AI for Lawyers: Why Your Judgment Still Matters
Colin Levy has read more legal AI hype than almost anyone. And he thinks most of it is useless.
Colin is General Counsel at Malbek, an adjunct professor at Albany Law School, and one of the most widely followed voices in legal tech. He's also the author of the new book Code Switched: A Lawyer's Guide to the Era of AI. When he says the noise isn't helping anyone, he means it in the most practical sense: lawyers are absorbing a lot of headlines and not enough understanding.
We sat down for a full episode of the Lawyer to Leader Podcast to talk about what AI for lawyers actually looks like right now. Not in five years, not in theory, but this week.
Here's my top takeaways.
We're in Chapter One, Not the Epilogue of AI for Lawyers
One of Colin's most useful framings: we've finished the prologue, and we're into chapter one. Nobody knows how many chapters there are.
That matters because a lot of lawyers are treating legal AI as either a solved problem ("we have a tool, we're fine") or a permanent threat ("it's going to replace us"). Both reactions skip the part where you actually have to learn something.
Legal AI isn't static. The lawyers who will navigate it best aren't the ones who adopted early and moved on. They're the ones building the habit of learning as the technology evolves.
Stop Saying Please to ChatGPT
One of the more memorable moments in our conversation: Colin's mild frustration with lawyers who say please and thank you to AI tools.
It's not a courtesy issue. It's a mindset issue. These tools have no feelings, no preferences, no awareness that you're even being polite. When you soften your prompts the way you'd soften a request to a colleague, you're not getting better outputs. You're just working from the wrong mental model.
The right model, according to Colin: assume it knows nothing. Give it more context than you think it needs. Be direct. Tell it exactly what you want and why. Then check what comes back.
That last part is non-negotiable.
Hallucinations Are a Feature, Not a Bug
AI hallucinations, the confident wrong answers, the fabricated citations, the plausible-sounding nonsense, aren't a glitch that will eventually get patched. They're structural. Large language models work by breaking text into tokens and matching patterns in their training data. They're not understanding your question. They're predicting what a good answer looks like.
That's why the same question, asked to two different AI tools, can produce two different answers. And why the same model can give you a different answer tomorrow than it did today.
Colin makes the point clearly: we've embedded imperfection into these tools, because they were built by imperfect people. The fix isn't a better model. It's a lawyer who exercises good judgment and knows to check.
The lawyers getting burned aren't using broken tools. They're skipping the part of the job where they apply their own skill to what the tool produced.
Your Judgment Is the Point
Colin's core thesis in Code Switched isn't that AI is good or bad for lawyers. It's that AI amplifies whatever you bring to it.
The more powerful the tool, the more skill you need to use it well. A hammer makes it easier to build a wall, but you still have to know where the nail goes. A chainsaw is faster than a handsaw, but you'd better know what you're cutting. The same logic applies to AI in legal practice.
What AI can't replicate is what lawyers actually get paid for: judgment, context, the ability to read people, understanding what the client actually needs versus what they're asking for. These aren't soft skills sitting adjacent to the real work.
The risk Colin keeps coming back to isn't that AI takes over your job. It's that you stop using the judgment that makes your job worth doing in the first place.
Week One of Legal AI Doesn't Have to Be Complicated
If you haven't started using AI tools, or you've started and stalled, Colin's advice is simple: pick one task you do repeatedly that's low-risk and time-consuming. Use AI for that. See what happens.
Don't roll it out firm- or department-wide. Don't set up a governance committee on day one. Just find the thing that's eating your time and isn't directly tied to client judgment, and run an experiment.
The goal of week one is to understand how the tool behaves, not to optimize a workflow. You can't do the second without doing the first.
The Ethics Question Most Lawyers Are Asking Wrong
Most of the legal AI ethics conversation is stuck on a yes/no question: should I use it?
Colin thinks that's the wrong question entirely. The more important question is: how am I using it?
Specifically: Are you protecting client confidentiality? Do you know whether the AI tool you're using is training itself on the data you're feeding it? Have you had an honest conversation with clients about what you're using and why?
These aren't hypotheticals. Lawyers have been sanctioned for submitting AI-generated briefs with fabricated citations. Others have shared confidential client information with tools that weren't treating it as confidential. Neither failure was caused by AI acting recklessly. Both were caused by a lawyer not doing their job.
The ethics of AI in legal practice aren't that different from the ethics of anything else in legal practice: know what you're doing, protect your clients, and check your work.
Listen to the Full Episode: Lawyer to Leader Podcast
We covered a lot more in the full conversation, including how AI agents work and why they require a different kind of oversight, what the billable hour model actually looks like under AI pressure, how to set up AI to challenge your thinking instead of just agreeing with you, and Colin's rapid-fire takes on the one myth he'd kill and the one tool every lawyer should try this month.
If you've been putting off learning this stuff because you weren't sure where to start, this is your starting point.
Watch or listen to the full episode here on YouTube, Spotify or Apple Podcasts, below.
Jonathan Cullen
Want to know where legal leadership skills land? Take the free Leadership Assessment at jonathancullencoaching.com or book a free discovery call with me here:
Full episode here on YouTube below or Spotify or Apple Podcasts, below.
More from the Lawyer to Leader Podcast
Subscribe to the Lawyer to Leader Podcast on YouTube, Spotify or Apple Podcasts for more episodes packed with insights to build your career, like these:
Running a Modern Law Practice in the AI Era (ft. Dylan Gibbs). A former Supreme Court of Canada clerk on why lawyers are so isolated in their careers, what a real peer community changes, and why legal AI is no longer something you can afford to put off.
Building a Speak-Up Culture for Lawyers (ft. Stephen Shedletzky). Formerly Head of Brand with Simon Sinek and author on leadership, on why silence is the most expensive thing happening inside your organization and what actually gets people to speak up.
Courage, Entrepreneurship & Boundaries: Building a Law Firm Young (ft. Sevana Poladian). Sevana walked away from a traditional legal career mid-COVID to build her own firm. We get into what litigation taught her about leadership, how she spots a bad client before signing them, and why she had to actively draw boundaries most young lawyers don't know they're allowed to have.
Comments