By Ivy B. Grey
Every conversation about artificial intelligence in law eventually arrives at the same destination: efficiency. Faster drafting, faster research, faster turnaround. To be clear, this productivity narrative is real, but a concerning story is emerging that firm leaders are not yet taking seriously enough.
Legal AI is creating a writing problem. Not the hallucination problem you’ve already read about but a different one: The profession is improving its output speed while systematically weakening the thinking that has always given legal writing its value.
Writing has always been about thinking
In law, writing is not the end product of thinking—it is the process by which thinking happens. When a lawyer works through a brief, memo or contract, they are not transcribing conclusions they have already reached. They are discovering them. The friction of putting an argument into words forces lawyers to clarify their thinking. That struggle is a feature, not a bug, in the legal workflow.
Cognitive science supports this principle. Research by Linda Flower and John R. Hayes demonstrated that writing is a recursive, goal-driven process where ideas are generated, evaluated and revised in real time. In other words, writing produces thinking, rather than merely recording it.
Nancy Sommers echoes this in her study, “Revision Strategies of Student Writers and Experienced Adult Writers,” noting that student writers see revision as linear and superficial, while experienced writers use revision to rethink their initial decisions and redefine their goals. In “The Revision Process in Legal Writing: Seeing Better to Write Better,” Christopher M. Anzidei criticizes the “assembly line” model that many novice legal writers adopt, arguing that revision is not a mechanical clean-up phase but a dynamic, recursive practice essential to meaning-making.
In other words, when lawyers write, they build the mental models that allow them to effectively advise clients and anticipate counterarguments. Unfortunately, that mental development doesn’t occur when the drafting is outsourced to legal AI.
What gets lost in the handoff
When lawyers accept AI-drafted work without deep engagement, they miss the intellectual work that would have revealed the keystone argument that elegantly blocks multiple lines of attack or the logical flaw in a line of argument that requires a strategic pivot. These are not minor issues that can be fixed during a proofreading round. They are reasoning failures, and no amount of competent editing catches what you didn’t know to look for when you hand off the critical thinking to AI.
Generative AI doesn’t “think.” It predicts what words are likely to follow other words based on patterns in training data. The output can be highly polished, organized well and syntactically correct. What it cannot be, however, is the product of genuine analysis because no genuine analysis occurred. The AI did not wrestle with the argument. It assembled language that looks like the result of someone who did.
A big problem for junior lawyers
The stakes are highest for lawyers early in their careers because they are still building the mental frameworks that expertise requires. Legal writing is one of the best ways of building those frameworks because it’s not just a communication skill—it is a reasoning skill.
Every time a junior lawyer works through a difficult draft, they are building the pattern recognition and the analytical habits that will eventually make them fast. Expert speed is built on a foundation of slow, deliberate effort. Skip the slow, hard work, and you don’t get the speed. You get the appearance of it.
When firms route AI-generated drafts directly to junior lawyers for light editing or the junior lawyers are encouraged to “take a crack at an initial draft using AI,” they may be clearing the to-do list while inadvertently eliminating the primary mechanism by which those lawyers were supposed to develop their critical thinking skills. While the efficiency gain is obvious, the capability loss won’t be—until no one on the team has the judgment to recognize that an argument is missing the mark.
At this point, some readers might be saying: “Wait a second—while some legal writing does demand a certain level of skill, a lot of it is templated and predictable, isn’t it?”
That may be true, but forms-driven practice comes with its own hazards. People who thoughtlessly rely on forms and people who thoughtlessly rely on AI will both suffer the same fate: There will be little room for them in the marketplace because they’re not bringing enough value.
What firms can do
None of the above is an argument against using AI in legal practice. The tools have genuine value, and the profession will continue to adopt them.
In fact, there are situations where AI can help someone write more effectively. For instance, you can use AI to help stitch together ideas that you have, smooth them out or challenge them. But you have to have the ideas first.
A useful prompt is: “I have a theory that X. I have lots of notes, please help me organize them to be used in an article that argues A, B and C because D, E and F and suggests Z as the solution. Then suggest counterarguments.” From there, a lawyer can go back and forth debating and adding resources and asking AI if their reading of the resources is correct and whether those resources really do support their arguments. Better yet, ask the AI to “fight” you and try to poke holes in your argument.
The challenge for firms, then, is not to avoid usage of AI tools but to find a way to use AI tools without sacrificing the intellectual rigor that defines good legal work.
A few concrete starting points: For starters, firms need to distinguish between the workflows where AI should accelerate production and those where the effort is the point—where the lawyer needs to own the thinking.
For their part, senior attorneys will need to provide meaningful feedback that shows how to walk through an argument and look for opportunities to strengthen or focus it. Additionally, firms should train junior lawyers to include goals and logical steps in their prompts, so that their generative AI use is about confirming and editing the ideas that they had already developed. This also invites the junior lawyers to think critically and look for holes, rather than encouraging them to just proofread for glaring errors.
Efficiency is a legitimate goal. But sustained use of AI for legal writing risks optimizing output while eroding the independent thinking that makes lawyers good at their jobs.
We’ve reached a point where the danger isn’t losing lawyers who can generate a brief. The danger is losing lawyers who can tell whether the brief is any good—and why. That gradual loss of critical thinking and reasoning is the hidden cost of AI-assisted legal writing.
Ivy B. Grey is a legal solutions adviser at iManage, a cloud-enabled document and email management system built for law firms, corporate legal departments and financial and accounting professionals. At iManage, Grey works as a legal subject matter expert and contributes to product strategy and thought leadership.
Mind Your Business is a series of columns written by lawyers, legal professionals and others within the legal industry. The purpose of these columns is to offer practical guidance for attorneys on how to run their practices, provide information about the latest trends in legal technology and how it can help lawyers work more efficiently, and strategies for building a thriving business.
Interested in contributing a column? Send a query to [email protected].