MAVERICK GROUPAI Book a 30-min call

All issues

September 2026 · 4 min read

Building a Team of AI Employees

By Carlos Pelayo, Founder of Maverick Group AI

A desk monitor showing a dashboard where a task moves between AI team member portraits before reaching a person at the end of the flow

A client asked me last month how we'd flagged something in their account before their own team even noticed it. I didn't have a clean answer, because I hadn't been the one watching. One of the specialists on my AI team had caught it hours earlier, quietly, while I was in a completely different meeting.

That's not the impressive part of what I've been building. It's actually the boring part. And that turned out to be the whole lesson.

A few months back I wrote about the job description nobody writes for their first AI employee: narrow the job, define one metric, let it own that lane. That advice still holds for your first one. What I didn't say then is what happens once you're past your first one, because that's where I got surprised twice.

The first surprise: it's not about splitting tasks smaller. It's about writing a real job description.

My instinct, once I had more than one AI employee, was to keep going narrower. Give each one an increasingly specific, increasingly small task, the way you'd keep slicing a pie. More AI employees should mean thinner and thinner slices.

That's not what actually works. What works is closer to how you'd build a human team: you don't hire someone to do exactly one atomic action all day, you hire someone for a role, a discipline, a defined set of related work they own end to end. A marketing hire doesn't just write one kind of email; they own marketing. An ops person doesn't just file one form; they own ops. I built my AI team the same way once I stopped trying to slice tasks and started writing actual job descriptions instead. One owns a discipline of research work. One owns a discipline of client communication drafting. One owns watching for a specific category of signal across everything else. Each has a real lane, not a single robotic action.

I didn't expect that the fix for "this AI is trying to do too much" wouldn't be smaller tasks. It would be a better job description.

The second surprise: the payoff wasn't where I expected it.

I assumed the win would show up in the flashy places. Big analysis, client-facing output, the kind of thing you'd want to screenshot. Some of that happened. But the actual difference in my week hasn't come from any of it.

It's come from the pile of small, recurring, nobody-wants-to-do-it work that used to eat an hour here and twenty minutes there, all week, every week. Watching for something that needs a reply before it goes stale. Pulling numbers together the same way every time. Catching the thing that would've sat unnoticed for three days. None of that is interesting to talk about. All of it used to be mine.

I didn't build this team to hand off interesting work. I built it, without quite meaning to, to hand off the tedious kind. That turned out to matter more.

What it's actually changed

I want to be honest about what this is and isn't. It's not a system I set up once and walked away from. I'm still tuning what each one owns, still deciding where I want a human in the loop before anything happens and where I don't, still learning where one job description needs to split into two. That part hasn't stopped, and I don't expect it to.

What has changed is what I spend my own attention on. The recurring stuff that used to sit at the bottom of my list, the stuff that never felt worth blocking time for but always needed doing, mostly isn't mine anymore. That's freed up more of my week than any single big win has.

But the real shift didn't happen when I had a handful of AI employees each doing their own job well. It happened when they started working together.

One picks something up, hands it to the next with the context already attached, and a third checks the result before anything reaches me. Nobody has to loop me in to pass the baton. A stack of separate tools each do their own thing and stop. A real team moves work between each other, the way people on an actual team do, and mine is built around how my specific business runs, not bolted on as something generic.

That's the part I didn't see coming when I started. I thought the value was in each individual hire. It's actually in how they work together.

If you're a business owner thinking about something like this, my honest advice hasn't changed from last time: start with one job description, well written, for one AI employee. But don't stop there. The bigger payoff shows up once you have a few of them and connect them, so they hand work to each other the way your own team would.

If you want to talk through what that could look like for your own business, reach out. I'm always glad to compare notes.

Ready to build your own team?

If you want help figuring out which job comes first and how to grow from there, book a 30-minute intro call. You'll get an honest read on where AI fits your business and where it doesn't, and you'll leave the call knowing exactly what comes next.

Book a 30-minute call

No pitch deck. An honest read on where AI fits.

Get the next article in your inbox

New articles land with subscribers first. No filler, unsubscribe anytime.