New Tools for An Age-Old Lawyer Problem
How can we teach young lawyers to handle live witnesses when the stakes are too high to let them actually handle live witnesses?
One of the most frequent concerns I hear from lawyers about AI isn't about hallucinated citations or client confidentiality. It's about people. If AI is doing the research memos, the first-draft briefs, and the document review we've always handed to junior attorneys, how do we train and develop the next generation? How can we equip new lawyers to take on the more sophisticated work that still requires a human touch?
These are fair concerns and warrant attention. But as I've mentioned in previous posts, I could envision that the very technology prompting these questions can also help answer them: AI as a training tool, not just a task-taker. And it's already being used in exactly this way. Recent reporting in the ABA Journal describes law firms adopting AI-powered deposition simulators to train their attorneys. The setup lets an attorney take a deposition against AI agents playing the witness, opposing counsel, and even the court reporter, then get real-time feedback on how they did.
This could potentially crack one of the oldest catch-22s in professional development for litigators: handling live witnesses. You can't call yourself a trial attorney, or even a litigator, unless you've done it. At my old law firm, experience with live witnesses was a threshold requirement for making partner. And the only way to get good at handling witnesses is to handle witnesses. Yet, the opportunities for junior lawyers to actually do so are scarce. Most clients understandably don't want to be guinea pigs and allow someone with no deposition or trial experience to handle a real deposition or trial examination, where the stakes are real and often high. So junior attorneys wait years for these opportunities, and the experience gap compounds. A simulator lets lawyers practice, build judgment, timing, and instincts in a setting that feels real, but isn’t on the client’s dime, and the only thing at risk is their own learning curve.
This is effectively the next evolution of the mock trial exercise, which has been used to train litigators for about as long as litigating has existed. But it's better in a lot of ways. Traditional mock trial training programs require substantial resources and attorney time. They are usually run as retreat-style gatherings that pull large groups of personnel away from billable work all at once and add expenses such as travel, meals, and hired actors. A simulator can be run on demand from anywhere, deliver real-time feedback, and use a fraction of the resources. Better still, it can be applied to a real case any time you need it, rather than once a year or every few years as part of a training institute or formal program. And it works in both directions: not only does this equip lawyers with the skills to handle live witnesses in the real world, but it also gives them a way to demonstrate that ability and inspire confidence in supervising attorneys and clients that they're ready to do it for real, which benefits everyone and only further enhances the quality of service delivered to the client. And it's not just associates who can benefit from this kind of exercise. From what's being reported, firms are running attorneys at every level through them, a reminder that skill-building doesn't stop at a particular class year.
Depositions are just the entry point. Once you can convincingly simulate a witness and a courtroom, the same approach extends across the skills litigators rarely get to rehearse: putting on a witness at trial, conducting cross-examination, delivering openings and closings, and even jury selection. You could run voir dire against a simulated panel, or test a case theory through an AI-generated mock jury and see how a verdict might land, all before you're standing in a real courtroom with a real client's outcome on the line.
The feedback layer is where this gets genuinely interesting. Pair generative AI with biometric tools, and a lawyer can see things a human observer might miss: where their pacing faltered during a tense exchange, whether their delivery lost the room, and how they actually came across under pressure. AI can flag the objections you missed, and, just as usefully, the ones you made too often. It can simulate not only the witness but the judge and the jury, so you're preparing against realistic behavior rather than a script. That kind of specific, repeatable feedback is hard to get from a busy supervising partner, and with AI, it's available on demand.
None of this replaces mentorship or real courtroom time. A simulator won't teach you to read a particular judge's mood or to think on your feet when a witness goes sideways in ways no model predicted (although you can train it to be a loose cannon, so you can practice expecting the unexpected). But it can get attorneys far more prepared at every stage before those moments arrive. That's the throughline in how I think about AI in our profession: used well, it's a tool that makes lawyers better, not obsolete. The training worry is real, but it's also an opportunity. The firms moving first aren't replacing their attorneys' development. They're accelerating it.
I'm curious how others are thinking about this. If you're at a firm experimenting with simulation or other AI-assisted training, what's working? And if the development question is the thing holding you back from AI, I'd genuinely like to hear how you're weighing it.
