The Future of Claims — Episode 5: Sean Griffin

Host: Andy Anderson | Guest: Sean Griffin

You're listening to "The Future of Claims," a show about the changes happening in the world of insurance claims. I'm your host, Andy Anderson. I've spent over a decade at the intersection of insurance and technology as a founder, a CEO, and a podcast host. We're going to sit down with some of the leading minds in claims to hear how they think technology, people, and organizations will transform claims over the next 10 years. Sean Griffin, thanks for coming on "The Future of Claims." Really appreciate you joining.

For those who aren't familiar with you and your background, I'd love if you'd give the quick one-minute version of who you are and the type of stuff that you do. Okay. One-minute version. I was born and raised in Milwaukee, Wisconsin. Go, Pack, go.

My high school graduating class had 14 people in it. From there, I went to Harvard University, which had quite a few more people in it, graduated with honors from there, went directly to Columbia Law School. Note that I am carefully leaving out the years in which I graduated, just in case you see this by video and might be fooled and think that I'm younger than I am. Hope springs eternal. After I graduated from Columbia, I clerked at federal court in Baltimore.

Then I went off to the Department of Justice for a few years to do commercial litigation and appeals, and from there I went into private practice, and that's when I got involved in insurance coverage and insurance defense work. Did a lot of trials, coverage and defense trials, and I have been loving insurance ever since. Insurance is incredibly boring until you actually get in it, and then it's like, for those of us who are insurance nerds, it's perennially fascinating. Well, at litigation, as they say about us, we find every case we're working on fascinating, and so I'll get involved in a case, I'll find it fascinating, then I'll tell somebody at a cocktail party about it, and then the expression on their face will let me know that is not fascinating at all. Either way, generally two have to refresh their drink, but their glass is still full after.

Yeah. Yes. My wife has definitely warned me several times not to talk about insurance at parties. So... Yeah.

I've learned it the hard way. Well, I love it, so we have a full hour, and I'm not going to get a drink anytime soon. So yeah, obviously you've been writing and posting and talking in various places around how AI in particular is impacting this world of insurance defense, and overall just the legal profession overall. So as we talk about cases, why don't we jump right into a case that's been in the headlines recently, which is this Hepner decision. If you could give folks who aren't familiar with it a little bit of background, and then how your reading of it and how it may impact the use of technology over the next few years.

Sure. So there was a CEO, his name was Hepner, and he knew he was under federal investigation, and he wisely hired an attorney to help him manage his way through this investigation and possible prosecution, various allegations of fraud, and other bad things he had allegedly done. However, he, not satisfied with having an attorney that he paid good money to, decided to run a few of his concerns through ChatGPT and see what ChatGPT had to say about his legal predicament. The feds showed up as expected. They grabbed his computer.

Also, he probably knew that was going to happen because his attorneys, I'm sure, told him. And his attorneys went through, I mean, the federal government went through it, found the inputs that he had inputted to ChatGPT, and they said, "Oh, we're going to take a look at these." And he tried to claim privilege over his questions to ChatGPT. I don't remember what the questions were. How guilty am I, perhaps? How to hide evidence, question mark, question mark, question mark.

I don't know what he said exactly, but once he re-involved his attorneys, his attorneys raised two claims. One, a claim of attorney-client privilege, and another claim of work product protection. The case went in front of Jed Rakoff, who was a former professor of mine at Columbia. Actually, he helped teach a seminar with Professor Coffee. And so, a great guy, great professor, and he wrote his opinion in this case, and he concluded that neither the attorney-client privilege nor the work product protection applied.

He said the attorney-client privilege is meant to protect confidential legal communications between an attorney and his client. And he said, "I can't help but notice that there is no attorney present here. There is just a robot. A robot cannot be an attorney, therefore you do not have attorney-client communications." Now, granted, he said, if you're using another person or something else to facilitate a communication with an attorney, that could qualify for the privilege, but there's no indication that you were doing that here. You just got started inputting your stuff into ChatGPT, and that's not an attorney-client communication.

