Why Communication Will Matter More for Lawyers in the AI Era
Taking part in the Law School AI Challenge this June made me think about how important it is for lawyers to use AI well and check its work carefully. The speed at which it could read large amounts of material, organize legal issues, and draft an opinion was remarkable. Its help allowed us to finish within the time limit, but we still had to decide which facts to rely on and which conclusions to reach.
After the competition, I began wondering about what came before any of that. Someone had already prepared the emails, contracts, and accounts of what had happened. What if, instead of those documents, I had been sitting across from a client who struggled to explain their situation? And once I had finished the opinion, could I have met the other side and actually negotiated a settlement?
As AI becomes better at the work
Finding precedents, analyzing legal doctrine, and drafting legal documents have long required substantial time and expertise. They are still difficult tasks, but the range of work AI can help with is growing quickly. Finding relevant information in a lengthy record, comparing similar cases, and drafting an argument now take far less time than before.
This is more than a personal impression. An experiment involving American law students found that AI assistance substantially reduced the time needed for several legal tasks. Improvements in the quality of the work, however, were less consistent. Faster work and work that no longer needs review are different claims.
Even so, if AI continues to become more capable and accurate, the time people spend reading materials, writing drafts, and revising them will likely shrink further. Knowing a large body of doctrine and precedent, or being able to draft quickly, may become less distinguishing than it is today. What, then, should I learn to do better?
I think much of the answer lies in communicating with people: drawing out the information a client has, gathering scattered documents, and establishing what happened. Then explaining the analysis to the client, finding common ground with the other side, and persuading investigators or a court. These parts of legal work, at its beginning and its end, are difficult to complete with an AI-generated document alone.
The facts do not arrive ready-made
In law school, we usually begin with a set of facts. We read about who entered into a contract, when it happened, and what followed, then identify the legal issues. The parties are conveniently labeled A and B. Important dates and actions are generally somewhere in the exam paper. We practice making sense of the information we have been given.
A real consultation may begin with much less. Suppose someone says, “I lent money to a friend, and they have not paid me back.” There is little to analyze from that sentence alone. When was the money sent, and what had they discussed beforehand? When was repayment expected? Has any of it been repaid? Does the friend also regard the money as a loan, or do they say it was an investment? Are there bank records and messages from the time?
Clients do not necessarily arrive knowing which facts have legal significance. They may talk at length about what upset them most while leaving out a crucial conversation they consider trivial. They may be reluctant to mention something unfavorable, or blend memory with inference when describing an old event. Without the right questions, important information may never emerge.
This takes more than working through a checklist. A lawyer needs to let the person tell their story, then ask specific follow-up questions where something needs clarification. They need to earn enough trust for uncomfortable facts to be disclosed while still addressing contradictions. They must distinguish what the client experienced directly from what the client assumes, and look for records that can verify the account.
Only after those accounts and documents have been organized by chronology and legal relationship is there a factual basis to analyze. In a sense, the lawyer has to assemble it. Unfavorable facts must be included, and what has not been established must remain uncertain. If the account is shaped around a preferred conclusion from the start, even sophisticated legal analysis may proceed from the wrong premises.
AI can help organize the material gathered and suggest questions that have been missed. But it cannot simply obtain a story the client has not told or a document the client has not provided. Thinking about what to give AI eventually leads back to an earlier question: whom should I ask, and what do I need to ask them?
There is still work to do after drafting
The conversation with a client continues after the analysis is complete. To choose a course of action, the client needs to understand the possible outcomes, risks, costs, and time involved. What is legally available may not quite match what the client actually wants.
Even a claim for repayment illustrates this. A client may initially say they want to pursue the matter to the end. A longer conversation may reveal that what matters most is recovering money quickly to cover living expenses. Someone else may care as much about an apology as the amount recovered. Understanding those priorities allows the lawyer and client to judge which offers to consider and which terms are difficult to concede. An opinion on the chances of winning does not settle those questions.
Negotiation also requires understanding the other side. Do they lack the money to pay now? Are they rejecting the proposal because of language admitting responsibility? Are they worried about an ongoing business relationship? Each answer changes the conversation. Even when the dispute appears to concern a single amount, payment timing, payment method, the wording of an apology, or the terms of a future relationship may create room for agreement. A lawyer has to make a proposal, hear the response, consult the client again, and adjust the terms.
Persuading investigators or a court is, of course, different from negotiating terms with the other party. It requires understanding what the investigators or judges find doubtful and answering those concerns with facts, evidence, and legal reasoning. Reading from a prepared statement will not always answer an unexpected question. Understanding the question and providing the necessary support may matter far more than sounding confident.
