Michele Kawamoto Perry had me on her podcast, Expressions of Leadership, to talk about leadership, AI adoption, curiosity, and the winding road that got me here.
We covered graphic arts, which my grandfather and my father both made their careers in, and so did I. Teaching myself HTML in 1995 by viewing source code on other people’s sites, because there were no books on it yet. The Squidoo years with Seth Godin. And where AI fits into all of it now.
Here’s the full conversation, with the key ideas expanded below.
Where The Expert Prompt started: a Python script I couldn’t write
I got ChatGPT in late 2022, opened it, and thought, “What am I going to do with this?”
I work with a lot of data. For years, I’d told myself I should learn Python because it could turn that data into reports. I bought two books in early 2022 and started working through them. It was going to take me a couple of years to get any good, and nobody on my team could write it either.
Then one day I noticed a JavaScript example sitting right there on the ChatGPT home screen. Not Python, but I thought, “Oh, it can write code.”
So I gave it my data and the report I’d been building by hand, and asked for a script that would do the job. It wrote it. I ran it. Work that took me hours came back in a second, and the report was right.
That was the moment.
What I didn’t yet have the language for was why it worked. I knew the data. I knew exactly what the report needed to look like. The only missing piece was the programming in the middle, and that’s the part I handed off.
My wife put a finer point on it years later.
I bought the domain for Bassists.com in 1999 and couldn’t build the thing I wanted. In 2025, I finally put together a prototype using Claude in a chat window, which was cumbersome, and then it sat. In April 2026, I built the real version with Claude Code over a weekend. Security checks, ADA compliance, all of it. I told my wife I couldn’t believe how easy it had been.
“Not everybody could do that,” she said.
“What do you mean?” I asked.
“You knew what to tell it.”
She’s right, and I owe her the idea. I’d been doing it for more than three years without noticing.
Why AI mandates fail with teams
I’ve seen a version of this more than once. “We’re going to use AI. Everybody gets an account. Start using it.”
That’s like handing Photoshop to someone who has never opened it and saying, “Start using this.” I’ve used Photoshop forever. Someone who hasn’t will stare at it and quietly decide they’re not a Photoshop person.
A mandate skips the two things that actually matter. It doesn’t address the fear, and it doesn’t address the people who simply don’t know where to begin. I was one of those people. I opened ChatGPT and had no idea what to do with it.
I lived through the dot-com era, and this AI boom reminds me of it. Wild valuations, big claims, a pendulum swinging hard in both directions. It sorted itself out then, and it will sort itself out now, but patience beats a mandate in the meantime. Have a plan. Do some training. Set up labs where the people already doing good work with AI guide everyone else.
What my father and Seth Godin taught me about leading experts
When I was promoted to assistant art director, I panicked and went to my father. He’d had a long, successful run as an art director, and he always talked warmly about the people on his team.
“Hire people better than you,” he said, “and give them room to succeed.”
That stuck, and over time I came to understand the real point of it. As a leader, it’s not how good I look. It’s how good we look. If you’re growth-minded, and I am, relentlessly, you don’t want a B team. You want the best people you can find, and then you want to get the obstacles out of their way. I always figured I worked for them.
Years later, I got the same lesson from Seth Godin, and then I got it daily for nine years.
I’d been a fan since Permission Marketing in 1999. When he posted on his blog looking for four people for a summer project, I emailed him and assumed I’d never hear back. He replied in 15 minutes. In 2005, I went up to New York to meet him and the other three.
I had one job: write the HTML and CSS for a prototype. Sitting there, I could see we weren’t documenting the project the way a web build needs to be documented, so I got up and took over the whiteboard. That night I emailed Seth. “I’m really sorry. I kind of took over.”
He wrote back, “Keep doing what you’re doing.”
We never built that prototype. It was too big an idea. I spent the summer scoping it instead, and the scope document ran about 100 pages. In August, he asked me to come to the meeting a day early. “I’d like you to run it,” he said. That became Squidoo, and I ran it for nine years before it was acquired.
