AI made "good enough" cheap for anyone. Design judgment is what makes it good: knowing what to keep, what to cut, what to push, and holding the standard AI can't hold on its own. The field is catching up. In 2026, AI-in-design research now finds taste and judgment matter more, not less. I'm a product design leader who builds the infrastructure to run at AI speed without losing that judgment: design systems, agent-orchestrated workflows, and the cross-functional bridges that let smaller teams deliver at enterprise scale. Hands-on at every level. That's not a compromise at my seniority. It's the job now.
How I Leverage AI Across the Full Product Lifecycle
AI isn't a separate initiative. It's integrated into how my teams work — from discovery through delivery.
AI-Accelerated Research
Synthesizing thousands of survey responses, usage patterns, and competitive signals in hours instead of weeks. AI finds the haystacks. My researcher finds the needles.
1000s of survey responses synthesized via Claude — foundation for coupon discovery research.
AI-Augmented Creation
Generate 5–10 directions in days, not weeks. AI gives you something to react to instead of starting from nothing. The cost to restart is effectively zero — that changes everything.
Code to Canvas. Bidirectional — code becomes editable design, design drives production code.
AI-Informed Operations
Built team workflows around AI tools at key decision points. Established evaluation criteria for when AI adds value — and critically, when it doesn't. A team of 5 delivering the work of 8–9.
~40% capacity gain through strategic AI integration across the design team.
From Brain Dump to Developer Handoff
A 10-step workflow where AI compresses everything before and after the craft moment — but the craft moment is human.
Brain Dump Dictation + raw context · Human
Scope & Plan Claude synthesizes into outline · AI
The Bad Build First draft — cost to restart: zero · AI
Creative Director + Team Find the sparks, cut the noise · Human
Refine Code-based iteration · Both
Expand Context Document learnings · Both
Code to Canvas Working UI pushed to Figma as editable frames · AI
Team + Design & Craft Flow state — novel solutions emerge · Human
Build Again Design context flows back via MCP · AI
Handoff Mocks + functional prototype · Both
Feature Design Process
5 phases, cross-functional roles, AI augmentation breakdowns, and key goals at every stage.
Five phases: Discovery (AI-Accelerated), Convergence (Human-Led), Build (Human + AI), Review (AI-Accelerated), and Post-Launch (Human + AI).
Authentication Redesign Under Pressure
A real-world example of the process above — compressing a multi-sprint authentication overhaul into weeks while leading the team and making the strategic calls simultaneously.
AI Didn't Do the Design Thinking. It Made Design Thinking Possible.
A risk-based authentication system had locked thousands of users out of their accounts. Phone-only verification was a dead end. I pulled this onto my own plate while my team was fully allocated across four workstreams — a player-coach moment under real pressure.
AI analyzed stuck-user data and modeled verification paths. It generated the rough first pass — not looking for final designs, just the right direction. Then I put on the creative director hat: finding the edge cases and error states AI will never think about on its own. The real craft happened in Figma, where solutions emerged I couldn't have prompted my way to.
The constraint: 1000s of users stuck in verification loops · email verification completing at 28%.
The result: 75K dead-ends prevented · email verification at 45%+ · 40% drop in support escalations · delivered in weeks vs. a multi-sprint timeline · IC and lead execution running simultaneously · a proactive email campaign to prevent recurrence.
Claude — Data Synthesis · AI — First-Pass Concepts · Figma — Craft & Precision · Player-Coach Leadership
AI compresses everything before and after the craft moment. But the craft moment is human. That's still where I have the most flow state. That's where novel solutions emerge that I couldn't have prompted my way to.
JD McCulley — Sr. Director of Product DesignMCP: The Bridge Between Design Intent and Production Code
Model Context Protocol is the standard that lets AI tools talk to specialized systems — design editors, file systems, APIs, databases. It's the wiring between AI and everything it touches. I use it every day.
Claude Code is the orchestrator. It connects to specialized MCP servers — Figma via Code to Canvas for bidirectional design, the file system for code output. One conversation, multiple tools, no context switching.
This portfolio is the proof. Every page you're reading was designed and built through MCP-connected pipelines before the tools were officially integrated. When Figma and Anthropic announced Code to Canvas on February 17, 2026, it formalized the workflow I'd already been running.
The craft moment is still human. MCP collapses the distance between idea and artifact. But taste, judgment, and the creative direction that makes output worth shipping — that's the designer's job. The protocol handles the plumbing. I handle the decisions.
This protocol is already reshaping product design. Code to Canvas is the first formal instance of design-tool MCP going mainstream — and the same architecture will connect products to AI agents tomorrow. A customer asking their AI to "find me a deal on toothpaste" at Dollar General. The designer who understands MCP at the tooling level now is the leader who architects agent-ready product experiences next.
The pipeline runs top to bottom: Creative Direction (design intent, constraints, taste) → Claude Code (the AI agent that orchestrates via MCP) → MCP Servers (Figma Code to Canvas · File System) → Production Output (this portfolio — designed and built live).
Design Systems as Strategic Infrastructure
A tokenized, well-structured design system is no longer a nice-to-have — it's the infrastructure that determines whether your team can move at the speed these tools enable.
Design system architecture, layered from the ground up: Design Tokens (colors, spacing, typography, motion), Components (buttons, inputs, badges, modals), Patterns (navigation, forms, cards, data display), Screens & Flows (full experiences composed from patterns), and an AI Acceleration Layer (Claude Code + Figma Code to Canvas + MCP). Built across both the DG and pOpshelf brands.
At Dollar General, I led the creation of multi-phase design systems for both the DG and pOpshelf brands. Tokenized naming conventions, shared component architecture, custom Figma workflows adopted across engineering teams. At the time, the value proposition was consistency and handoff efficiency.
Without that systematic foundation, you cannot iterate at the pace AI makes possible. Your components need to be structured so the system can work with them. Your tokens need to be consistent. Your naming conventions need to be clear enough that both humans and AI agents can navigate them.
The companies that don't have this foundation are going to find themselves stuck — all the speed tools, no systematic way to use them. The design system is what makes the accelerated product lifecycle actually work.
With a design system: AI generates within constraints · consistent across both brands · engineering handoff is clean · iteration speed compounds.
Without a design system: AI output is inconsistent · every screen is bespoke · dev rebuilds from scratch · speed tools, no system to use them.
Where AI Falls Short
Knowing where AI doesn't work is what separates a design leader from someone who just lists "AI" on their resume.
Real Users Break Things AI Can't Predict
In a split tender usability test, participant seven failed a task that six passed. If we'd stopped at six, we'd have shipped a broken flow. Synthetic users give directional signal — but they don't misread a button label because they're in a hurry.
Precision Requires Design Judgment
AI generates maximalist output — it throws chips, icons, and cards everywhere. The value is knowing what to keep, what to cut, what to refine. The designer's job shifts from producer to creative director and curator. Volume is AI. Taste is human.
Let's talk about design-led AI.
Product design leader building teams that deliver at enterprise scale — operating as creative director and curator of AI-augmented output, staying hands-on in the work.