Hey friend,
I've been chewing on something you said a while back. You told me you only give AI about an hour a week, and you asked how you're ever supposed to learn it that way, and whether it's even worth trying.
I want to answer that honestly, which means this is tough love. Emphasis on the love. If today isn't the day for it, put this down and come back to it later.
Short version: eventually an hour a week is all you'll need. You can't learn it in an hour a week.
That gap is the whole thing.
You describe yourself as spread wide and thin. Marketing you picked up over the years, people development, process, training, writing, some design. Nothing you'd call mastery. I'm a generalist too, so I know exactly what that self-assessment feels like from the inside.
I think you've got the read backwards.
A generalist understands a lot of things at a foundational level. Call it 80% of the way to expertise in each one. That was a liability back when the market paid for depth. With AI in the mix, I'd rather have your resume than a specialist's.
Here's what I mean when I say "use AI," because I don't mean asking Claude a question and reading the answer. I mean training it. Building context, building a corpus. I've called it a brain when we've talked about it. And then directing the thing.
The conductor doesn't play the violin. They point at the violin when it's time for the violin.
They know when the woodwinds come in and whether the percussion is drowning everybody out. They can read the music for every instrument in the room, they know the caliber of each player, they know exactly where each one is weak. Then they just direct.
The conductor is the ultimate generalist, and it's a seat you're already qualified for. You won't sit in it without spending gobs and gobs of time first.
Think about it like onboarding, because that's really what it is. You bring a new associate in, you walk them through the process, you fix their first few passes, and you spend real hours doing it. Then you stop. They report to you, they tell you what they ran into, you adjust.
You don't stay in their business forever.
Claude shows up like that on day one. It already knows a lot of general stuff, it'll analyze something and tell you what it thinks about it. What you actually want is to hand it a job and trust that it comes back done, and that only exists on the far side of the training.
An old boss of mine used to say your whole job is to train someone to replace you. Same job here. Train the AI to replace you.
Let me show you what that looked like for me, because I'm not describing something theoretical.
A client sends me a reference video she likes. I go try to make the thing myself, and I usually land close but not exact.
So I break the reference down the way I'd break it down in my head. Colors, pacing, what kind of voiceover, what music, the transitions, how long it runs, how it made me feel watching it. I hand all of that to Claude along with what I think she's responding to. Then I ask her what she likes about it, she tells me, and I feed that in too.
For a while I'm the middleman between the client and the system. That's the training, and it's not more complicated than that.
Then I'm not needed for it. She sends a reference now and my system breaks it down the same way I would, and Claude edits my videos. Something that took me eight hours of teaching started running on its own in days. Not weeks. Days.
You've got that exact loop sitting right there in your client's feedback. If Claude helped you build the thing in the first place, and it knows what you sent and what came back, it can process her notes and make the changes.
I hear a version of this constantly, and I wonder if you're waiting on it too: it's going to get so good that I won't have to learn it, I'll just talk to it.
That's a lie people find comforting, and it isn't coming.
However good the models get, the system still has to know you. Your process, your standards, what "done" looks like when you're the one looking at it. Nobody hands you that part. Better models don't reduce the hours, because the hours were never about the model.
I'm not going to undersell what this cost me, either. I don't want you walking in thinking it's free.
I sold everything I had and closed my old business. I ditched every past client and planted a flag in something completely new, at what might be both the best and the worst possible moment. AI has the worst reputation it has ever had. You get treated like a villain for working in it.
Finding the first hour was harder than all of that.
I had a standing weekly lunch with someone for years. An hour and a half, and I always walked out feeling better about myself. A cheerleader, the person who told me I was doing everything right. They also didn't believe in AI, and I could feel the judgment land before I got through a sentence about it.
I traded that hour. It didn't appear on its own. I took it from somewhere that hurt.
So here's the tough love part.
When somebody tells me they don't have time to learn something, it's the most boring excuse I've ever heard. I don't think you're lazy. I think the pain hasn't reached the point where you're willing to sacrifice something for the new thing, and maybe that just has to happen on its own schedule.
But I believe with every part of me that there's a point coming where not having these skills makes you unemployable.
So spend the money on Claude Code, give it one real job you already do every week, and train it the way you'd train a new hire. I already wrote up the mechanics of teaching it your voice and I'll send you that. The hours are the part I can't send you.
You have the brain and the range to dominate with this technology. That's the part that bugs me about the hour.
I love you, my friend. Tell me which job you pick.