In April 2022, I shut down my podcast.
Marketing Hyperdrive made it 12 episodes. That's not a typo. Twelve.
So I did what any determined creator would do. I studied what successful podcasters were doing differently. I changed the format. I changed the frequency. I paid for editing tools that promised to make production painless. Then I launched a brand new show, The Virtual Event Strategist, with a sharper focus and a better plan.
It died a few months later. Also around 12 episodes.
Two shows. Two funerals. And here's what took me years to see: between those funerals, I had changed everything - the tools, the format, the frequency, even the topic - everything except the one variable that killed them both.
The structure of the work never changed.
The experiment I didn't know I was running
Look at what I actually did in 2022. Every time the podcast started drowning me, I made a purchase. A new editing tool. A new production workflow. A new format that promised less overhead. I treated podcasting as a tooling problem, and I kept buying my way toward a solution.
But the tools were never the problem. The problem was that a podcast is not one job. It's a host, plus a researcher, plus a producer, plus an editor, plus a promoter. I had hired exactly one of those people - me - and left every other seat empty. The work from those empty seats didn't disappear. It landed on my nights and weekends, and eventually it buried both shows.
Without meaning to, I had run a controlled experiment. Swap every variable, hold the org structure constant, watch the outcome repeat. The tools were present the whole time.
The seats were empty.
Now zoom out
In 2025, MIT's Project NANDA published a report called The GenAI Divide, and one number from it traveled everywhere: 95% of enterprise AI pilots were failing - producing no measurable return. You've probably seen that stat. It opens half the AI keynotes in America right now, and it's almost always followed by the same diagnosis: companies bought licenses and didn't train anyone. The fix, we're told, is better enablement.
I don't buy it. You can't train your way out of a design problem.
Those pilots didn't fail because nobody taught the team to prompt. They failed because the company made a software purchase when the situation called for a staffing decision. Somebody bought seats of an AI tool, dropped them into the existing tech stack next to the CRM and the project trackers, and waited for productivity to show up.
Nobody wrote the role. Nobody defined what the AI was responsible for, what it would hand off, who it reported to, or how anyone would know it was doing a good job. We would never onboard a human that way. We do it to AI every single day, and then we blame the technology when it wanders around the office accomplishing nothing.
So here's the premise I've come to build my entire business around:
AI belongs on your org chart, not your tech stack.
Not to replace the people you have - to finally staff the work no one was ever hired to do.
The work no one was ever hired to do
Every business runs on two org charts. There's the official one, which shows what people were hired to do. And there's the real one, which lives in everyone's calendar and shows what they actually spend their week doing.
The gap between those two charts is what I call Admin Drag - your best people losing most of their week to work they were never hired to do. The producer hired to build relationships who spends her afternoons re-keying data. The account manager hired to serve clients who spends his mornings assembling reports. None of that work appears on the official org chart. All of it is real, and somebody is paying for it in hours, margin, and burnout.
I lived this one personally. Even after my own shows died, I was hosting interview podcasts in my day job, and prepping for a single guest took me more than two hours - reviewing their onboarding form, studying their website and LinkedIn, choosing a topic, rewriting their bio, drafting the episode description and questions, packaging it all into a prep doc. I time-blocked my calendar for it. I dreaded it.
And here's the confession: when a guest canceled before I'd done the prep, I felt relief.
Sit with that for a second. A professional host, quietly rooting against his own show. That's what Admin Drag does. It doesn't just consume hours - it turns the work you love into work you hope falls through.
The hire that changed everything
In May 2024, I launched a new show - the one that became The AI Hat Podcast. And before my very first interview, I did something I had never done as a podcaster.
I wrote a job description.
I hired a Showrunner. Not a tool. A role, with defined responsibilities on both sides of the microphone. Pre-show, my Showrunner studies the guest's LinkedIn and onboarding form, recommends five angles based on their background and the show's premise, proposes titles, then drafts the intro, a custom guest bio, three takeaways, and ten interview questions. Post-show, that same Showrunner takes the transcript and drives the show notes, the blog post, the newsletter, and the social content.
