By Tony Russo, Mile Marker, Vice President of AI Innovation & Strategy
Our agency’s open roles right now skew senior. We need judgment on the floor today, so we’re hiring for it directly. In so doing, we’re choosing the near-term need while watching the long-term bill accrue.
Here’s why that bill exists. Every job early in my career started with work nobody wanted: pulling the reports, reconciling the numbers that didn’t match, building the decks. Tedious, low-value, and an agent could do all of it today in seconds.
It was also the training program.
You learn by reconciling two thousand numbers until the wrong one starts to feel wrong before you can explain why. Judgment isn’t taught, it’s accumulated, in the boring reps nobody enjoys and everybody wants to skip. And agentic AI is very, very good at exactly those reps.
So we as an industry must recognize and acknowledge that we’re handing the bottom rung of the ladder to machines. The pitch is the one you’ve heard at every conference this year: let the agents take the grunt work, move your people up to the high-value thinking, everybody wins. It sounds inarguable. It even sounds humane.
But the grunt work is the ladder. You don’t get to keep the senior strategist and automate the path that made them. Take away the two thousand boring reps and you haven’t freed the junior to do better work; you’ve removed the only way the junior was ever going to become the senior. We’re not just automating tasks. We’re automating the internship, and the assistants, and the junior planners and analysts, and calling it efficiency.
This is a challenge facing every business deploying agents to cut “manual” effort . What makes this all so tricky is that it doesn’t feel like a mistake while you’re making it. It feels like a win.
Output goes up, costs come down. The demo dazzles the client, boardrooms celebrate the “innovation.” None of the dashboards you actually watch will show you the thing you’re losing, because the thing you’re losing is a person who doesn’t exist yet: the senior, five years out, who never got built. Bench erosion doesn’t have a line on the P&L. You don’t notice the missing seniors until the day you reach for one and they aren’t there. By then, the fix is a decade out, because that’s how long judgment takes to grow.
The traditional apprenticeship produced three things as a byproduct, and they’re the three things the agent era can’t easily replace.
- Problem diagnosis: Finding the real friction point instead of the one the client named, a skill you only get from seeing enough messy briefs to know when a brief is misleading.
- Standard-setting: Knowing the difference between “good enough” and “great,” — a skill you get from having your own work torn apart until the bar lives inside you.
- Edge-case detection: Catching the one percent of the time the agent is fluently, confidently wrong. That’s the new hard skill, and the hardest to teach, because failure doesn’t look like failure. It looks like finished work. You can’t spot this unless you’ve done the reps to know what the truth looks like.
Agents can now do the tasks those three skills were learned on. Remove the task, and you remove the training. We’re not just losing labor. We’re losing the conditions that manufacture judgment.
You should still run the development pieces: university programs, internships, temp-to-perm, but they don’t replace what the old bottom-rung reps built. It’s a thinner apprenticeship than the one the agents are eroding, not a real substitute for it. That’s not a solution. That’s a tradeoff.
If you’re early in your career, the move is the same regardless of where you land: over-invest in judgment, not output speed. The market is about to flood with people who can do and starve for people who can direct.
And if you’re running a shop, understand the bet you’re placing whether you mean to or not. The firms cutting deepest into their junior ranks today are mortgaging their senior bench to do it. In five-to-seven years, we may be looking at an entire cohort who can operate the agents but never built the instinct to know when one is wrong. The firms that understood this early will be the ones left with people who can actually direct, putting them at a strong advantage.
Output is the commodity now. Judgment takes a decade to grow and one budget cycle to stop growing.
