Episode 013 BigLaw, Privilege and an Unexpected "Wow!" Moment

Welcome to AI Tools for Practicing Lawyers. Practical, no-nonsense guidance to help attorneys put AI power to work in their practice right now. Most lawyers they're still talking about AI like it's a productivity tool. But if you're a white-collar defense lawyer handling investigations, compliance, cross-border discovery, and sensitive communications, AI becomes something very different: a privilege issue, a governance issue, and potentially an evidence generation machine.

My name is Ron Trescher. I'm Heather Gardner. And I'm very happy to introduce our guest for today, who is Matt Lafferman. He's a partner at Denton's, does white-collar and government investigations, cross-border investigations, crypto and blockchain matters.

He's on the AI task force at Denton's. He advises on AI tools, workflows, and policies. Matt, it is awesome to have you on our show today. Great.

Yeah, it's great uh great to be on and thanks for inviting me. So, all right, before we get into privilege and discovery, I'm curious, because you're our first big law guest, I'm curious to he to understand, at least in Denton's, how AI is being adopted. Is there is there a struggle to get AI adopted, or is it pretty much, yeah, let's jump in with both feet, or is it something different? Um, I would say that across the industry, what I'm seeing, and I'm sure that you know ever all attorneys are somewhat familiar with this, lawyers are a little conservative in adopting newer technologies, new ways of doing things generally, it's just the nature of law practice, a little bit backwards looking.

I think that is also applying to AI. I think you've seen a lot of, I think you've talked on this, talked on this podcast about a lot of very famous, or I guess ethics cases that have kind of arose that have gotten a lot of media attention. And I think I've definitely seen in the last couple of months as you know, AI is being rolled out of the firm across the industry, older, maybe more reluctant attorneys who are more set in their ways, have their own established business, um, established practice, are a little bit more reticent to engaging in AI.

That said, I think Denton's is pretty ahead of the curve in terms of adopting AI. We've been an early adopter. I'm very proactive in training, giving guidance to attorneys. I've been part of that process.

Um, attorneys certainly on my team, associates on my team, are instructed to kind of use AI, um, build out files, incorporating AI into the tool, incorporating the file into the AI tool, depending on what tool we're using, and you know, running queries throughout the case to essentially manufacture, like uh put together first at least first drafts and uh first drafts of analysis as well. All right, so you mentioned tools. What's what's in your tech stack? I would say I mean I have a quite a bit of different tools.

I would say my favorite from a publicly popular accessible tool is Claude. Right now I think Claude is, I think it used to be about maybe six, nine months ago. I think Chat GDP was pretty popular. Claude is uh, I think surpassed it, especially for work-related tasks.

I would say, you know, obviously, uh, you know, no I wouldn't recommend putting confidential material, client material into Claude, as I've as you've talked about numerous times on this podcast. Um that said, it it is very good at putting together marketing materials, you know, um drafting up first drafts of um, you know, client alerts, things like that. It's quite good. And then I think the firm's main tool, they have a couple.

Clear brief is quite good, drafting. I really like Harvey. I think Harvey has in the last nine months has just improved marketably in terms of what it can do. And I think you've seen that across all tools.

I mean, I think there was a comment earlier this month by Ken Griffin of Citadel. He uh was you know sitting at Stanford Business School and asked about AI. He's re he's a pretty uh a pretty uh negative commentator historically. I think back in January he referred to AI as garbage.

Um about like uh earlier this month, he he changed his tune drastically, came home one Friday, he said he was depressed because the work that usually has has become so much more sophisticated, these AI tools, the work that's usually been done with people with masters and PhDs in finance that took weeks or months is now being done by AI agents over the course of like hours and days. And he said he's basically said that like this is beyond just kind of mid-tier, his words, mid-tier white-collar jobs.

You know, more advanced roles are being automated with the Gentic AI. And I think, you know, certainly that is really helping lawyers who require a little bit more sophisticated analysis than more historically, it's been used for administrative and ministerial tasks. It's now being stepped up and being used for preliminary analysis, preliminary drafts, and it's quite good at that. All right, so let's unpack a little bit about what you just said, which is very interesting to me.

Harvey, we know, um, is uh an enterprise-level tool, uh, and it's built so that you can talk about confidential client information safely in a siloed environment. You talked about Claude, not necessarily wanting to put confidential client information in there. What tier are you using of Claude? Uh we have we have a um enterprise version at the firm, um, and then we have a uh you know a standard public version as well.