That's you talking basically the same as you're talking to a friend. That part of the communication is hard for me to dispute. I really can't really criticize that part, and I've seen very few people who have, at least ones who understand that he did address the idea that it's easy to imagine a situation in which you could use ChatGPT to facilitate your conversations with an attorney, but that involves hypothesizing a lot of facts that we just don't see in this case. Set those to the side. The second one is work product protection, and this is one of the things I've got involved in when I was a very junior associate at another law firm.

Had to do a lot of research into work product. I read a lot of treatises about it, a lot of cases about it, and I've kept that interest up quite a while. I know Todd Presnell, but I hold my own. And the work product protection is Federal Rule of Civil Procedure 26 B3, and one of my soapboxes. I have many, but one of my soapboxes that everybody keeps calling it attorney work product, and it's not attorney work product.

By the very language of the rule itself, and even the language of Hipkin v. Taylor, which created work product protection, is not attorney work product. Anything you do in a reasonable anticipation of litigation ... is ordinarily protected from discovery by the other party, and that could be either by an attorney or by the party themselves. The rule says that very explicitly.

But my former Professor Rekoff has skipped over this distinction. He said, well, this attorney work product, the ChatGPT is not an attorney, therefore you can't claim work product protection. Now, as a matter of law, that's just wrong, but his other claim was that, well, you have to keep this confidential. You can't input it into any old thing because ChatGPT tells you right up front in your terms of service, nothing you input is confidential. Now, you could pay for versions that are confidential, but he did not pay for that version.

He paid for the free version. So he said, basically, you are telling a third party your business, and therefore, once you waive that, you have waived your work product protection. Now, others have pointed out that you only waive work product protection if you waive it to your opposing party, not necessarily if you waive it to somebody else. That's a very fine distinction that not all courts follow. I get where they're coming from, and they're not wrong.

But it's hard to say if I put, for example, my trial memo up on a billboard, it's hard to say that I had not waived work product protection. Similarly, if you input your narrow, otherwise work product-protected material into ChatGPT, which then uses it and spits it out in another context, hard to say that you're maintaining any sense of confidentiality there. So I can understand where he's coming from in that perspective. So in that respect, it's hard to say that he's wrong, especially with the ChatGPT disclaimer right up front that what you put in here is not confidential. Nobody pays attention to that, but I think if you are under federal investigation, you can't be too careful about such things.

And perhaps Heptinstall should've been a bit more careful. But then again, I guess if you're a careful person, you'd probably would not be under federal investigation to begin with, so. Yeah. Allegedly some of the things he had done. Allegedly, he's definitely under federal investigation.

It probably was not because he was such a careful guy, but then again, as Reggie pointed out, we should let the legal procedure take its course, and so- Yeah, that's a great summation and analysis of those cases. And I think, and honestly, for someone who's not an attorney but have followed that case a little bit, I think there seems to be a broad reading of this decision as potentially imperiling the use of AI generally in the legal world and concerns that it may create issues of privilege when particularly attorneys are using AI generally in the work they need to represent their clients. I guess I'd love for you to speak to that issue, particularly because at least the description of this case seems to draw some pretty clear lines that most of the time would be not crossed should AI be used by attorneys in working cases for their clients. You would think. This is the case, the facts of which lead to a certain result.

It's hard to argue another result. Yeah. But having said that, there was another case out of Eastern District of Michigan where a pro se person actually got work product protection because they said, well, the person's tried to litigate their case, and you should need to let them litigate their case. And then you could say, well, why did the pro se person out of Eastern Michigan get it and not this guy? Which is a question I can't fairly answer.

I would say, listen, ChatGPT is not really supposed to be confidential, and I say that quite glibly, but if you could imagine a situation 30 years ago where somebody said, "Okay, Sean, we want to see all of your Lexis searches. We want to see what you're researching about this case to see if you can get out of it." And I think most attorneys would intuitively feel that that is a different situation. Hey, wait a minute, that's my work product, he'd say. That's how Sean is thinking about the case. What I decide to research, the inquiries I put in, that is certainly work product.

I don't think I'd find anybody who'd disagree with me even 30 years ago, so. Yeah. Yeah. Go ahead. Yeah.

