Most leaders I talk to haven't drawn the line between what they own and what their systems handle. Not on purpose, anyway. They've drawn it by accident, usually around whatever their software happens to let them automate. Not around what they actually want to own.
That's backwards. And right now, in a year where everyone's bolting AI onto everything, it's the difference between a leader who scales and one who quietly becomes a bottleneck or a rubber stamp.
The line is where your judgment is the thing being paid for
Here's how I think about it. AI gets the drafts, the research, the scheduling, the monitoring. The first 80% of any artifact. I get who we say yes to. What we promise. How we recover from a miss.
That's the line. It's the place where your judgment is the thing being paid for. Not your typing. Not your ability to put a prompt into ChatGPT. Your judgment.
If you decide every small thing, you become the bottleneck and the whole system chokes waiting on you. If you decide none of it, you stop being useful. The work of leading right now isn't the people work, and it isn't the tool work. It's deciding which calls belong to you and which belong to your systems, and then defending that line on purpose.
I've drawn this line before. It just used to be people.
I'm not theorizing here. Before any of this was software, I did it real time. I was a drill instructor at MCRD San Diego. I ran systems-based training for America's sons coming into the Corps, and the honest job description was forecasting failure modes. I simply tried to find the thing that's going to break our performance, BEFORE it broke, and build the thing that catches it. Find the weak link in a squad before it broke us. Sometimes that was a recruit about to leave his footlocker unsecured, which is catastrophic. Sometimes it was something much smaller: his shoes not at the right angle, in the right order, under his rack.
So we had to build systems to find and catch these things. At the beginning, that was the drill instructors, constantly vigilant. Over time we'd train the squad leaders and the guide, the lead recruits in the platoon, to spot these things, and it would trickle down to the fire team leaders. Eventually every recruit, ideally, would develop a sense of self-reliance, and the group as a whole would have hard-earned teamwork.
DI school was just basic training for attention to detail, persistence, and the basic how-to-make-Marines of recruit training. Once I got to the Trenches at Lima Company, 3rd Recruit Training Battalion, I took some on the job training (OJT). Sometimes that looked like smoking a kid at every available opportunity because his sneakers weren't under his rack in the right order, exactly how he was taught. Until he realized the Marine Corps wasn't going to change for him. Until he met his own personal Jesus, soaked in sweat and agony, and thought: "You know what? I AM the problem, and I need to fix myself." Sometimes the system was breaking it down to the base level, sometimes fighting against years of habits and learned behaviors.
The best worst job I ever had. And it taught me the lesson I'm still running on: leadership is taking the time to teach a system, whether that system is people or software, how to achieve a goal. Organize it, direct it, inspire it toward something. Then know where it's going to fail before it does, because that part is yours.
The system was people then. It's software now. Same skill, different application. One mind, any domain.
"Isn't that just AI slop?"
I get this one a lot, so let me take it head-on, with the receipt.
When I first started having a system produce work in my name, it was clean and it was soulless. Technically fine. Empty. So I did the work most people skip. I fed it my own words. My journals, my recorded thinking, the things I actually believe and the things I'd never say. I taught it my no's, the phrases I reject on sight, the way I question a thing before I trust it. I learned to put what was in my head into words that could become action.
Now when something ships under my name, my judgment is already in it, because I put it there on purpose.
So is that slop? What even is slop. Slop is output with no operator behind it. No experience, no point of view, no one who can tell you why the system produced what it did, or catch it when it's wrong. The value that's missing from most AI content is exactly that: the experience of the operator, the person actually living it.
Mine isn't missing, because I did the work to put mine in. You're not paying me to put words into a model. You're paying me because I understand why the system is producing this output, and I can tell when it's wrong. That's the forecasting-failure-modes job again. Different uniform.
Why I drew my line this year
I drew mine this year, building Capers Ventures with Claude as my second brain. I wanted my perspective out at a speed I couldn't reach alone, to the people I can actually help.
Here's the line in practice. My system can pull every fact, draft every page, watch every number. What it cannot do is decide which source to trust, or know on its own when something it reported with total confidence is actually wrong. That call is mine. So I built the system to do what it does best, and I kept the judgment for myself. I have to allow the system to do what it does best, as long as I'm willing to free up, to uninvest my ego.
The world is not new, by the way. The tools changed. Teaching a system toward a goal, and owning the calls that require a human, did not. The best leaders I've watched could see the bigger picture and do the hard things, and that took a rigor most people don't want to commit to. It's easy to do what everybody else is doing. It's harder to say, I don't like this, and you're not going to like this either while we fix it, but when it's done it's a better answer.
I'm willing to fight hard to build something so I don't have to fight hard later.
Your move
Draw the line on purpose this week. Three steps, and none of them are technical. You don't have to be technical. You have to be structural.
- Write down, in one sentence, the decisions that belong to you and no one else, human or machine. That's your line.
- Feed your system your actual experience, not just prompts. Your words, your standards, the calls you've made and what they cost. That's what keeps the output yours instead of slop.
- Let everything below the line run without you. You still have to do the work of teaching it and turning it on. But you stop being the bottleneck.
That's not giving up control. That's the only version of control that scales.
If you want to talk through where your line should sit, that's the kind of thing I actually enjoy. Reach out.