Most pitches for enterprise AI promise a leveler. Lift the bottom of the team. Close the gap. Make the median produce like the senior.
What's actually happening runs the other way. AI amplifies talent. The strongest performers on a team gained more from the rollout than anyone else, and the gap between them and the rest widened beneath the polish. The math on a great hire shifted accordingly.
It's not a floor raise. It's a ceiling lift. And the people sitting under the new ceiling are exactly the ones the org chart already had ranked at the top.
Where the Lift Lands
For senior knowledge workers, the slow part of the job was never the first eighty percent of a draft. The slow part was the last twenty: the version that goes to a regulator, a partner, a board. AI compressed everything before that twenty percent and barely touched what came after.
A corporate development analyst used to spend two days assembling a target overview before the director would see a draft. With AI assistance the assembly takes a few hours. The director who used to make one directional call on Wednesday morning now makes four across Tuesday afternoon and Wednesday morning, each pass tighter than the one before. The analyst is faster. The director is operating on a different cadence entirely.
The senior didn't get faster at typing. The senior got more chances to think before the work shipped.
What the Amplifier Doesn't Touch
Production cost collapsed. The cost of evaluating production did not.
A senior underwriter reading an AI-drafted credit memo catches that the sourcing on a comparable transaction is from a sector that doesn't quite map. The memo reads cleanly. The analysis is plausible. The recommendation looks defensible on its face. Seven minutes in, the senior flags two specific points and sends it back. Fifteen years of writing memos built that catch. AI didn't supply it. AI just removed the wait between drafts.
That accumulated catch-rate is the part of seniority that has to be earned. AI amplifies talent that's already calibrated. Where calibration is missing, the amplifier has nothing to multiply.
What the Polish Stops Telling You
Polish was always a proxy for skill. It got cheaper to produce.
A pricing analyst's recommendation now reads as cleanly as a director's. Structure holds. Tone is right. The numbers tie. What the prose can no longer carry is the writer's grasp of why this conclusion and not the next one over. The off-key sentence that used to give a senior reviewer a reason to slow down is gone. The reasoning behind the polished sentence is whatever the writer brought to the prompt.
Most review processes were never designed to test underlying judgment directly. They tested visible craft as a stand-in. The stand-in stopped being reliable. Review either gets sharper and slower, or it gets faster and looser.
The teams that already had strong reviewers, the ones reading for reasoning rather than craft, lost less in the transition. They still have something the prose can't fake. The teams running on craft-as-proxy lost a calibration tool they hadn't realized they were using.
What Compounds at the Top
Strong performers extract more from AI than anyone else, and the gap widens with use.
They notice where the model tends to drift, update their prompting, and stop falling for the failures that catch newer users. Each correction is a rep on their judgment. By month nine of a rollout, the senior is not only doing more work, they are getting better at framing what to delegate to the model and what to keep close. AI amplifies talent that already had something to multiply. Compound interest only works with principal.
The numbers on retention shift downstream of this. The cost of losing a senior was always understated, because the comp band priced the headcount and not the calibration. The calibration just got more productive. Replacing a senior now means replacing not a role but a few years of compounding judgment that AI made faster but did not create.
A handful of teams have noticed and adjusted quietly. Most still treat the senior as one head, indexed the same way they were before the amplifier showed up.
What's Slowly Going Missing
The second-order effect is the one the org chart misses.
The path that produced this generation of senior performers ran through unaided reps. The mid-level analyst wrote the first draft, got torn up in review, and learned from the bleeding. Polished output from the start means less prose to bleed on. Less prose to bleed on means less learning per cycle. The reps that built taste are getting fewer, on the teams that need the next batch of seniors most.
This isn't a one-year problem. On a five-year horizon it's structurally significant. Seniors today are getting more reps; juniors today are getting fewer of the right kind. Calibration accumulates unevenly, and the unevenness shows up later, when the firm tries to promote into a role nobody has been training for.
Strong teams will figure out how to put the formative reps back. They'll design review to read judgment, not just craft. They'll invest in the friction that used to happen on its own. Most teams will not, because the dashboard says throughput and throughput looks great.
Where That Leaves the Math
The upside is bigger than most rollouts have priced in. The best people on the team are doing more of their best work than they ever have. The signal of who they are is sharper. The cost of losing them is higher. Retention, recruiting, and internal mobility all shifted in the same direction at once, and most comp bands haven't caught up.
AI amplifies talent that's already there. Whether the rest of the team rises with it is a different question. The amplifier won't answer it.