.

08
Launch & Learn
My Role
After launch, I track user feedback, support tickets, and analytics. I maintain a design debt log of small problems we shipped with, along with planned fixes for the next sprint.
AI Laverage
Summarizing feedback into themes. I verify the themes are real (not invented by the AI) before sharing with the team.
Tools
Claude, ChatGPT
Deliverables
Weekly insight notes, design debt log, next-sprint priorities.
AI multi-tool workflow - the best use of my time and tokens
I am using this jumping strategy to ensure that I don't waste hours or deplete the AI's memory capacity on minor modifications inside more expensive models.

1. Generate initial designs in Claude
I start here because Claude delivers better quality out of the box compared to other tools.

2. Push to Figma
I move the generated design to Figma to manually tweak layouts and fine-tune properties.

3. Iterate in Codex
I bring that polished Figma baseline into Codex: bulk editing, swapping elements, and refining.
(Codex uses 3 to 4 times fewer tokens than Claude and processes changes much faster)
I bring that polished Figma baseline into Codex: bulk editing, swapping elements, and refining.
(Codex uses 3 to 4 times fewer tokens than Claude and processes changes much faster)

4. Push to Figma
Once iterations in Codex are complete, I push the updated screens back to Figma.

5. Bring into Claude
Finally, bring the refined layout back into Claude when I'm ready to run large-scale design explorations or generate clean handoff code for developers