AI-Pilled Designer
When I was admitted to IxDIA, the letter said I showed significant promise but lacked some key interactive experience. It was a fair call. I could design an app, prototype it and explain exactly how it should behave. Then I handed it to someone else to make it real.
That handoff was the edge of my job for most of my career. I tried learning to code more than once. It never stuck, and my ideas stayed on the canvas.
Over the last year that edge disappeared. Somewhere along the way, I became an AI-pilled designer: the designer who reaches for AI first and ends up building the thing instead of handing it off. This post is about what that actually means, told through the things I have built.
The line moved
In the past year I have shipped things I would once have left as mockups:
- Human Atlas, an interactive site that maps psychology frameworks and mental models
- Our Jodi, a React Native matchmaking app
- Nurture, a design system with a full component library deployed on Storybook
- Courtside, a sports community app for finding courts, players, teams and tournaments
I am still not a frontend developer. I did not become a product design engineer over a weekend. What changed is that the tools finally speak a language designers already know: intent, constraints and judgment.
It clicked when I built and shipped an internal app at work on my own. The engineering team was surprised that one designer could make it happen, and so was I.
Leave the canvas
Our tools quietly teach us to think inside their boundaries. Open Figma and you think in frames. Open Photoshop and you think in images. You start with what the software allows, then squeeze the idea into it.
A coding app like Claude Code works the other way. It starts almost empty, a blank box and a blinking cursor, but it can reach your files, your images, your design system, your codebase and the other apps you use. Instead of working inside one canvas, you ask all of that context to work together.
I still open Figma most days. But the ideas that used to die in a Figma file now have somewhere else to go. The point is not to become a command line person. It is to stop letting the tool decide what is possible.
Taste became the spec
The first thing AI exposed was my own vague feedback. "Make it cleaner" gets you nothing from a client, and it gets you nothing from a model either.
I felt this most while building a design system generator: you give it a brand color and a platform, and it produces tokens, type scales, spacing, grids and exports. To make it work I had to write down rules I had applied by instinct for years. Why this type ratio. Why this spacing step. When contrast is enough.
That is the critique muscle IxDIA trained. Taste is not liking good things. It is being able to say why they work, clearly enough that someone else, or something else, can act on it.
If you are not sure where to start, copy people. Recreate work by designers you admire, not to publish it, but to take their decisions apart and see what made them work. Once you can explain your references, you stop asking AI to guess your taste and start directing it.
Designing with AI, and for it
At Kouper Health, AI voice agents help guide patients through care after they leave the hospital. I do not design the agents themselves. I design the experience around them: the patient engagement tools and the touchpoints where people meet the agent. That work taught me that when the AI is the product, design shifts from screens to trust. Does the patient know what is happening, what comes next, and how to reach a human?
The moment this became real for me was a research project on speech-to-text models. I compared the options and presented which one we should use. Going in, it looked like a technical choice. Coming out, I saw it was a design decision. If the model mishears a medication name or a date, the agent has to ask again, and the patient starts to doubt it. If it responds too slowly, people talk over it. The model you pick quietly shapes the whole conversation flow.
On the side, I help build Floutwork, an AI workspace that brings a team's tools and AI models into one place. Building it as a designer taught me to work with AI the way I would with a sharp junior teammate. It is fast and capable, but it was not in the meeting. So I give it the goal, the audience, the constraints, examples, and what I do not want. Then I review, correct and decide what happens next.
There is no perfect prompt that ships the finished thing in one shot. The skill is direction: define the problem, assemble the context, and judge what comes back.
There is still plenty of back and forth. That is not AI failing. It is the new creative process: I am no longer pushing every pixel, I am directing the work, reviewing it and deciding what happens next.
Turning judgment into systems
The bigger shift came when I stopped repeating myself. Once a workflow works, I package it as a skill: instructions, examples and templates an AI can reuse without me explaining it again.
I have started doing this outside design too. I am studying screenwriting, and I am building a custom script-writing skill from the material I learn: formatting rules, structure, what script readers look for. Every correction I make becomes knowledge I no longer carry in my head.
This is what building Nurture, a design system, first taught me, and what AI has made far more powerful. Good design knowledge should be portable. When it lives in a system, other people can use it without waiting in my queue.
You also do not have to start from zero. Plenty of builders share their skills openly. Take one, understand it, and adapt it to your own process.
Build without permission
Designers have always grown through work nobody asked for: the poster nobody commissioned, the experiment you make just to understand something. AI does not change that instinct. It removes the distance between the idea and the thing.
Most of what I build now starts that way. Human Atlas began because I wanted one place to explore the psychology frameworks I kept coming back to. Courtside goes back further. It was my individual thesis project at IxDIA, a problem I chose for myself. Back then it ended as a design. Now I am building it into a real product. Both taught me more than any brief would have.
More room to create
I expected AI to make me faster. I did not expect it to make me more creative. Resizing, reformatting, exporting and versioning are small on their own, but together they fill the gaps where better ideas hide. Handing those off gave me time back for the parts of design I love: exploring, sketching and solving problems.
It also changed my role. Working with AI quietly turns every designer into a manager, not of people, but of context, intention and quality. The job is less about making every asset and more about creating the conditions where good work can happen.
So what is an AI-pilled designer?
If AI can produce the asset, what happens to the designer? I think the answer splits us in two. Some of us will keep producing variations of the same work, faster. Others will turn creative judgment into leverage.
To me, being AI-pilled is not about knowing every model or installing every plugin. It is being the person who can move an idea from intention to something real, in other people's hands.
What I would tell IxDIA students now
- Build one real thing end-to-end outside Figma or directly in a coding app. Not the one-prompt website everyone posts. Pick a problem you actually have and take it all the way to someone else's hands.
- Learn enough to ship. You do not need to be an engineer. You do need to know how project files fit together, what GitHub is for, and what has to happen before real people can use your work.
- Practice saying why. Save references and write one sentence on what makes each one work: the grid, the type, the spacing. That vocabulary is how you direct AI.
- Hand AI one repetitive task. Something real that you know well. Automate one step, use it on real work, then add the next.
The program that once said I lacked interactive experience gave me the thing that matters most now: judgment about people, intent and systems. AI did not replace that. It gave it somewhere to go.
Your first attempt will probably be rough. Keep going. You are looking for the moment when something that would have ended in your head or a Figma file works in someone else's hands.
Designers were always builders. The tools have finally caught up.