He's a former design exec at Apple, EA, and Nestlé — 20+ years of building teams and translating design into language executives care about. He's coached 2,500+ designers from Google, Zillow, Uber, and beyond.
We spoke about 3 things to build your Personal Operating System with AI 👇
1. The context layer
Your job isn't execution anymore. It's holding the story.
Ryan used a TV analogy: think show-runner, not writer.
The show-runner doesn't write every scene. They make sure the story holds together from beginning to end. Every scene, every beat — serves the bigger arc.
That's what AI needs from you now.
Most product teams can't answer a simple question: "What are the 8 behaviours that tell us this product is actually working?"
Not because they're bad at their jobs. Because nobody wrote it down.
That's your context layer. The story only you can hold. And if you don't own it explicitly — the model will make it up. Badly.
2. Codify your expertise
Here's the thing nobody tells you about senior designers: the better you get, the harder it is to explain how you think.
Your brain is running at nanoseconds. Pattern matching. Reading rooms. Making calls. But ask you to write it down? Blank.
Ryan's fix: stop thinking in skills, start thinking in situations.
What scenarios keep coming up? How did you navigate them? What worked, what didn't?
That's your judgment. Not "be a better communicator" — but "when a stakeholder shows up tense, I stop making logical arguments and switch to coaching mode."
Specific. Testable. Something a model can actually use.
He exported FigJam decision trees, recorded voice memos of his thinking, handed it all to Claude. The model didn't guess. It had a map.
3. Build AI agents with your judgment
Start here: what's the 30-minute conversation you have on repeat?
Same question. Same logic. Same advice.
Every week.
That's your first agent.
Ryan's process is basically software development minus the code. Built in one chat. Test in a brand new one — so the model has no memory of how it was built. Log what breaks. Fix it. Repeat.
The goal isn't a perfect agent. It's one that sounds like you.
And the ROI math is simple: one 30-minute conversation doesn't justify three days of building. But 240 of them a year? Now you're getting your time back.
Ryan said something that stuck with me:
"AI models are prediction machines. They predict no matter what. Give them your logic, or they'll make it up."
Remember this: your expertise is the prompt, everything else is just autocomplete.
Ryan runs a hands-on one-day workshop for experienced design leaders who want to turn their judgment into AI agents that work as you do. No coding required.
We also went deep on whether building agents is worth the time, what Claude Cowork unlocks for non-coders, and how to test like a developer — without touching a line of code. Watch the full ask my anything ↓
You received this email because you've downloaded a resource, signed up to hear from me, or you bought something from me. Change yourpreferencesorunsubscribeor .
