'I Don't Have Time to Set It Up.' You Set It Up While You Use It.

I don't set up my AI agents first and use them later. I set them up while I use them. Every task I hand an agent, instead of doing it myself, teaches it something. That is the setup.

I made that point on a live training in late October, when someone asked how anyone finds the time for all this. Here's what I told them:

"A lot of people will say, 'Dude, I just don't have time to set it up.' And I'm like, what the flip are you talking about? Like I'm setting up as I use it."

I Could Have Fixed It Myself

Here's what that looks like on a normal day.

My agents run in Grokbot, and they have their own computer. "My day is spent a ton inside of their computer." That's not an exaggeration.

Not long ago, a hero section on a client website needed fixing. The background video and its overlay weren't going full width. I gave my agents instructions, and they worked it until the whole thing was ready.

And here's the part that matters. I didn't have to hand that off.

"I know I could have went and fixed that hero background. Easy. I'm a designer turned developer turned HubSpot nerd. But I spend time in their computer because everything I asked them to do on their system makes them smarter."

That's the whole trade. When I do the work myself, I get the fix. When I do the work through them, I get the fix and an agent that knows how to do it next time.

What They Learn Doesn't Stay on Their Computer

You might be thinking, okay, but what happens to all that learning?

My agents work from my main brain, the folder of plain files that holds everything about me, my business and my clients. If you haven't built one yet, start with What Is a Second Brain?

And the learning flows back. As I explained on the call, "anything it does in here, it knows to sync back to the main brain." So the agent "learned how to fix hero images in HubSpot themes," and "now it's just in its basic memory of understanding."

That has a second benefit. The tool can change. The learning stays mine.

Trust Gets Built the Same Way

This is the part I didn't expect.

"If I actually do the work through them, meaning I get them to do it, but I'm watching it as I'm doing other stuff, now all of a sudden I start to build, and ladies and gentlemen, this is gonna sound so weird, I start to build trust in the agents that I've hired to do the work that I need them to do."

You can't schedule trust. You don't get it from a setup weekend. You get it from watching an agent do real work, again and again, and seeing it get it right.

And once the trust is there, the ask gets small. "Now it's nothing for me to ask, hey, can you just go do this thing and let me know when it's done so I can double check it?"

"Wow, doesn't that sound like if you had an assistant, you'd be like, go do this thing so I can do the last 10% and go check it, but it can be an AI agent that is for that thing."

If waiting until it's perfect is what's holding you back, read Perfection and Control Are a Speed Limit. You want it perfect before you trust it, so you never trust it.

Watch It Work While You Work

You don't have to sit and stare at it, either.

On the training, I showed an agent adding a client's new podcast episodes, and their transcripts, to the client's community platform. "Meanwhile, I'm working on another of my five" monitors. "And it's actually working and I can watch it work because I can open its computer and literally watch it work while I'm working."

That's how the setup fits into a full day. The agent works on its screen. I work on mine. I glance over, I learn how it thinks, and I step in when it needs me.

I wrote about why that lowers my stress, not just my hours, in How AI Helps Me Stress Less.

Keep the Decisions That Matter

Doing the work through your agents doesn't mean handing them every call.

Take the short videos my team builds for my ministry articles. I told the agents to use their tools, "let our writer write the script, let our post-production guy look at the edits that need to be made, bring me the human in the loop decisions when you spend tokens or when you want me to redo a shot."

So when a new version comes back, I listen to it and say, "hey, redo, redo nine and 11." That's my decision. They do the rest.

If you want the three words I use for every one of those decisions, read Go, Tweak or Hold.

I have a task I could do myself, but I want you to do it so you learn how. Here it is: [the task].

Show me your plan first. Then do the work and let me know when it's done so I can double check it.

Bring me any decision that spends money or tokens, or that I'd want to redo. Don't make those calls on your own.

When we're finished, tell me what you learned and save it where you'll find it next time, so I don't have to explain this again.

That's the starter. The deeper conversation starters for building an agent team that learns as it works live in the AI Second Brain Workshop and my coaching.

Treat Them Like a Human Employee

I'll say it the way I said it on the call. "Again, I don't mean to sound weird, but the more I talk to these agents and treat them like they would be a human employee, the better results I get in what they're giving me and doing for me."

Think about a new hire. You don't spend a month writing their manual before they touch real work. You give them real work, you watch, you correct and you check. That's how they get good.

It's changing me, too. "The more that I pay attention to building a system that I can actually have do work and educate me along the way, the better results I get as becoming the human that I'm trying to become, which is an AI native who came from a one-room log cabin with no running water."

If you want to see how I hire and run these agents, start with Agents, Skills and Grokbot. And if you want the idea underneath all of it, read Human-Powered, AI-Assisted.

Who This Is Not For

If you want a finished, perfect system before you hand it a single real task, this isn't for you. You'll be setting up for a long time.

And if you want agents you never have to check, this isn't for you either. My ask is always the same: "let me know when it's done so I can double check it." The last 10% is still mine.

Frequently Asked Questions

Do I need to set up my AI agents before I use them? No. I set mine up as I use them. Every task I hand them, instead of doing it myself, teaches them something.

Isn't it faster to just do the task myself? Sometimes it's easier. I could have fixed that hero background myself, easy. But "everything I asked them to do on their system makes them smarter," so the next one is theirs.

How do I start trusting my AI agents? Do real work through them and watch them do it. Trust builds from seeing them get it right, again and again.

Where does what my agents learn go? Mine sync it back to my main brain. If I ever left the tool, what we learned together would still be there.

What should I still decide myself? Anything that spends tokens, and anything you'd want redone, like a shot in a video. Have your agents bring you those decisions.