AI is your junior colleague

I've been managing people for about a decade. Working with AI has made me a better manager—not because AI taught me anything about leadership or empathy, but because it forced me to get specific about what I actually want.

When you hand a task to a person, shared context does most of the work. They've been in the meetings. They know the culture. They know what you mean by "make it clean." AI has none of that. Every task starts from zero, so you have to think about what you're asking for, what good looks like, and what matters before you hit send. That's management training, whether you realize it or not.

I first noticed this when I moved Western Gallery from WooCommerce to Shopify. This was before Claude Code—just a chat window. I needed to customize the Shopify theme in ways that would've meant hiring a developer or spending months figuring it out piece by piece myself. In fact, I'd been working on it for close to two years, a piece at a time, when I could fit it in. Instead, I described what I wanted to Claude, reviewed what it gave me, told it what was wrong, and we went back and forth on Slack. The whole thing launched in a week. Did every code snippet come out perfect right away? No. Were there instances where there were 32 revisions I had to go through? Yes. But it got done quickly, and I was able to go live.

The thing that surprised me wasn't the speed. It was that I was doing the same thing I'd do with a new hire—setting scope, checking work, giving feedback, redirecting when it went sideways. The only difference was the feedback loop took seconds instead of days.

This got even clearer when I was directing creative at Glide. My process there had a lot of steps before a designer touched anything—I'd lead discovery calls with clients, put together creative briefs, build wireframes, write the messaging. AI became useful in that pre-handoff phase. There were times I needed to show specific interactive functionality so the designer and developers could understand what I had in my head. Instead of trying to storyboard it on paper, I'd write a prompt, take the output to Figma Make, and get something eighty percent of the way to a real wireframe. The designer styled it up. The developer already had most of the front-end code from the Figma Make output. What used to be a multi-day game of telephone turned into one afternoon.

But AI wasn't replacing anyone in that chain. It was speeding up the handoff between my head and theirs.

Where it fell short is where many juniors do: it doesn't know when to stop. The messaging drafts were always too long, too generic, trying to cover every angle at once. The wireframes needed trimming, rearranging, and augmenting in places. I'd generate a few versions, pull the best parts from each, add what I thought was missing, stitch them together, and then edit another layer on top of that. Less "accept the output" and more "direct it, curate it, fix it." Which is just management.

When I first started managing people at UT Dallas, I had student workers. It was my first time supervising anyone, and the thing that hit me was the range. Even after I'd filtered by 1000x during hiring, I had one student—a junior in computer science—who could not do the most basic thing on their own. Not lazy. Just a gap between what was on their resume and what they could actually do. I had another student who's probably the best employee I've ever worked with. I'd hire him again tomorrow. I'd work for him if the project was right. But most of the students who came through fell somewhere in the middle. Capable but inconsistent. Fine with clear direction, overwhelmed without it. That's where AI is. An average junior employee.

Of course, the output isn't perfect. The hype wants you to believe you're getting a senior employee for free. You're not. What you're getting is a hundred times faster than your worst hire, ten times faster than a decent junior, and about the same quality. That math is still incredible. But you still have to manage it. You still have to know what you want. You still have to catch the mistakes and bring the thing it can't bring for itself—the sense of what's right that only comes from having done the work and lived with the consequences. And the major drawback is that they stay a junior. They don't learn enough to grow into more responsibility.

The people I see struggling with AI are usually struggling with delegation. They hand off too much with too little direction and wonder why the result is bad. Or they refuse to hand off anything because they can't stand imperfection. Both of those are management problems, not technology problems.

The habits I built managing AI have continued to make me a better manager of people. I'm clearer in my briefs now. I'm more specific about what "done" looks like. I give tighter, more useful feedback instead of vague dissatisfaction. AI didn't teach me those things—but the reps did. Many cycles of prompt, review, revise, compressed into weeks. It built the muscle in a way that years of managing people hadn't quite finished building.

So if you've been avoiding AI because it feels like cheating — try thinking about it differently. You're not replacing yourself. You're making your first hire.