So it's just that, one, is this part of that relationship where there is an expectation of confidentiality and established law- Mm ... around that, and there's questions of intent, and obviously the circumstances very much matter. And it does seem like there's quite a bit of settled law around the overall use of technology. And I was seeing, again, following various posts around, there were earlier points around Lexis, around the use of email, around the use of cloud systems. We've traveled somewhat similar ground in different points in time when new technology has arrived, and it seems like the courts have settled these questions.

But I don't want to put words in your mouth, and I'm sure you don't want it in a judge's mouth either, but- Well, it's interesting that you raise that because remember the... Oh, come on, who was that woman in Florida who was accused of killing her children, convicted of it? Casey Anthony, that was her name. That's right. And part of their evidence they used against her, they would've found all of her Google searches.

They said, okay, her Google searches is how do I dispose of a body? How do I get away with murder? That thing. And nobody jumped up and said, "Hey, wait a minute, that was in anticipation of litigation." Now, if you do it before you commit your crime, a different story, but- Yeah ... I'm sure that there might've been an argument to be made once she knew the police were closing in on her, if she was still doing Google searches about how to get away with a crime, conceivably she could say, "Hey, wait a minute, I should get some protection over that." But as far as I know, that did not come up.

I'm not a big true crime person. Maybe they did at law. But basically, inputting ChatGPT is more akin to inputting a Google search. I think most people would agree with me, although they'd probably struggle to explain exactly why. And how about the, again, there's two issues.

It's the individual non-attorney doing things on themselves. I seem to understand the point around it. It isn't necessarily have to be an attorney's work as long as it's in expectation of litigation. Yeah. But I think let's go more clearly.

There seems to be some gray area there, which I think I understand, but quite specifically for the attorneys, the use of technology, if they're using tools that are confidential, if they're using it with the expectation that it's privileged, the fact that it is AI itself does not seem damning in terms of this is going to violate privilege. I think that's exactly, I think the bright line you need to draw, that attorneys need to draw, they're using AI is AI that is closed. That means they don't use your inputs to train data. They keep everything. They have the rights over everything.

It's all still confidential to you or your firm. And the AI that's open that they absolutely going to use your queries to train data. And if they have to ask which one you have, you have the open one. So your invoices would let you that you have a closed one, no problem, and then you need to proceed accordingly. Meaning even though I pay for all these services, I do not input anything into AI that I think I would want to assert privilege over.

I just don't. I don't say, "Hey, how can my client get away with this?" And so that'd be the same thing to put out public anyway, but I certainly don't upload it to the cloud where anybody can conceivably get ahold of it. It is possible through interrogating ChatGPT to get it to spit out something, especially if the thing is there. If that Sean Griffin regularly uses AI, ChatGPT or whatever, to prepare his trial outline, well then, if you're clever enough, you could say, "Hey, what's up? I was Sean Griffin, prepare for this case.

I have a gift to get some next month." Well, ChatGPT will happily tell you. So yeah, let's go down this road because you've also posted a lot about attorneys using ChatGPT or some open system and getting a lot of fake citations and whatnot. That seems to be largely an issue that seems to happen with the models, the GPT. GPT actually stands for something pretty nerdy, but I think it should stand for general purpose tool, because that's what it is. It's not an attorney.

There are much more legal-specific AI. At least it seems like you've been posting that you use some of them. I use a lot of them. A lot of it I use just to stay on top of AI's capabilities. Yep.

So a lot of times I personally will ask AI the same questions I tell people not to ask, just to see if it's gotten any better or what have you. Yep. Just play with it. But a lot of them have been implicated in this because again, AI is just a tool, and if you have a screw you need to drill into the board, pull out your hammer and don't pull out a can of spray paint. Use the appropriate tool for that, a screwdriver.

And same for AI. It's just a tool, so there's no use blaming that tool for your messing up. And all of the tools have been implicated because they're not really designed to find the right case for your proposition. They have learned what a case looks like, and they'll make up a case and make up something that sounds like it could be in a case, but that's very helpful in that respect. But in the terms of reliably finding a case, they simply can't do that.

And I've tried to correct it. Yeah. I've said, "No, no, no. That's not a real case." And then it said, "Oh, you're absolutely right." It'll either cite the same case again or cite you a different made-up case, but you can't get it to cite a real case. Even using Lexus or some of the legal specific- Lexus or Westlaw, they're better at this because they have access to the cases.