Two things about Seth stayed with me.
The first happened before we ever talked business. He picked me up at the airport, and at the gate on the way out he looked at the attendant’s name tag and said, “Hi, (her name). How are you today?” She’d been slumped and clearly having a rough day, and I watched her transform. He took a second to treat a person like a person.
The second was how he ran the room. In meetings with Google, Amazon, and eBay, everybody’s eyes were on Seth, and he’d push himself down. “I don’t know anything. Corey knows everything.” He lifted everyone else up. He wanted the company flat, and in practice, everyone had a lane and real authority within it. We’d all talk it through, and then the person whose turf it was had final say.
I’ve worked that way ever since with developers, project managers, and editors. Experts don’t like people mucking with their work and making it harder. Give them the call.
The editor who came around to AI on his own
My entire editorial team had concerns about using AI, and they voiced them. My senior editor among them.
If I had a mandate, it was this: “I want you to try it.” That’s all.
He tried it, willingly, and I left it alone. A few months later, he started sharing things with me. Some of it was work. Some of it was a personal project, research on a hobby he wanted to learn, which is the part that told me it had landed. He wasn’t doing it because someone told him to. He saw its value and looked for other places to use it.
At work, it went further than that. He surfaced information about our audience I hadn’t come across. We built a plan around it, and as we rolled it out, we saw growth.
He got there because he was curious, and because nobody stood over him with a deadline while he figured it out.
That’s the case for enablement over mandates. The people doing the work already have the expertise. Your job is to help them aim it, then get out of the way and let them be the ones who found something.
Knowing when to close the laptop: time-boxing AI work
Here’s the part I don’t hear many AI advocates admit.
This stuff can get addictive. When I was building Bassists.com, I say it took a weekend, and it did, but I was falling asleep at the computer at 2 in the morning because I didn’t want to stop. Success comes so fast that you keep going. And going.
So I time box it now. I decide how long I’m going to work on something, and I stop when I hit it.
My wife plays video games at night. Claude is my video game. That’s fun, and it’s also exactly why the limit has to exist.
Curiosity is the real AI skill
The one thing that has held across every chapter of my career is that I go look at the new thing.
At 11, I bought a home computer and taught myself BASIC. In the mid-1990s, my CEO at a catalog company said we needed a website, and since there were no books on HTML yet, I learned it by viewing the source of sites I liked and reverse-engineering them. I found Viaweb through a tiny ad in a trade magazine, and back then, when you called, Paul Graham answered the phone. I built a store in about an hour at a company that printed and mailed 6,000,000 catalogs a year.
Same instinct, decades later, when I opened ChatGPT with no idea what it was for.
And it doesn’t have to be work. Last year I decided to grow tomatoes from seed, having never farmed anything in my life. When the seedlings came up, I had no clue whether they looked right, so I took a photo and handed it to ChatGPT.
It asked, “What’s the weather like at night?” It told me the evening temps were too cold, and I felt like I was going to kill the plants before it guided me through the process.
It also told me they were getting spindly from reaching for light and that I needed a heat lamp. I had a spare desk lamp with a bunch of settings, so I pulled the specs off my Amazon order and asked if it would work. “Use the third setting,” it said.
After that, I sent a photo of the seedlings and a screenshot of the forecast every day, and got a plan back.
The celery is the part that got me. I asked one question: the lower leaves looked paler than the top. Was that bad? It said the color was fine.
Then it told me two things I hadn’t asked about.
My soil was too packed because a downpour earlier that day had compacted it. I never mentioned the rain. And it noticed I’d planted green onions next to the celery, which it said was a good move, because they repel the insects that go after celery.
I asked about leaf color. It read the whole photo.
I didn’t need any help with the tomatoes this year. I learned it by doing it.
That’s the whole idea, really. Bring what you know, let AI fill the gaps, and stay curious enough to keep poking at things.
Just go to bed before 2 a.m.
This is the thinking behind my consulting work. If your team wants to get real value out of AI, here’s how we’d work together.
