Legal Practitioner’s Guide to LLMs

I recently attended the Bloomberg Law Symposium, where I had the opportunity to hear from two very different worlds: academic researchers presenting their latest findings, and legal practitioners and law firm representatives speaking candidly about the challenges their field is facing — what’s broken, and what they need.

What struck me most was the gap between what Computer Science researchers understand about AI and how legal practitioners actually experience it. There’s a common claim that generative AI (including large language models) has lowered the barrier to entry and expanded access to justice. But from what I’ve observed, that’s not quite the reality. If anything, it may have shifted the barrier rather than removed it: the assumption baked into that optimistic view is that users already know how to operate these tools correctly. That’s a significant assumption. Computer Science and law are fundamentally different fields with very different skill sets, and what feels intuitive to a researcher may feel entirely foreign to a practicing attorney. There’s also the question of trust, which varies widely from person to person. Speaking for myself, I’ve always approached AI with a degree of skepticism; not because it isn’t useful, but because different models are built for different purposes, and frankly, nothing has reached the level of true “general intelligence” yet. As a result, I’ve been noticing a pattern of misuse in the legal field that, with enough awareness, could largely be avoided.

We are living through a chaotic moment of technological disruption. A rapidly evolving technology is upending practices that have been stable for decades. In education, for instance, the same problem sets and exam questions we once relied on no longer serve their purpose, because AI can solve them effortlessly. But disruption doesn’t mean obsolescence; it means adaptation. We need to find a way to co-exist and co-evolve with AI. A useful analogy is the calculator: once we had reliable, ubiquitous access to calculators, we stopped requiring students to do arithmetic by hand. The skill shifted, from computation to knowing how to use the tool correctly and understanding what it can and cannot do. I think AI is no different. That said, the legal field presents some particularly distinctive challenges that make this transition more complex than it might be in mathematics or computer science, with confidentiality being a prime example, and those are worth examining carefully.

Coming away from those conversations, and after discussing it with my advisor, I decided to launch this blog series. My goal is to synthesize recent research in a way that’s accessible and useful to legal practitioners, making it essentially a practical guide to navigating generative AI in law, including how not to use it. Hopefully this becomes a valuable resource to legal practitioners.

======