Podcast host and CEO of JMan Seminars, Jeremias “JMan” Maneiro, argues that a machine writing chapter markers is ok but a machine writing the thing you said isn’t ok.
Something goes wrong around week three.
A host discovers AI, and the beginning is the honeymoon phase. A new tool that you are so excited about, and you can’t wait to free up so much time with AI automations.
Show notes appear like magic. The transcript lands before the kettle has boiled. Then one afternoon in week three, they listen back to an episode and something is off.
The cold open sounds like a cold open. The questions sound like questions. Every piece of it is technically fine, but none of it sounds like them.
Most people quit right there.
I have trained more than 50,000 professionals on practical AI, and I host one show and co-host another, so I have watched this from both sides.
The pattern never changes. Adoption in the room is close to total. Then it falls off a cliff somewhere between day fourteen and day thirty. The tools did not fail. People aimed them at the wrong half of the job.
Here is the half they should be aiming at.
Nobody subscribed to your podcast for your show notes. Nobody subscribed for your chapter markers, your transcript, your episode titles, your metadata, your clip selection, or the ninety minutes you lose every week turning one episode into eight social posts.
That is assembly work. It is real work, it eats your evenings, and it is completely invisible to the listener. Hand every bit of it over.
For a two-person show, that is not a small savings. Transcription and clip finding alone gives you back the better part of a day. Feed the transcript back in, and you have a first draft of show notes in the time it takes to make some tea.
Guest research that used to mean an hour of scrolling becomes a five-minute briefing you will actually read.
On my own shows, the transcript is the hinge. Everything downstream in your workflow comes from it, the notes, the clips, the titles, the posts, it all comes out of that one file, so a single good input with a skill that you create in Claude Cowork or ChatGPT feeds a week’s worth of output. None of it touches the show. All of it touches your calendar and saves you valuable time.
Then there is the other pile, and this is where shows quietly lose momentum.
People subscribed for you. For the way you set up a story, for the question nobody else would think to ask, for the thing you say when a guest surprises you.
The AI-written script cannot actively listen and ask better questions. A good host doesn’t check boxes down a list of questions that they have to finish by the end of the episode.
The moment AI starts writing your cold open, drafting your questions, picking your topics, or smoothing your delivery, you are not saving time anymore. You are trading away the main reason why your audience was tuning in that cannot be copied or duplicated. You.
I call this voice drift, and it is sneaky, because no single episode sounds wrong. Week one you sound like yourself. By week three, the tool has sanded off the edges, and the edges were the entire product.
Your unique value is what makes your podcast, YOUR podcast. Your regulars will not write in to complain. They will just stop finishing episodes.
There is a simple test that sorts the two piles, and it works every time. If a listener found out AI made this, would you be embarrassed?
Nobody is upset that a machine wrote your chapter markers. Everybody is upset that a machine wrote the thing you said. That line has nothing to do with platform policy or disclosure rules. It is about what people came for.
Which brings me back to week three.
The reason so many independent hosts give up on AI is that they start at the end. Writing is the hardest part of making a show, so it is the first thing anyone tries to hand off.
It produces something that reads fine, feels wrong, and leaves you faintly embarrassed. So you close the laptop, decide the whole category is all hype, and never get around to the boring wins that would have handed you five hours a week for the rest of your life.
Start at the boring end instead. Automate one piece of your work this month, just one, and leave everything with your fingerprints on it completely alone.
When that piece is genuinely saving you time, add another. You will end up with a stack you actually keep, which puts you ahead of most of the people currently telling you how to use AI.
You can start by breaking down your systems and processes into micro tasks, so you know exactly what it is you do with your podcast. Some of us just put in the work and don’t document the process. It’s the critical step that many people miss.
The rule is short enough to remember on a bad day. Automate the assembly. Never the voice.
Jeremias “JMan” Maneiro is the Chief Edutainment Officer of JMan Seminars, host of the AI Agent Advantage podcast and co-host of Outside the Corporate Box. He has trained more than 50,000 professionals on practical AI, is a co-founder of Big Brain Chatbots and the creator of AI-Cademy, and is a licensed associate real estate broker in New York.
Website – https://jmanai.com
Linkedin – https://www.linkedin.com/in/jmansells/
