I run a horde of AI agents and I no longer see the details. A CEO doesn't see them either. So why does it still feel wrong to me?
I haven't written a line of code in months. I've written about that before, along with the question of why AI tires me out differently. Today something quieter is on my mind: what is this actually doing to me? The speed has gone up massively. What used to take weeks is standing after a few days. And with the speed comes pressure that nobody puts on me except myself: to crank out more projects. It's possible now, after all. So why not one more?
I used to look at an idea for hours. Turn it over, play through scenarios, check use cases. At some point a small prototype would follow, and then I'd watch how the first users dealt with it. With Pyngup I spent a whole day with two test accounts and learned that a button can look perfect in the design and still feel off in real use. Today agents take care of details like that, and I'm losing my eye for them. Is that gone? Feels like it.
Micromanaging machines
Then again: nah, there's another thought, and I can't shake it. Say I work at a big company and suddenly get more responsibility. People below me whom I keep busy with tasks. I couldn't look at every detail anymore either. And if I did anyway, people would accuse me of micromanaging, and rightly so. I'd have my status calls, a bit of insight into what's happening further down, and I'd have to leave the rest to the people.
That's exactly what I'm doing now, just with different staff. I run a horde of AI employees: designers, UI/UX people, SEO experts, copywriters, marketing managers, SEA pros, developers. What that looks like technically is in the article on vibe coding. Each of them is busy with their own task, and I no longer see the details. Why should that be any more reprehensible than at a corporation?
Sure: the objection comes right away. "But at a corporation it's still humans taking care of the details." Ok. And what exactly is the difference? How people report upwards is up to me. I build the structure through which I keep an overview of the processes. An AI can report upwards too. If too much gets reported to me, I want to steer down into the details, and there I am, micromanaging again. If I let people do their thing, everything's fine: I'm not micromanaging, and I don't catch the details. That's true for humans just as much as for agents. So where's the point?
Maybe it's with me
It took me a while before it occurred to me that the point might not lie with the staff at all, but with me. I want the feeling of having everything under control. Knowing what's running, which processes are changing and which are just emerging. I know that feeling well. I once tried out 17 apps to finally get an overview of my own tasks. None of them gave it to me.
And now it actually gets funny. I'm still a sole proprietor, but with my flock of agents I'm sitting in the chair of a corporate leader. I can build layers of management, people for control, people for execution, an agent that checks the others and one that sums up for me in the evening what happened. In principle, nothing distinguishes me from a corporate manager for now. A one-man corporation, then. Wild.
So maybe the question is more of a philosophical one, and maybe it applies to both, the CEO and the lone fighter with a laptop: How do I find calm? How do I get clarity? How do I build a structure that I can keep an overview of and that I trust, without everything slipping away from me?
Nobody took the before away from me
While writing this, I noticed something. Further up I wrote that I used to look at an idea for hours and play through scenarios. Those were never details in the code. That was the before, the time before anyone writes a line. No agent took that away from me. I gave it up myself, because everything goes so fast now. Hmmm. Uncomfortable, but also good: what I miss, I can get back without firing a single agent.
For the CEO it's probably no different. Does the good boss read more reports than the bad one? I don't think so. I rather think the good one has cleared more time for thinking.
I once wrote down: tasks use exactly the space you give them. The sentence works the other way round too. If the agents build in three days what used to take me three weeks, there's time left over. I can fill it with seven new projects or with one that I take a proper look at. Where cramming everything in leads, I've written about in the context of burnout: as long as the system stays the same, it comes back, even with new tools.
What I want to try
- An idea stays with me for a day before an agent touches it. With a notepad, without a terminal.
- One report a day is enough, and it's one that tells me what has changed.
- Once a week I deliberately climb down into one detail, like a factory visit. A spot check on a fixed date isn't micromanagement to me.
- I watch the first users myself again. I still have the report on it written for me.
Whether that's enough, I don't know. I suspect that in the end trust isn't a question of structure at all. The structure just makes it easier for me to decide in favour of it. When sailing I once called this good seamanship: accepting mistakes and giving the crew another chance, again and again. Whether the crew is made of humans or agents, the wind couldn't care less.
By the way, I started this text exactly like that: with an idea I carried around for a while and turned over and over before anything was on the screen. Felt good. The laptop lid stays open now, otherwise the agents fall asleep. I'm heading out for a bit anyway.
Comments