My AI is a brilliant fresh graduate. So was I, once.
Twenty years in, and I finally have the output I could never afford as a one-man shop. Same hours, ten to a hundred times more work getting done. I should be thrilled, and most days I am, but I’ll be honest about the part nobody warns you about: it didn’t make the days lighter. It moved the wall.
The way I get it is by onboarding AI agents the way an agency onboards juniors. I keep oversight, I give direction, I say what to do and, more often, what not to do right now, and I clean up after them. Some days it feels like the most capable team I’ve ever run. Other days I catch myself not knowing whether I’m extending myself or slowly replacing myself, and I leave that question open, because I genuinely don’t have the answer yet.
The thing about a junior, any junior, is this. They drive perfectly by the rules, and the rules are not the road.
Last week I was scaling a Manual CPC account up. Beyond raising the daily budget, the bids had to go up too, for higher positions and to catch more of the traffic that was sitting just out of reach. The AI raised the ad group bids. Textbook, exactly what the manual says. Except in Manual CPC the ad group bid is only the fallback for the keywords that don’t have a bid of their own, and most of the keywords that actually spend on that account had their own bids set months ago. Every one of those ignored the lift. The keywords doing the work never moved. Half the scale-up was cosmetic.
So I sent it back to do the thing that actually mattered, lift the bids at the keyword level, where the money is. The textbook move and the correct move were not the same move, and the only reason I knew to check is twenty years of watching that ad group number quietly fail to do what the interface implies it does. Fresh graduates think they know it all too. I certainly did.
That is most of my day now. Not doing the work, but reading what came back, finding the one move that was right on paper and wrong in the account, and sending it back. It is quieter than the old job, and somehow heavier.
The clearest proof that twenty years isn’t a fixed rulebook is that Google keeps editing the rulebook under me.
There was a time you set a campaign live and got traffic inside thirty minutes, sometimes faster. Then automated bidding came along with its learning period, and that quick certainty stopped being the default. The rulebook my AI memorized is a snapshot of this year’s Google. Mine got rewritten more than once, and surviving the rewrite is a different skill than knowing the current version. It’s the same instinct behind every time I’ve ignored Google’s own on-screen recommendation: those are written for the average account, not for yours.
For a while I thought that gap, the road versus the rulebook, was the whole story. The reassuring one, where the experienced human stays a step ahead. I don’t fully believe that anymore, and the reason sits closer to home than Google.
The bottleneck in my work stopped being my hands a while ago. It became my judgment, and, if I’m honest with you, my energy. A few hours of directing confident juniors and cleaning up after them leaves me more drained than doing the work myself ever did. There’s a finite number of good decisions in me on any given day, and the AI doesn’t refill that number, it spends it faster. So I’ve started treating that finite judgment as the thing I reserve, deliberately, for the road-not-rulebook calls that nothing in a manual can make for you.
I didn’t apply for this role, by the way. The part I built a career and a reputation on, the doing it with my own hands, is now the part I hand off first. And if you look at any company with people in it, a senior directing juniors and catching the confident mistakes, isn’t this just that, wearing new clothes?
I won’t pretend to know how this ends. I used to tell myself the junior was becoming a medior in front of me, leveling up week by week. Rereading my own accounts, I’m not sure that’s what’s happening. It feels less like a colleague climbing toward independence and more like an extension of me, and an extension is never better than the person it extends, on their most tired day. So I honestly don’t know whether I’m training my replacement or just building a longer reach for the same two hands.
I still think that for genuinely good work, the kind a client pays a premium for, my input will always be needed in the loop. Then again, that is almost certainly what every senior believed right before the road changed under them. My last famous words, maybe.
The one number I’m sure of is the one I can’t 10x: the good decisions I’ve got in me before dinner. Ask me again in a year, and if I still sound this certain, hold these words against me.