Was it perfect on day one? No - and that's exactly the point. My Showrunner got onboarded with loose instructions, and after 20 interviews I sat down and did what any good manager does after a new hire's first quarter: a performance review. I formalized the role, tightened the responsibilities, and documented what we'd learned together. You don't do performance reviews on software. You do them on crew.
Now, the receipts. Interview prep went from more than two hours to about 20 minutes. Guests compliment the custom bio I read on air - almost every single one. And the shows that used to die at 12 episodes? The AI Hat Podcast just passed 120, and it's still going.
Twelve was my ceiling for five years. The tools never moved it. One job description did.
Say the quiet part
I know what some of you are thinking, because someone thinks it in every room I speak to: this is how the replacement starts. Put AI on the org chart, and eventually a human comes off it.
Look closer at my org chart. The Showrunner seat, the research seat, the production seat - whose name was crossed out to make room? Nobody's. Those seats were always empty. The work in them was never staffed; it was smeared invisibly across my nights, my weekends, and my dread. The Digital Crew didn't take anyone's job. It took the work that was slowly taking mine.
During the Apollo program, roughly 400,000 people worked in support roles so that three astronauts could fly each mission. Nobody watched mission control light up and said the ground crew replaced Neil Armstrong. The crew is what made the mission possible - and made the astronaut more valuable, not less.
That's what a Digital Crew is for. Not replacing your people. Giving them their jobs back.
Headcount discipline works both ways
Because AI sits on my org chart and not in my tech stack, I hold it to org chart rules. Every seat has to trace back to real Admin Drag in my real business. My Showrunner exists because prep and production killed two podcasts. My AI Chief of Staff exists because strategy work was eating my weeks.
And some seats stay empty. There's an Editor seat on my chart right now that I could fill tomorrow - the technology is ready. I haven't filled it, because I'm still scoping the role, and you don't hire for a seat until the work justifies it. Meanwhile my Showrunner is up for a promotion, taking on more of the publishing workload. Roles grow. Roles wait. That's how staffing works.
One caution here, because the market is about to get loud on this point. Every vendor in the space now sells AI employees - agents you can supposedly hire off the shelf, productive on day one. But an agent you didn't write the job description for is still a software purchase wearing a name tag. The role, the scope, the reporting line, the definition of done - that design work is yours, because nobody else knows where your Admin Drag actually lives.
Notice what that discipline solves. The reason so many AI rollouts turn into a hammer looking for nails is that a software purchase demands usage to justify its line item. A staffing decision demands a defined role to justify the hire. When you write the job description first, you are forced to examine the work itself - what it is, where it lives, whether it should exist at all - before AI ever touches it.
That's the part nobody tells you: writing the job description is the redesign. And it's exactly the step the 95% skipped.
Log 08-2026 · Industry Press
This argument, in the industry press
I published this here on July 8, 2026. Since then I've made the case twice more for the insurance industry, and each room needed a different half of it.
InsuranceNewsNet ran my piece on July 28, answering the question that follows me out of every keynote: whether AI is coming for the job. Part of it, yes, and it should. That part is the Admin Drag half, the re-keying and the form-filling nobody was hired to do, and the agents I meet who are most confident about their own future tend to be the ones who handed it over first.
Independent Agent, the Big "I" magazine, ran the other half on August 19. That one opens in a training room outside Pittsburgh, where we inventoried every AI tool already in use at the agency and the list ran screen after screen while the CEO watched. Adoption was going fine. What the agency did not have was anybody in charge of it. The piece walks through naming an owner, writing the policy, and moving one workflow from experiment to embedded.
Look at your real org chart
Here's what I'd ask you to do this week. Pull up your org chart - the official one. Then open your calendar, or your best producer's calendar, and compare. Every recurring block of work that doesn't map to a role someone was hired for is a seat that's been empty for years, with the cost hidden in plain sight.
Now imagine that seat with a name and a job description instead of a license and a login.
Because when you stop asking which AI tools to buy and start asking which seats to staff, everything about adoption changes - the rollout, the training, the ROI conversation, all of it. That's the work I do now, and it's the argument I'll be making from every stage I'm invited to.
AI belongs on your org chart, not your tech stack.
What's the first seat you'd staff?
