Thinking Like a Lawyer in the Age of AI
AI’s growing role in legal education is sharpening the focus on which forms of judgment matter most and how to teach them deliberately.
Across legal education, a familiar worry keeps surfacing: Is AI eroding the critical thinking that law school exists to build?
That question does not have a simple yes-or-no answer. But a recent conversation among Bobby Chesney, Dean of Texas Law; Tim Duncan, Director of AI at BU Law; Jonah Perlin, Professor of Legal Practice at Georgetown Law, Jeff Kelly, AI and Innovation Partner at Nelson Mullins, and Harvey Legal Innovation Partner, Megan McMillin, pointed toward a more useful set of questions.
Is critical thinking actually declining, or is AI making existing gaps more visible? As AI changes the work through which students and junior lawyers have traditionally learned, who is responsible for developing judgment? And what should an AI responsible use policy solve for in the first place?
AI does not make “thinking like a lawyer,” less important. It does, however, force schools, firms, and individual lawyers to be more deliberate about how good legal judgment is developed, practiced, and assessed.
Redefine the Target: Judgment, not Knowledge
When information and first drafts are increasingly easy to generate, a lawyer’s value shifts toward what they do with that raw material. Jeff Kelly described AI as “flattening what we'd consider knowledge,” with the differentiating work coming after that: discerning what matters, choosing, refining, and defending a position, and recognizing that “sometimes less is more.” For lawyers, that can mean resisting volume for its own sake: narrowing an argument, removing immaterial detail, and deciding which points will actually move the analysis.
Bobby Chesney calls the practice of testing an argument’s foundations, weaknesses, and implications “intellectual engineering.” Faced with a position or argument, a lawyer should instinctively ask, “What are the strengths and the underpinnings of that position? What are the weaknesses? What are the implications?”
AI makes one's capacity to engage in this intellectual engineering more valuable, not less. A model can produce a fluent answer in seconds, but fluency is not the same as sound reasoning. Lawyers still need to test the premises, identify what is missing, and decide whether the conclusion holds.
Move Beyond the Simple “Decline” Story
Classroom experience complicates the idea that AI is simply weakening student performance. Jonah Perlin stressed that measuring critical thinking is difficult and that law schools still rely on imperfect proxies. From a writing perspective, he said: “I have not found that my students are any worse or better. If anything, what I've seen is maybe the bottom of the class has sort of moved up a little bit, and I think that may be the product of using AI tools or learning with AI tools.”
That does not make the risk of students outsourcing too much of their thinking any less real.. Tim Duncan identified “cognitive surrender, de-skilling” as “a very real concern with AI.” Chesney placed the concern in a longer historical frame in a June memo to his faculty, describing cognitive de-skilling as a worry “as old as Plato's Phaedrus.” New technologies have long raised the possibility that something will be lost when a tool takes over work people once did themselves.
The useful stance is neither panic nor complacency. AI may expose judgment gaps that were already present, even as it creates new ones when users surrender too much of the thinking. That visibility creates an opportunity. Instead of hoping students absorb judgment “by osmosis,” schools and firms can identify the capabilities that must remain distinctly human and build training around them.
Question worth asking: Where is human judgment genuinely necessary, and how do we teach students to recognize those moments before they rely on AI?
Redesign the Path from School to Practice
For years, the division of labor was relatively clear: law schools taught doctrine and legal reasoning, while firms provided practical “reps” through junior work. AI is putting parts of that arrangement under pressure, but the change is more nuanced than the wholesale disappearance of associate work. Jeff Kelly was explicit that “I don't totally agree with that premise.”
The pressure is nevertheless visible. Kelly recalled a private equity client telling the firm “they're not paying for associate work for due diligence.” That kind of work has historically been one of the ways junior lawyers learned by doing. If some of those assignments shrink or change, the profession will need other ways to provide the underlying experience.
Clinics, externships, and experiential education offer one starting point. Chesney argued for giving students more “pseudo reps” before graduation so an employer sees a new graduate as “six months ahead of where they usually are,” particularly as “the large firm pyramid gets to be more of an obelisk.” Formal partnerships between firms and law schools could extend that approach, allowing students to see how lawyers use AI on live matters rather than learning about it only in the abstract.
