The Thomson Reuters Australian legal market update for the first half of 2025/26 exhibited a remarkable finding: non-equity partners recorded a third straight productivity decline, and hours worked by junior and mid-level associates were also slightly down, while senior associates and equity partners logged more hours to keep total numbers stable. The report's own gloss on this is that GenAI is now widely being used for research, drafting, and document review, which are precisely the tasks the bottom of the pyramid used to do.
Michael Legg at UNSW has been making a related argument for some time, calling it the 'flattening pyramid'. The traditional firm structure, with many juniors at the base, fewer partners at the top, and a leverage model that turns associate hours into partner profit, assumes that routine work has to be done by humans who climb a ladder of competence over a decade. AI breaks that assumption.
Equally interesting is what the productivity data does not yet show. Total hours are stable, and the efficiency gains AI is producing are not yet flowing through to clients in the form of lower bills, which is why research from Axiom suggests firms are not yet feeling pressure to drop their rates for AI-assisted work.
A traditional law firm's economics depend on a wide gap between what it costs to produce an hour of legal work and what that hour is billed at, and the gap is widest at the bottom of the pyramid. When AI compresses or eliminates the work those juniors used to do, it changes who is producing the margin, and the answer is increasingly the senior associates and partners themselves, who are both more expensive to deploy and harder to scale.
This is the part of the shift that is structurally awkward for firms. A managing partner can adopt AI tools quickly, but they cannot restructure a partnership compensation model, a graduate intake program, or a client-facing pricing schedule on the same timeline. The technology moves in months, while the institutional arrangements move in years. Firms that are not aware of this will discover it through a slow erosion of leverage that becomes visible in profitability data two years after the cause has set in.
For firms thinking about this seriously, the question is no longer whether to adopt AI, but how to adopt it in a way that does not collapse the leverage model before a new pricing structure is in place. The firms that figure out value-based or outcome-based pricing for AI-assisted work will set the pace. The firms that do not will spend the next decade defending billable hours against clients who can see what their work actually costs.
At Habeas, AI is not a productivity tool sitting on top of an unchanged business model, but a forcing function on the business model itself. We grew with AI rather than before it, which is why our pricing, our team structure, and the way we deliver legal research were all designed for a world where the bottom of the pyramid is no longer where the work happens. The firms that thrive over the next decade will be the ones that make the same choice deliberately, before the market makes it for them.
