Transformation vs Adoption
My mindset right now is transformation vs adoption.
Most teams talk about adopting AI — slotting it into the existing workflow, keeping the workflow intact. I'm interested in the other direction: letting AI change what the workflow is for. When the tool stops being a way to render the same artifact faster, and starts being a way to do a different kind of thinking, the job itself reshapes.
That's the frame I'm working from. Everything else on this page is a consequence of it.
Design Time Into Thinking Time
The double-duty problem
In Figma, every prototype I built was carrying two stories at once. There was the tool story — the thing engineering was going to build, the components and the states and the edges. And there was the artifact story — the customer experience the tool was supposed to create. Both stories had to live in the same fixed sequential flow. Both had to resonate across a wide population of customers who didn't share a context. One designer, two narratives, one shot.
The cognitive load of association
The trick I was asking customers to do, every time, was to associate my fixed story with the specific gap they had on their service team. That association was the work. It was invisible work — they did it in their heads — but it was the part that determined whether the demo landed. If they couldn't map my story to their situation, the value never showed up. We were dressing up Figma frames and hoping the bridge would form.
The AI-native flip
A prototype built on real infrastructure doesn't need to be sequential. The customer can bring their own case. The thing in front of them responds to their service team, their edge cases, their CRUD-and-empty-and-error states. The value stops being story-fitting and starts being service-knowledge specific. That's the part that feels like magic, but it isn't magic — it's the removal of the association load.
The reallocation
Here's the part that actually matters for the design discipline. When the tool doesn't absorb your time anymore, the only thing left is thinking about the customer. Research, service blueprints, customer journeys — these have been treated as nice-to-haves for years because most of the calendar was burning on getting a design reviewed and approved in Figma. They aren't nice-to-haves. They're the must-haves. They're what's left when the manufacturing of artifacts stops being the day.
The Rhythm
It may start as uncomfortable until you understand the rhythm or you create one, then the rhythm creates patterns for others to follow. You socialize to bring others to that plateau, then move into the unknown and start all over until it becomes comfortable.
The pattern
Macy's. Albertsons. AWS. The arc looks like job hops, but underneath it's the same loop, repeated. Arrive somewhere new. Learn the landscape. Find the gaps the people who'd been there had stopped seeing. Build the mechanism that closes the gap. Push past the established norms when they're in the way. Share the mechanism so other people can run it without me. Then look for the next stretch — growth doesn't fall in your lap, and the frontier is where I want to be.
The reason it works is the rhythm. Discomfort is the start state — it has to be — and the work is to find or invent the cadence that makes the unfamiliar legible. Once that cadence exists, other people can follow it. Once they're following it, the plateau forms. Once the plateau forms, it's time to move on. Moving on doesn't always mean out — the next unknown can be inside the same walls, if the walls are big enough.
Why this matters now
At AWS right now I have a carved-out role with autonomy: room to question internal guidelines, to stay on the bleeding edge, to share theories about what AI-native actually means. I'm not looking to sit back and cruise yet. I want to stay relevant — and the frontier is where that happens.
The Pairing
The same person keeps bringing me with them.
For 15 years I've been the builder half of the same design-executive pairing. Sprout Designs. Macy's. Albertsons. AWS. Four companies, one partnership. The pattern isn't an accident, and the part worth naming isn't the loyalty — it's the structure of what two people can do together when the roles are separate and the trust is absolute.
Strategy and build, not strategy then build
A design executive sees the opportunity. The market shift, the political landscape, the thing the org isn't paying attention to yet. What they don't see — and shouldn't have to see — is the specific mechanism that turns the opportunity into shipped software. That's the builder's job. In a conventional org, strategy hands a brief to execution, execution sends clarifying questions back, and a quarter of calendar burns in the translation gap.
The pairing collapses the gap because the conversation never stops. Strategy proposes; the builder interrogates — asks the questions that make it better, surfaces the constraints the exec can't see from where they sit, pushes back on the pieces that won't survive contact with reality. By the time the exec walks out of the meeting, the strategy is sharper and a build plan exists. No brief. No handoff. No translation layer.
Why it moves faster than a department
A department optimizes for repeatability and coverage. A pairing optimizes for throughput. The department has to serve many strategies simultaneously; the pairing serves one at a time, deeply, and ships. The department debates ownership; the pairing has one of each and knows who does what. The department's political cost of moving is high; the pairing's is zero. When a hard call has to be made, two people can make it over coffee.
The trade is obvious: the pairing can't cover everything. What it gives up in breadth it earns back in speed and in the quality of the thing that ships — because the builder was in the room when the idea was still soft, and the exec is still in the room when the build hits its first wall.
The evidence
Macy's: three codebases to one responsive framework; 50+ component variants to one coded source of truth. Albertsons: UX practice built from zero during a digital transformation; an ADA-compliant picking app that moved +213% visibility and +14% picks/hour across 2,000 stores. AWS: an AI-native platform in daily use by 2,000+ people, a multi-agent spec system that turns a raw idea into a funded specification in under 15 minutes, a self-serve onboarding mechanism that took the platform from 3 to 10 design systems by removing the human bottleneck.
Each of those is the same pattern. Executive thesis, builder translation, shipped platform. The companies change. The shape of the work doesn't.