The problem I found when I was using them is it's an enthusiastic but not entirely competent first-year associate. And so if you're asking a really basic question like, "What is the statute of limitations for wrongful death in DC?" It'll tell you two years, no problem. There you go. But if you ask a slightly touch with a little bit more nuance, it's going to throw you off. For example, a couple of years ago, I asked the AI whether you were obligated to give the additional insured a copy of the CGL policy.

Is there a statute along those lines? And they said, "Yes, there was a statute of DC and they called it the life insurance statute." And I said, "It's not life insurance statute, it's CGL. You're required to give them that policy." And knew the answer, but it just could not get its head around the fact that there might be a difference between a CGL policy and a life insurance policy. It just would not get that. And again, any question that an experienced attorney will have to research, ChatGPT is not going to be terribly helpful for you.

Even Perplexity, it's better because it's designed to answer questions, it'll show you what it shows you. But if you ask a legal question, it'll look at the rules, it'll start looking at the law firm websites to get the answer, which probably be, but it's not really what you want to do in terms of the hardcore research that most insurance attorneys- Defense teams are drowning. Tighter deadlines, thinner margins, and the plaintiff firms armed with AI and litigation defense. That's why we built Oracle. I'm a former insurance executive.

My co-founder is a former litigation defense partner. We built Oracle specifically for the work defense teams actually do inside a fully closed, secure system. The result, faster, more comprehensive work, better outcomes for carriers, increased profitability for firms. Today, Oracle is being used by top national insurance defense firms and regional boutique specialists handling workers' comp, trucking, construction defect, med mal, toxic torts, and everything in between. Oracle, AI for the defense.

Learn more at oracleim.com. That's O-R-A-C-L-A-I-M.com. I mean, being the builder of an AI tool for defense attorneys, legal research is not an area that we do, partly because we don't have the repositories that Westlaw or Lexis, but also We haven't seen that for those very nuanced questions, I think, and I think AI is getting better, but it's a challenge. Yeah. You have to teach humans for three years to get them to learn how to read a case properly and analyze it, see what's really behind it.

And even then, you only know but so much. So computers, it's going to take them a while before they really analyze certain things. And they might be better at some things than others, but right now, you'd have to be crazy to rely on them in their current state. And ask me again next year, I might have a totally different answer, but right now I'd say don't do it. Is there anywhere you're bullish in terms of the use of AI in legal?

Oh, lots of ways. For example, it's really good at summarizing big, massive documents, especially if you just need to get the gist. A client might ask you, "Okay, we've just given you 1,000 documents. What do they say?" You could run that through your AI program and say, "Okay, listen, here's what the documents look like." Again, you're not going to necessarily run into court on that, but a good AI program could summarize those documents very well. I use AI all the time for proofreading documents.

It's really good at proofreading. Not as good as you would think. If I have a brief, I will often run it through more than one AI program. I'll fire up ChatGPT, and Claude, and Gemini, and there'll be overlap, but they will not all find all mistakes. Each one of them will miss a few things that I wish they'd caught.

AI is really good for forcing me to do something I don't want to get started on. Let's say I want to write an article that I promised to write. Oh, geez, I don't want to do that, but you could put in an outline or just give it an idea. It'll write something. Will it be as good as you like?

No, it won't be, but at least it'll get you started. It's more easy to revise something than it is to write something from whole cloth, so you can get started on that. One of the FDCC members, Jimmy Wilkins, gave a demonstration in which he used it to write letters to opposing counsel. I would call them Rule 11 letters. They call them something else in Mississippi, but basically, "Hey, you bum, you didn't give us these documents.

You better shape up." And since there are so many examples of those out there, AI's really good at writing those. You give it a few facts, you create those out, no problem. So those are three use cases that I and other attorneys I respect use AI for on a regular basis, and you could easily think of more. If you have a closed system, you could use AI to summarize your emails if you're overwhelmed by emails. People do that.

Again, I emphasize this needs to be a closed system. You need to ensure that it's closed, this is the policy, so on and so forth. But if you have that, you have assurances of that, then a lot of your emails that you've kicked off to the side, it could say, "Okay, Andy wants to tell you that you should've worn a different shirt for this video call," whatever, and okay. "Sorry, Andy." You look great, Shawn, if that was your concern. Red's your color.

