The legal profession has always been built on billable hours, and no task generated more of them than discovery — the grinding process of reviewing thousands, sometimes millions, of documents to find the handful relevant to a case. For decades, this was the domain of bleary-eyed junior associates and contract attorneys, hunched over boxes of paper or clicking through endless PDFs at three in the morning. That world is vanishing faster than most clients realize.
Large language models and their predecessors have fundamentally altered how law firms approach the most labor-intensive phase of litigation. What once required teams of dozens working for months can now be accomplished by a handful of attorneys supervising AI systems that never tire, never miss a keyword, and increasingly understand context in ways that would have seemed like science fiction to lawyers trained even a decade ago.
The economics of obsolescence
The numbers tell a stark story. Document review traditionally consumed somewhere between half and three-quarters of total litigation costs in complex commercial cases. Law firms charged premium rates for this work, and clients paid because they had no alternative. Technology-assisted review, as the industry euphemistically calls it, has compressed those costs dramatically while simultaneously improving accuracy.
The implications ripple outward in uncomfortable directions. Contract attorney positions — once a reliable if unglamorous entry point into the profession — have contracted significantly. Law schools continue producing graduates at roughly the same rate, but one of the traditional absorption mechanisms for legal talent has fundamentally changed. The associates who remain in discovery work have seen their roles shift from readers to supervisors, training and correcting AI systems rather than doing the reading themselves.
What the machines actually do
Modern legal AI operates on principles that would be familiar to anyone who has used a sophisticated search engine, but with crucial refinements. The systems learn from attorney decisions about relevance, building models that can predict how a human reviewer would categorize documents they have never seen. They identify conceptual relationships that keyword searches miss entirely — understanding, for instance, that a discussion of "the project" in one email chain refers to the same matter called "Phoenix" in another.
The technology excels at consistency in ways humans cannot match. A tired attorney reviewing their five-hundredth document of the day will inevitably make different judgment calls than they did on document fifty. The AI applies the same standards to document five million as it did to document five. Courts have increasingly accepted these systems, with judges noting that technology-assisted review often produces more defensible results than traditional human review.
The expertise paradox
Here lies the profession's uncomfortable truth: the work AI handles best is precisely the work that once trained young lawyers to become good older lawyers. Reading thousands of documents in a complex case taught associates how businesses actually operate, how executives communicate, how deals go wrong. That institutional knowledge, absorbed through tedium, informed the judgment that made senior partners valuable.
Law firms are still working out how to develop talent when the developmental grunt work has been automated. Some have created artificial training programs, others rely more heavily on mentorship, but no consensus has emerged on how to build the next generation of litigators when the traditional apprenticeship model has been disrupted.
Our take
The legal profession's AI transformation is neither the apocalypse some fear nor the utopia others promise. It is something more mundane and more significant: a fundamental restructuring of how legal work gets done, who does it, and what skills matter. The lawyers who thrive will be those who understand both the technology's capabilities and its limitations — who can train the systems, interpret their outputs, and know when human judgment remains irreplaceable. The profession will survive, but it will not look the same, and the young lawyers entering it today are training for jobs that may not exist by the time they reach partnership.
