The Unannounced Change
Partnership criteria at major law firms are not published. They are communicated through informal signals, mentorship conversations, and the slow accumulation of feedback from reviews. This is, for our purposes, a feature: it allows criteria to change without announcement.
What we are documenting in this piece is a shift that has been underway for approximately eighteen months and that most associates at the affected firms cannot yet articulate clearly, even though they sense it. The skills that earned partnership in 2020 — and that partnership committees still nominally say they are looking for — are increasingly insufficient. New criteria are being layered on top of the old ones.
The Three New Competencies
Based on interviews with twenty-three partners at AmLaw 100 firms who are actively involved in associate evaluations, three competencies are consistently emerging as differentiators for the partnership track in the AI era.
First: AI workflow fluency. Associates who can design and manage AI-assisted workflows — not just use AI tools passively, but actively configure, prompt, and evaluate them — are producing substantially more client-facing output at the same billable time. Partners notice this. "An associate who brings me a first draft of a thirty-page agreement in two hours instead of two days is demonstrating the same judgment I care about, packaged in a different delivery mechanism," one partner told us.
Second: Client communication on AI. Associates who can explain to clients what AI did and did not do in their matter — clearly, without either over-promoting or under-informing — are building client trust in a way that is increasingly visible. "The clients who ask about AI are our most sophisticated clients. They want their lawyers to be fluent, not defensive," a capital markets partner observed.
Third: Quality detection. The ability to read AI output and identify its failures — not just obvious hallucinations, but structural weaknesses, missing counterarguments, doctrinal blind spots — is emerging as a core attorney skill. Associates who have developed this skill can work faster and maintain the quality bar. Those who have not are creating liability in both directions: over-reliance on AI they cannot evaluate, and under-use because they do not trust outputs they cannot verify.
What This Means for the Current Associate Class
The associates who are well-positioned for partnership in five years are the ones who are actively building these three competencies now, not waiting for their firm's training program to deliver them.
The practical implication: treat every AI-assisted work product as a training opportunity. When you use AI to draft a section, go back after the matter is closed and evaluate what the AI got right and wrong, and why. Build your own taxonomy of failure modes. This is the skill that partners are developing through experience — associates who develop it proactively will arrive at partnership readiness faster.
The associates who are at risk are the ones who have drifted into passive AI use — who use the tool because their firm provides it and senior attorneys expect it, but have not developed the critical evaluation layer. They are producing output efficiently but not building the judgment that partnership requires.