That's another thing, the answers, you can. Yeah, it's sycophantic. It is sycophantic. It's always, even when you catch a mistake, "Oh, great catch, Shawn." Yeah, I know it was. That's why I did it.

But there are a lot of ways they could use it, and a lot of times they will talk it all like I'm very bearish on AI because it does make a lot of mistakes, but people think it's a genie that, or one of those sci-fi robots that always has the exact correct answer. And I said it is like a sci-fi robot, but it's like C-3PO in "Empire Strikes Back" when he tells Han Solo that the odds of surviving an asteroid field are 3,720 to one. He's just wrong. Han Solo's doing no harm. His chances are much better than that.

It's more likely that he'll get through with no problem. So yeah, ignore that guy. You have to know what C-3PO's good at. He's a protocol droid. Yeah, and I like where you're going here, which is, I think don't make perfect the enemy of good, and certainly in the world of insurance defense that we see a lot of the issues that you're bringing up, which is that folks just have too much to do and not enough time.

Mm-hmm. Right? That I think they need to be informed. They need to understand how to do it, the limitations of its capabilities, but it can be a huge help, which I think, and don't throw the baby out with the bathwater. Very much so.

It's a useful tool, just like my screwdriver's a useful tool, but you can't use it to do calculus. That's not what a screwdriver- Totally ... calculus is for. And one of the things you made a post a couple of recently that I saw about the use of AI and how it changes nature of work to some degree, and I think there's perhaps this belief that AI's going to just make everything faster, so everything's going to take less time. And I think at least the post that I saw, the suggestion was it may do some of that around the edges, but it actually may be particularly good for actually improving the quality or the depth of work around particular topics.

Can you talk a little bit about that, what you posted and your reactions and thoughts on that? Yeah. So AI, as I said, it can summarize documents very well. Certain things it does very well, and most attorneys I know who use it, people ask, "Well, what's the use case? Why would an attorney use that?

Attorneys get paid by the hour. Why would they want to get work done faster?" Most attorneys I know are already overwhelmed with work. They're probably isn't getting more work, it's how to do the work that they have, and AI can be a huge help with that. You mentioned not letting the perfect be the enemy of the good. If the client is demanding those documents be summarized quickly, you could say, "Listen, here's a quick AI summary of what I did for the documents.

These are the three key documents. I've looked them over myself. These are correct. Let's talk about those." You can compress a lot of work into the time you have available, so that's a very good way of doing that, too. On the flip side, what I tell a lot of attorneys is, "Listen, if you're an attorney, you probably think you're an above average writer." That's because attorneys have huge egos.

Not because we write well, but because attorneys by definition have huge egos. And so just keep in mind that the generative AI platforms, they all work by the same way. They basically go through the entire Internet, they take in everything, and they use it as part of their training data. So if you remember when you played with Play-Doh as a kid, you grabbed all the Play-Doh together, you mush it all together, you had one big ball of Play-Doh, and it'd be hard to separate out the blue from the red from the yellow. It was all mushed together, one big average ball.

And ChatGPT, Gemini, Claude, they'll give you a big average ball. But if you think you're above average, well, then it's going to drag you down to average, isn't it? Your writing, it's going to take all the bits and pieces that make you above average, take those out, and make it average. Now, if you're below average, it's a huge boon because it'll bring you up to average. That just drags you down, too, and I've noticed when I have it proofread stuff, it's always flattening out some of the nuances that I need, especially in an insurance case where the nuance, the devil's in the details, and you need to put those back in.

You need to make sure that it understands, well, not it understands. You have to understand what you want to do, and you need to stick by it no matter what the robot is telling you to do. Yeah. I think I find myself And maybe you, I'm not sure you do, I find it does best when I give it the rough structure of what I want to say. I can give it an outline, even like bullet points, that these are the ideas I want to cover, and then it does a good job of...

I find the quality of the prose that it writes pretty good. The sentences are well-constructed. They're grammatically correct. And it takes those ideas and it helps them flow well together. I would love to say, like everyone else, I'm above average, but I also know that I'm perhaps overly verbose and repetitive in some areas, so I've also noticed it's quite good at polishing the edges around some of that writing.