The responsibility cannot sit with institutions alone. As the automatic apprenticeship model thins, students and associates will also need to seek out feedback, stretch assignments, and opportunities to practice rather than assuming the traditional career path will supply them.
Question worth asking: If some routine work where associates once learned is disappearing, which reps should schools provide, which should firms commit to, and what responsibility does the individual learner have to seek them out?
Teach Durable Skills Through Real Tools
One response can be summarized as “teach meta-skills, not tools.” But that clean distinction breaks down in practice. Perlin offers a more useful formulation: “We need to use tools to teach meta skills,” because no one knows which products will dominate in three or six years. The goal cannot be proficiency in a single interface. It must be adaptability, process awareness, discernment, responsible use, and the confidence to challenge an output.
Those habits still need a place to be practiced. Constructing a careful prompt, interrogating a response, and verifying an answer within one platform can teach an approach that transfers to the next. A platform like Harvey can therefore serve as the conduit rather than the point. The tool provides the environment in which students practice a durable habit of mind.
Doing that well will require investment in the people teaching it. Duncan observed that “the people that are trying to teach AI in the law schools are people that are also learning themselves.” Faculty development and support from practitioners will be essential. So will a broader view of legal work as a system, not simply a series of discrete tasks. As Kelly put it, AI “rewards system thinking in a way that traditionally the tools we had available haven't.”
Question worth asking: Are we teaching a product, or a transferable habit of mind that will survive the next product? And who, within the faculty or beyond it, is equipped to teach that habit?
“Both/and,” not “AI or not”
Schools do not have to choose between AI literacy and traditional skill-building. Chesney’s memo makes the relationship clear: the two goals “are not in a zero-sum relationship. It is possible to pursue them both with excellence.” It treats skills, assessment integrity, and educational rigor as distinct challenges, each requiring a more considered response than a single policy can provide.
At Texas Law, that means pairing required AI-literacy training with deliberate protection for unmediated thinking. Perlin’s approach to assessment follows similar logic: “anything that's done outside of class, we have to presume that someone is going to use AI,” making intentional, in-class, AI-free settings more important as places to build and demonstrate core skills.
Chesney’s memo reaches the same conclusion. As shortcuts outside the classroom multiply, “the minutes of class meetings become an ever-more valuable asset,” creating an opening for “a revival of the Socratic method, not in a jerky Kingsfieldian way, but in a thoughtful way,” along with “laptops down in 1L courses.”
Before deciding where to require, permit, or remove AI, a school needs to define the outcome it wants. Once the desired capabilities are clear, individual policy choices become easier to evaluate. Speed alone is an incomplete target. Perlin’s closing reframe is useful here: generative AI “doesn't make those tasks necessarily more efficient. They make it more effective,” allowing a lawyer to work “bigger, better, more thoroughly than I ever could.”
Question worth asking: Before debating an AI policy, what is our metric of success? Are we optimizing for speed, or for deeper and more thorough legal reasoning?
Further reading: Tim Duncan has since written about why legal education should look beyond its own walls to decades of cognitive-science research, including Kahneman’s work, on how people think and learn. Read his piece.
What Should Change From Here?
The practical agenda begins with treating judgment as something that can be intentionally developed, not an instinct students are expected to acquire along the way. It means creating new forms of practice as traditional junior work evolves, building faculty capability, and teaching durable habits through the tools lawyers will actually use.
It also means recognizing what new lawyers can contribute. Kelly urged the profession to treat the “beginner's mindset” students bring as a competitive advantage during a period of transformation, and to recognize process design as a skill worth developing and rewarding.
Legal education can make AI literacy as ubiquitous as legal research tools became a generation ago while still protecting intentional, unmediated thinking time, especially in the 1L year. Those goals reinforce one another when the objective is clear: not simply faster legal work, but better judgment about what the work requires.
Continue the Conversation
Go deeper with the panel
Hear the full discussion with Bobby Chesney, Tim Duncan, Jonah Perlin, and Jeff Kelly on how AI is changing critical thought development for law students and associates.
Build practical AI skills with Harvey Academy
If you are a law student with access to Harvey, visit Harvey Academy to explore student-focused courses, including Preparing for Legal Work, and build practical AI fluency for the work ahead.
