Teaching the Next Generation of Lawyers to Think With AI

I have had an enjoyable and eclectic career as an entrepreneur, executive, lawyer, technologist, and now a legal educator. The unifying theme has been working on opportunities and challenges created by new technology — whether building a tech-driven startup, as head of technology at the first new federal agency post-Internet, or advising organizations responding to major technological changes on their business strategy.

Law school students today have grown up witnessing more change in their world than any generation before. Recently, they have seen a massive decrease in computer science jobs largely due to AI adoption and are now understandably concerned about the impact AI may have on their legal careers.

I’d like to prepare students to enter the profession as it will exist in the future, and to be a leader in an environment where the pace of change is accelerating due to AI. Instead of fearing AI, I try to get students to see the widespread adoption of AI in the law as an opportunity.

Teaching law has always included teaching substantive subjects such as torts, contracts, and property. Just as importantly, legal education teaches students to analyze problems through objective, methodical reasoning: isolating relevant facts, applying legal rules precisely, and evaluating competing arguments without being controlled by emotion or bias. As we all know, Law schools like to say that they teach students to “think like a lawyer.”

AI will dramatically change legal workflows and processes, but its greatest impact will be on how lawyers think and the cognitive powers they can bring to bear on a problem with the extraordinary power of AI at their fingertips. Law schools will now have to teach students to think like a lawyer using AI.

I want students at BU to take advantage of the law school’s AI offerings to develop both the ability and the confidence to become leaders in the AI-driven transformation of legal practice.

I believe in taking calculated risks and working with emerging technologies before their implications are fully understood. AI seemed like an extraordinary opportunity to bring together my background in law and technology to investigate, learn, and teach about what may become the most significant change in legal education since the creation of the modern casebook and law schools nearly a century and a half ago.

What I got right was jumping in with both feet. The technology was changing too quickly to understand from a distance. Teaching it, experimenting with it, and watching students use it has given me a much clearer view of both AI’s potential and its risks.

What has surprised me most is the almost magical leap in AI capability over such a short period. The progress is not linear, it is exponential. That speed of development has reinforced my belief that legal education cannot wait for AI growth to stabilize before engaging with it. It may never stabilize in the conventional sense.

Last semester, every student in my Artificial Intelligence and the Law course was asked to learn Harvey and apply it to an actual legal use case they had encountered during a summer clerkship or through one of BU Law’s experiential-learning programs.

Each student then presented a live demonstration of what they had created to the class. I was actually astonished by what the students were able to produce. First, they generally learned how to use Harvey far more quickly than I expected, and second, they were very creative in how they applied AI to real-world legal problems.

This generation is far more capable of adapting technology for its purposes than prior generations. Perhaps that is because of a lifetime of access to apps on their phones and laptops, or perhaps students have already been using AI far more than we realize. In any event, law schools are generally going to be challenged to find instructors who are more AI capable than many of their students.

The best place to begin is with assignments that resemble work lawyers actually perform: legal research, drafting, document review, summarizing complex materials, developing arguments, and preparing advice for a client. These exercises allow students to see immediately where AI can improve their work and where human judgment remains indispensable.

Experiential courses and clinics are especially promising because students encounter real, often unstructured legal problems. AI can help them organize information, identify issues, explore alternative approaches, and produce an initial draft — but students must still determine whether the analysis is accurate, appropriate, and responsive to the client’s needs.

One useful exercise is to have students analyze a problem initially without AI and then revisit it using AI. Ask them to compare the two approaches: What did AI identify that they missed? Where did the platform improve their work, and where was its response incomplete, misleading, or wrong?

The important point is that the assignment should not reward students merely for generating an AI response. It should require them to interrogate that response, improve upon it, and remain accountable for the final judgment.

The risk is not that students will use AI. The risk is that they will accept its output without understanding, testing, or taking responsibility for it. We describe that as the difference between cognitive offloading and cognitive surrender.

Cognitive offloading occurs when a student deliberately uses AI to extend their own capabilities — perhaps to organize information, test an argument, identify an overlooked issue, or improve a draft. Cognitive surrender occurs when the student substitutes the system’s answer for their own thinking and accepts it uncritically.

Assignments should therefore require students to make their reasoning visible. They should explain what they asked AI to do, how they evaluated its response, what they verified independently, what they rejected or changed, and why they stand behind the final product

I have used AI not only to help create course content, but to build a structured process for developing an entire new course.