"Okay, we've made this point. We don't need to make it four times in a row." Yeah, it's very good at that, and I want to emphasize the point you said. The more guidance you give generative AI, the better the output is, but that's same as any other associate or any other attorney you talk to. If you just let a junior associate run off on their own, you're going to get whatever you're going to get, and really you have nobody to blame but yourself because how's the junior associate supposed to know any better? If you give AI a very rough outline, "I want you to write a Rule 11 letter to this guy, tell him he's a bozo," they'll just make up some reason he's a bozo, stick it in there, and off you go.

But if you say, "Okay, here are the letters we've exchanged. You get the gist of what we're arguing about. Tell him he's a bozo and he needs to respond to me by Friday at 5:00 PM," okay, you're going to get a much better letter, especially if you guide it along the points it needs to hit and everything else. Now, if you're thinking about this, you're saying, "Well, Sean, all you're doing is basically tricking yourself into writing an outline, aren't you?" And then ChatGPT fills it in, and then you write the rest. So you really still are doing 75% of the work, and which is, I think that's true, but that still saves me 25% of the work that I can use to turn to some other client's needs.

So that 25% is very valuable to me. That's a- Yeah ... an investment. And I do think, that's a point that I've heard quite a bit, which is maybe a counterintuitive one, that AI is actually, in its current state, most useful for the people who already know the space because you can catch where it may be going off track and you're like, "Okay, I know this is in the hands of an amateur." You believe it because it is quite convincing. It speaks with false bravado.

So I'd love to think though, I don't know if you've thought about how AI may impact the journey of young attorneys and young associates. It's going to be a difficult issue because when I was coming up, back in the Jurassic era, I'd get off my horse and buggy, pick up my quill pen and... But the reason how we trained was a large part, they'd have us do document review. They'd send us to a big warehouse and go through the documents or read the documents and try to figure out how the documents got into the case, and then you summarize that for the partner. That was how you learned how a case goes.

Now, of course, it'd be ridiculous to do such a thing. No client would pay for that. They just run it through some AI platform. You get the idea, and then you put it into some program, and you do a bunch of targeted searches, and that's the way to do it. And it's a much better way to do it in terms of efficiency, in terms of how many documents you can get, especially since with the advent of email, also 30 some odd years ago, there are a lot more documents than there used to be.

So that method was no longer really sustainable, but that does take away one way in which associates learn. I also wrote memos for my partners and for clients. They'd say, "Sean, research the preemption doctrine" or whatever. I think that might have been one of the first memos I wrote as a summer associate. And I go back, and a five-page memo about what preemption is, they're like, "Oh, this is really good." But now, you could probably get Perplexity, Claude to tell you what the preemption doctrine is.

Now, you can trust any of the cases it puts in there. You have to find those out on your own, but if you just want to know what the preemption doctrine is, hey, there it is. Now, is it going to apply to that case well? Maybe, maybe not. That's another thing the attorney has to do, but your question was about how a junior associate can learn.

Well, if you learn what works and what doesn't work by actually going through and doing that work yourself, that's how you have to know. Much like a pilot, they have an automatic pilot on the plane, but they don't just set the plane, turn on the button, and then hop off the plane as they go fly off to Missouri or whatever. The pilot's still there with his hands on the wheel, making sure that everything's going okay because it's too important to be left to AI. And same way for attorneys. Yeah, you can have the AI there helping you even by itself.

If you're a junior associate, yeah, you could have AI write your memo, but you're going to stay a junior associate forever if you don't get out there and get your hands dirty yourself. But I do fear that yet another avenue of training for junior associates is going away with the advent of AI because a lot of the work that they did just is not practicable for them to do anymore. Yeah. Yeah, and I think there will be, I know you're heavily involved in some of these associations. We've become pretty involved ourselves, and I know I've been talking with some of the leadership about how do we think about training, how do we think about in this world, how do we...