I mean, I use it in my spare time to play with and build out software uh applications. Um I would also say there's a couple of other tools I really like. I think Copilot's great with email. You know, one of the things I would also say about Harvey is that you have to be very careful about like putting confidential materials in, especially in my practice.

Um, and one of the benefits of Harvey is that it is designed for that. And there's obviously search functions that you know are off-limits if you're doing more um, you know, public record searches, things like that. I think it it connects to the internet, and you have to be very careful about running confidential inputs in regards to those. But I think generally speaking, that's the one I use.

And then I have access to a couple of you know client databases and a software, and I use some of their some of their software, and they have quite good tools, I would say, as a general point throughout the technology industry, which are some of my clients. Um, you've obviously seen some reductions in force across the industry. I think you've seen a couple, um Amazon, a couple in the industry that are being announced. Um so I think in all of those cases you're you're seeing it become uh you know the standard practice, and I think that's posed a little bit of a challenge definitely in those companies for in-house council, which have a different is a different challenge than maybe, you know, maybe a smaller company where it hasn't been such a quick adopter to AI practice.

You know, Matt, I I want to go back because to me, the ability to put confidential client information into an AI tool really elevates your ability to put out higher quality work and different kinds of work. And what I heard you say was that you're comfortable putting that kind of information into Harvey. Is there any kind of information that you wouldn't put into Harvey? I think you just have to, I think that um, you know, from my perspective right now, I wouldn't necessarily think so.

I haven't really thought through that issue yet in in depth, but I would say preliminarily, I can't think of any off the top of my head. I think it's a quite a good tool. I mean, I think well one of the things you've talked about, I think on this, you've talked about your three-legged stool. I think you've talked about USV Hepner.

Um, both of both of those instances are great examples of why you know you have to be very careful on using what selecting what ADA AI tool you use. So I think Heppner, again, Heppner was a little bit of different facts than you kind of your standard uh work product case. Um, you know, the defendant in that case had produced like some documents memorializing communications with Claude, wasn't done at the direction of counsel. You know, essentially the government sought that work product.

You know, there was a there was a claim that that was uh attorney client privilege, uh attorney work product. The court granted the motion for the government, handed over the materials. One of the things that the Judge Rakoff was in that case is out of the Southern District of New York, he discussed one of the things he covered was um the AI tool specifically. And he talked about how the tool that was used, which was uh Clawed, uh claimed confidentiality.

So specifically, he pointed to the privacy policy, which was in effect at the time of when the defendant ran these searches, and he pointed out that the it collects data on the prompts, entered, and the outputs generated. It uses the data to train the AI and it may disclose the data to a host of third parties. Instances are like, you know, things you need to pay attention to. And I think one of the things that Harvey's done a great job at is kind of is adjusting for those, you know, those realities and essentially being uh confidential.

Obviously, you have to look at there's a variety of different factors I think you have to look at when you consider the safety of a certain tool from a from a confidentiality and privilege perspective. But again, there's a corollary there in terms of, yeah, it's a third-party agent and it's automated, but it's still operating as a third-party agent. And in a lot of ways, it's it's not as any different from you know a paralegal or you know, investigator or things like that. Right.

And in Hepner, the clawed tier that he was working in was not the enterprise tier. Yes. Right? It was it was the pro tier.

I mean, that's the thing that is becoming more and more obvious to people who are in the AI space, and probably not obvious at all to people who are not spending a lot of time here, that ChatGPT is not ChatGPT. ChatGPT is there's the free version, there's the plus version, there's the Teams version, there's the enterprise version, and they all have different levels of security and privacy controls. Absolutely. And that's a hard thing, I think, for users who are less sophisticated to really understand.

How do you communicate that idea to your clients and to other people in your sphere? I mean, I think that generally speaking, as in most law firms, everything has to be cleared with the client ahead of time. And part of that representation is that everything will be, you know, handled with confidentiality. I would also say that one of the things I recommend my clients to do when they're adopting AI tools throughout their investigative functions, litigation functions, just throughout their legal department, certainly in-house, is to put together a policy.

And there's, you know, on the AI use protocol. Depends on obviously on the sophistication of the company. There's a challenge with AI for in-house counsel, which is you have this mixed role, right? You're kind of doing a little bit of legal work, but you're also handling the business.

When there's a business decision that reflects business and legal evaluations, the business aspects aren't privileged simply because legal considerations are present. You need to kind of consider that. So operating the right tool, um, you have to choose the right tool, have a have a policy that increases the defensibility of privilege. One of those aspects is, you know, ensuring that the tool you select is privileged, but also maintaining other taking other precautions.

