After we got home from Mexico last week, Julia and I flew to New York. We sat on a panel about how creators can use AI to run their businesses. Slow Creator Fund and Delphi hosted it. Matthew Hussey and Gabby Bernstein spoke first with Dara, Delphi’s CEO. Julia and I went next, alongside William Lasry. The Varun Shetty, VP of Media Partnerships at OpenAI closed.
Here are the ideas I keep coming back to from the day.
1. “Good” isn’t good enough anymore
Matthew said something on stage I’ll butcher in paraphrase, but the gist was: AI has raised the floor. Everyone can now write at a 7 out of 10. He’s a writer. He builds his career on words. And it bothers him that the bar for “good writing” has collapsed.
He’s right. If anyone can produce 7/10 work in seconds, 7/10 work no longer earns attention.
You’d think the answer is more polish. It’s not. The answer is being incredible at speaking to exactly who your audience is. AI raised the average. To matter now, you have to be specific, sharp, and unmistakably you.
Those were his words, these are mine: “The only way to compete is you + AI. Not AI alone. Not you alone.”
That doesn’t mean flashy AI graphics in every video, or agents running your whole business in the background. It means: figure out what people actually value from you — or what they could value but haven’t seen yet — and use AI to give them more of it, or a deeper version of it.
AI is a tool. Better: a coworker. And the thing most business operators are missing is protecting the craft they’re known for while handing off everything around it.
2. The two ways I’ve learned the most about AI
Too many people treat learning AI as the objective. The best learning is a byproduct of chasing a real result. Two things have moved the needle for me.
Carve out dedicated time. I went to the Slow hackathon in San Francisco earlier this year. I’d already been vibe coding on nights and weekends (vibe coding = building software with AI). But I wanted a retreat — somewhere away from Julia, away from the kids, with experts, on one objective. That weekend produced vibeshop.juliaberolzheimer.com, a shoppable experience built from years of Julia’s daily looks. I would not have built that in nights-and-weekends mode. The container made it possible.
You can build a mini version of this for yourself. Block a weekend. Set one goal: “By Sunday night I’ll have X launched” or “I’ll have Y problem solved.” Treat it like a sprint, not a study session.
Use it in moments of desperation. Major unlocks for me have come at 9pm when I’m slammed hours of work ahead of me and don’t want to do the busy work, to do the meaningful work.
The clearest example: I was badly behind on campaign emails. dozens of threads, hundreds of total emails. I was about to grind through the inbox the slow way. As a last-ditch move I asked Claude to build me a Kanban priority board of every campaign I had to catch up on. I expected a list.
It came back with a full board — and then went two moves further than I’d asked. It anticipated which campaigns would need follow-ups, flagged dependencies between them, and pre-drafted the next actions. It predicted moves ahead of where I was.
That was the boundary moment. I’d been asking AI to do the task. The shift was asking it to think like it had my job. Once you push past one boundary, the next is easier. A few weeks of that and you’re doing things you didn’t think were possible.
3. Stop looking for the perfect tool. Use a coworker.
The OpenAI VP closed with how their trying to help the creator economy. They’ve thought hard about how to support creators — dashboards, skills, custom tools. His takeaway after all of it: just start using Codex, their general-purpose coding-and-thinking coworker. The equivalent of Cowork or Claude Code.
Teach a person to fish.
He could have pitched a creator-specific product. He pitched the Swiss Army knife instead. He believes — and I do too — that the dashboard approach assumes someone else knows what you need. The coworker approach assumes you do. You bring the problem. AI brings the labor. That’s the relationship that scales with you.
These tools can dig into your analytics, run your operations, help you frame a script, or build something custom for your business. Most creators are still using them like a fancier Google search. There’s so much more on the other side of treating them like a colleague.
What I’d tell a creator starting from zero
Pick one annoying back-office task this week (email triage, invoicing, scheduling, analytics, contracts). Hand it to AI. See how close it gets.
Block a Saturday. One project, start to finish. No phone, no kids, no partner. Just you and the model. Set a finish line.
Stop chasing tools. Pick one general-purpose coworker — Claude, Codex, whatever — and go deep. Depth beats breadth.
The business who’ll matter in the future aren’t the ones with the flashiest AI gimmicks. They’re the ones who used AI to free up time for the work only they can do.
Reply and tell me: What are you incredible at that an AI coworker can support you with? Curious what comes up.