You rebuild the training process, the training manuals, create courses where you show them how do you use AI responsibly, how does it help you do this work, how can it be... Again, it's a different path, but the sky has been falling for Chicken Little and many others for a long, long time. I don't think it's the end of lawyers, the end of associate training. It's just going to be different like it always, like everything is with every new technology. We've been nattering on for a little while.

I think this has been great. Any other areas that you think you'd want the audience to think about or know? What are the pieces in this space that if you had a proverbial soapbox, you wish- You wish the world of insurance defense, both on the claims side, on the in-house side, as well as outside counsel side, knew and was thinking about? I would say a couple of things. For the insurance side, there are a lot of programs out there that claim to make the insurance process, whatever part of it you decide, underwriting, evaluation, whatever, simpler.

I think those have their place. They're useful tools, but you have to have a human thoroughly involved at all times. That's because if something goes sideways, if somebody accuses you of bad faith, you can't just say, "Well, the gen AI keep up with this." You can't argue with it. That's not going to get you very far at all. But if you do have a human looking over these things, again, the pilot who knows what's going on, a lot of these tools can be very helpful.

I'm a big fan of all these AI tools. The second thing I would say with both in-house attorneys and outside attorneys is if your firm or your company's going to be using AI, make sure you have an AI policy in place. A lot of the cases I see where AI has gone sideways is clearly the result of having either no AI policy or cursory or ineffective AI policy, which is the same as having none at all. And the people we laugh the hardest at are people who clearly do not have an AI policy, clearly do not keep a human in the loop at all times, clearly do not keep their eye on what was going on. They just let people run amok with whatever they're going to do, and people do what they do.

And you can't say they violated the firm policy because you don't have one, and then the whole firm ends up getting sanctioned based on one attorney who submitted an AI-written brief of fake citations or did something else goofy with AI. So those are the two things I would tell everybody listening to this. Yeah. I know you've done some cyber work, and I think that unfortunately, having done a lot of that and worked in cyber insurance for a while, the humans usually are the problem. Yes.

And I think that that adage that if you make the most secure thing also the easiest thing, that's a good general rule. And if you just ignore it and pretend it's the ostrich is not a good way to go through life. Mm-hmm. And what I worry about the most from a security perspective is honestly for this community that has, as we've talked about, too much to do, not enough time, that you essentially have an easy button that is one browser tab away that creates all of those issues. It's open.

It's not a closed system. It is not trained on anything. It is not built. It is a general purpose tool. I've literally had this conversation with a very good associate, and he's like, "Yeah, our firm says we can't use AI, so I just do everything on my iPad, not on the firm computer." Well, that's exactly it because I do a lot of cybersecurity work, and anybody with security in their field knows that just like water seeks its own level, people will seek the easiest option, and people will seek it out, and they will happily sacrifice security for convenience.

And so you hide your head in the sand, as you said, if you think your associates, your coworkers are not using AI, they're just not telling you about it. And so you want to have a policy to govern it, even if your policy is don't use AI. Give it a shot. But people will seek out the easiest way to do their job, and that's just your policy needs to take that into account and be realistic about it. So I think that's a great point you raised.

Well, I know I would love to get you more involved in some of the efforts that we're doing with some of these organizations around training, around education, because I think it's much needed, and then obviously you've thought about these things deeply from multiple directions. This has been really fun. The guidance around Hepner and your close reading of that based on your experience, I think is invaluable in particular. Anything else you want folks to know before we wrap things up? Nope.

I think they can catch up with me on LinkedIn if they really have burning questions or with you as well. But yeah. Cool. We'll put links to your LinkedIn, to your firm so people can find you. And if people were looking to hire you for legal work, what is the specific stuff that is best?

What's the slow pitch up the middle? It's baseball season already, so for you. Baseball season already. Well, mostly I do commercial litigation, which means if you are involved in litigation involving a business and there's a contract involved, or you're suing another business, I'm your guy. If there's fraud involved, I'm your guy.

Also, as you probably have gathered, I do cybersecurity, privacy, AI work, both consulting and litigation. So if you have questions about that, please give me a call. Happy to help out. Cool. Awesome.

Sean, thank you so much. Really appreciate you being on "The Future of Claims" podcast, and we'll see you maybe here again or at one of the various events that we're going to be at. So thanks so much. Great. You're welcome.