I created an AI script based on the decisions I made while designing the first week of my new seminar on AI-assisted legal reasoning. The script and I interact to guide the development of each subsequent class and produce a defined set of deliverables for each class including a class overview with learning objectives, readings and assignments, a detailed lecture outline, and a complete slide deck with speaker notes.

For me, that is a good example of thinking with AI. The system does not decide what the course should teach or substitute for my judgment as a professor. It helps me translate my ideas and the lessons learned from building the first class into a disciplined, repeatable development process. That allows me to work more efficiently while also making the course more coherent, rigorous, and consistent from week to week.

There is an important caveat however: that while AI helps me produce a course in far less time, I am not fully prepared to teach it until I have spent additional time reviewing, thinking, and weighing what has been created using AI.

The course begins by looking at human cognition: intuition, deliberative reasoning, cognitive bias, and moves on to how technologies over history have allowed us to expand our mental capabilities. It then introduces Shaw and Nave’s Tri-System Theory, which treats artificial cognition as a third system operating alongside intuitive and deliberative human thought.

Students will apply that framework to actual legal research, writing, and analytical tasks. They will compare work performed independently with work performed using AI and examine how AI changes not only the final product, but also the questions they asked, the issues they noticed, and the conclusions they reached.

The course therefore requires a kind of metacognition — thinking about how you are thinking. Students must identify when they are using AI deliberately to extend their capabilities, which we call cognitive offloading, and when they are beginning to accept its output without sufficient scrutiny, which we call cognitive surrender.

The goal is to develop lawyers who understand how AI affects their reasoning and who can use its extraordinary capabilities without surrendering the independent judgment at the center of professional responsibility.

I have learned that faculty resistance is not a single problem. Some professors have principled concerns about accuracy, ethics, intellectual development, or the effect of AI on assessment. Others are interested but do not have the considerable time to master a rapidly changing technology. Still others are uncomfortable becoming beginners in front of students who may already have more practical experience with AI than they do.

Faculty benefit from concrete examples created by colleagues, practical support, and permission to experiment without pretending to be experts. In this environment, credibility comes from intellectual honesty and careful judgment, not from claiming mastery of a complicated technology that is changing so quickly.

Law schools must recognize they have a real capacity problem. Professors are being asked to learn the technology, reconsider their teaching methods, and redesign courses while continuing all of their existing responsibilities. Realistically, adoption requires time, training, access, and institutional support.

Prospective students now ask law schools how they are engaging with AI and increasingly consider the answer when deciding what school to attend. Employers want to know whether graduates can use AI when they enter legal practice and whether they can use them responsibly. Influential alumni also expect their law schools to engage seriously with changes reshaping the profession. Like it or not, law schools need answers for all three groups — and those answers cannot merely create the appearance of innovation.

At BU Law, the response has grown from a dedicated Artificial Intelligence and the Law course into a broader institutional effort. The new certificate program adds or modifies almost ten courses. Many are skills-oriented, while others incorporate AI into doctrinal subjects. We are also working to integrate AI into experiential programs and access-to-justice work where BU Law has demonstrated leadership for many years.

Many brilliant people teach and work at law schools, but these are not organizations that change easily. Law schools have operated in broadly recognizable ways for well over a century. That stability serves a real purpose, particularly in transmitting respect for the law, disciplined reasoning, and professional responsibility to each new generation of lawyers.

The consequence is that changing a law school can feel like turning an ocean liner. It takes patience and time. The wheel may already be turning, but the ship does not immediately move in a new direction.

I have learned that early progress comes from creating tangible examples: a course students value, access to a professional platform, a faculty member willing to redesign an assignment, or a certificate program that gives separate efforts a coherent structure. Those examples make the change concrete and give others something they can evaluate, improve, and join.

Institutional transformation will nevertheless take years. In the meantime, schools need leadership that can create momentum while respecting the enduring values that legal education is right to protect.

Harvey gives students an opportunity to work with a professional legal AI platform before they enter practice. That matters because it moves AI education beyond consumer chatbots and abstract classroom discussion into the environment in which lawyers are beginning to perform real work.

Its greatest value is that students can experience the full cycle of AI-assisted lawyering: defining a legal task, supplying context, evaluating the response, refining their approach, and deciding what must be verified or changed. They begin to understand that professional competence with AI is not a prompt-writing trick. It is a disciplined interaction between technological capability and legal judgment.

Providing access through the law school also promotes equity. Students should not develop professionally important AI skills only if they can afford a particular tool, happen to work for an employer that provides one, or know how to find their own way into the technology. A common platform gives every student the opportunity to learn and allows faculty to establish shared expectations for responsible use.