AI Adoption Inverts the Pyramid

The displacement story says senior expertise absorbs the first hit from AI. The vendors' own data tells the opposite story: junior workers are the heaviest users inside adopting firms, and the senior-expert narrative does not survive contact with their own logs.

Chatterji, Holtz, Rakholia, Tambe, and Weeratunga's August 2026 working paper measured the inversion at scale, and the gap is the story. Across 1,500-plus organizations and more than 17 million messages logged on ChatGPT Enterprise, usage intensity peaks at the bottom of the ladder. Early-career adoption is the dominant signal, not a footnote.

The paper is vendor-sourced, with all the caveats that implies, and the authors flag the results as subject to change. The pattern is what survives those caveats. Junior knowledge work is the first category to be unbundled into AI-mediated tasks, because the work itself is closest to the model's strengths: writing, synthesis, communication, and the loops that absorb a new hire's calendar. Senior judgment stays scarce because it was never the bottleneck the model was built to replace. The repeatable move is the one to carry to the next story. When AI arrives inside a firm, the load shows up first where the work is most substitutable, which is rarely where the credentials sit. The next time a working paper leads with senior displacement, ask who actually typed the prompts.