The same applies to advisory work. Explaining the risks in a contract clause is only part of the task. To produce an agreement the parties can use, a lawyer also needs to understand why the business team wants a particular term and why the other company refuses to change it. Whether in litigation or advisory practice, legal analysis reaches a real decision through conversations between people.
What we practice for exams
These thoughts bring me back to my own studies. We spend much of law school preparing for case-analysis and record-based exams: reading the supplied materials, identifying issues, and applying legal doctrine. Within a limited time, we must catch every relevant point, recall the necessary precedents, and write a coherent answer.
One semester was enough to show me why this training matters. Without understanding the elements of a legal claim, it is difficult to know what to ask. Without knowing the relevant cases, it is difficult to recognize which part of a client’s account deserves attention. A focus on communication still requires a thorough legal education. Knowing the law well is what makes precise questions possible.
Record-based exams are not simply exercises in reading material that contains the answer. They require us to compare conflicting accounts and evidence, assess credibility, and determine legal significance. But however complicated or contradictory the materials are, the record has already been placed in our hands. There is little opportunity to ask a client what else may be available, obtain new documents, and revise an initial view. Adapting an argument to another person’s response after submitting the answer also falls outside the exam.
Moot courts, legal clinics, and practical placements offer some of these opportunities. As students, though, we naturally devote much of our time to the abilities on which we are tested. If AI is going to help more with research and drafting, I would like legal education to give us more practice in learning about a matter from the beginning and putting the resulting advice into effect.
A client instead of an exam paper
The standardized patients used in the practical component of Korea’s medical licensing examination offer one possible model. As the examination authority explains, these participants portray patients according to a fictional scenario and help evaluate the candidate’s clinical performance. Could legal education similarly have students interview a simulated client with a particular set of circumstances?
Instead of providing a complete account of the case, the exercise could begin with no more than “I want to recover the money I sent to a friend.” Students would listen and ask questions to establish what happened. Asking about a conversation could produce the relevant messages; asking about the transfer could bring out bank records. The scenario could include uncertain memories and unfavorable facts the client finds difficult to discuss.
The sequence of questions might resemble a game of Twenty Questions, but counting how many hidden facts a student discovers would be an inadequate measure. We should also look at whether the student understood why a question mattered, avoided leading the client toward a predetermined answer, and followed up when the account conflicted with the documents. After the interview, students could separately identify established facts, unresolved questions, and additional material they need to obtain.
They could then use AI for research and drafting. This would make it possible to see how a missed question affects the later analysis. Even with the same AI tool, different facts and documents gathered during the interview could produce very different results. Reviewing the finished opinion alongside a recording of the interview would give students a much clearer idea of what to improve than a general instruction to “communicate better.”
An assignment that continues after submission
After the interview and analysis, students could attempt mediation or negotiate a settlement with classmates representing the other side. In addition to shared facts, each side could receive confidential information about its client’s circumstances and priorities. The exercise would then differ from a debate in which everyone begins knowing what the other side thinks.
For example, one client might be willing to accept a lower amount in return for prompt payment. The other might be less concerned about the total than about having to pay it all at once. Students would struggle to discover that difference by simply repeating their opening demands. They would need to ask questions, make proposals, and understand why an offer was rejected before finding terms both sides could accept. New documents or a change in a client’s wishes could be introduced along the way.
Assessment should not depend only on whether the students reach an agreement or obtain a higher amount. It should also consider whether they understood their client’s priorities, explained the alternatives to settlement, and stayed within their authority when making concessions. Sometimes the available terms make it better not to settle. Explaining that judgment clearly to the client should also be part of the exercise.
Following one matter through interviewing, gathering documents, AI-assisted analysis, negotiation, and a final explanation could make it much clearer how legal knowledge is used. Not every course needs to take this form. I would simply like more opportunities to apply what we learn to problems between people, discover where we fall short, and return to our studies with those gaps in mind.
The kind of lawyer I want to become
AI will likely become better at supporting conversations and negotiations too. It can help organize an interview, suggest a follow-up question, or play the other side in a practice negotiation. I would not claim that communication is something AI will never be able to do. But the person using its analysis to understand a client’s situation and help make a decision will still need to listen carefully and explain their judgment clearly.
When I left software development for law school, I wanted to become a lawyer who understood technology. At first, I thought a great deal about the “understood technology” part. The more I study law and use legal AI, the more closely I also consider what kind of lawyer I should become.
For now, there is a mountain of doctrine and case law to learn, and I need to study it properly. Alongside that work, I want to practice hearing the full account of someone who needs that knowledge, noticing a problem that has not yet come to light, and explaining a difficult decision in words they can understand. If AI gives me back some of the time spent searching for sources and drafting documents, I want to be the lawyer who uses it to listen more carefully to a client and work through the next decision together.