And one of which is, you know, mandatory human review, right? Making sure not just that it's recommended that everyone review for accuracy, it is mandatory that all attorneys incorporate their mental processes, opinions, thoughts, and that increases the case for opinion work product. Because we we believe that all of the work that we do in relation to AI and in terms of output is opinion work product, right? And we don't want the you know, the government, you know, opposing counsel and litigation to make this uh make the argument that it's fact work product, right?

It's uh doesn't incorporate, it's just more of a compilation of facts that is much more discoverable, especially in you know grand jury matters where all they have to do, the government can't has a you know a burden of proving that it it's uh they have a substantial need for it and it's too burdensome to go to another party for it versus opinion work product has much stronger protections and discovery, and those are kind of the goals in terms of having clients like essentially establishing that you know client's work product is opinion work product.

Heather, you're in the trenches all the time dealing with lawyers and law firms. Do you have any sense that there's sensitivity as to the different tiers of these AI products? Not really. I do still hear a lot of hesitation, but uh Matt has brought up some things that we've talked about a lot on the show, and that is how big law has this gap, right, from smaller firms, right, because they just have access to such better, you know, more robust tools and software.

And I think that AI is really changing that quickly. I I know that they have, you know, better AI tools, but the the AI tools that are available to everyone are becoming better and better. In the few years they've existed, they've advanced tremendously. I mean, they've gone from summarizing a legal brief to being able to do complex legal, you know, drafting.

And the way that his firm is using it as operational infrastructure is something that smaller, you know, boutique and solo firms could be doing if they weren't, if they were willing to, and probably many are. So I feel like it will start to help close that gap of accessibility. So yeah, I I I'm hoping so. Uh, you know, one of the things that Matt brought up is my beloved three-legged stool, which kind of sounds a little bit silly, but it's really a very important concept.

You know, Matt, do you I don't know if you had a chance to look at a couple of legal documents that I drafted with the massive assistance from Claude, which are the uh notice of intent to use AI in discovery and the proposed language of uh a protective order for a Rule 26F. Uh now, do you have any comments or thoughts about those documents or ways that lawyers could implement those or or similar documents and their litigation matters? Yeah, that's a great question. I think that um some of those materials are would be great, you know, in terms of a productive order.

I think it should go a little bit different in terms of it depends on how you want to disclose that to opposing counsel. You know, obviously there's a series of case considerations about whether or not you want to go down that route, whether or not you even think that there is any sort of need to do so. If you're, you know, obviously I think if you're putting your own client material into the tool, I don't I think that, you know, that having a rule 26 uh disclosure uh or having a protective order, I I'm not so sure of whether or not there's validity to that.

It just opens you up to like potential discovery. Now, if you're using other clients' material or other, I'm sorry, opposing party material, that depends, right? It what are the it's gonna depend on how the confidentiality is is uh defined. Um maybe it includes, you know, as you know, a lot of times when you define confidentiality, maybe it's just as simple as putting in a sentence saying confidential material can be accessed by you know attorneys, their agents, paralegals, court reporters, and you know, put into AI tools.

So it kind of depends on the matter. I think uh like what you've provided is extremely helpful. Um, certainly, like, you know, obviously it's not gonna apply to every single case. No one can make a template that applies to every single case.

Um but I do think it's very helpful. I would say, in addition to that, as like the lawyer leg, I think that um every firm needs to have some sort of protocol, you know, policy, as I said earlier. And this helps you, again, like if there is a court inquiry in terms of what it is, privilege. I think that the challenge you're gonna have with this is that there's not a lot of case law in there.

There is that out there. There is some case law that's saying outputs are a work product. I think there's a a case from last year that I thought was very helpful: Tremblee versus OpenAI. Just for the anyone who's interested in looking it up, it's you uh 2024 U.

S. District Lexus 141362. And it's not that's not a hallucinated citation. Not a hallucinated citation.

Just an unpublished Lexus case. But it is still very helpful in the court considering, you know, what's fact product whether outputs are fact work product, um, you know, uh opinion work product, and you know, how those types of you know material is going to be waived, right? In that particular case, there was an allegation that like um that open AI had infringed on the plaintiff's copyright because open AI's chat GDP was training using copyrighted books. Sure.

Um, you know, part of the plaintiff's work before filing the case, the council ran prompts and received outputs from ChatGDP as part of the testing process. Um and then OpenAI came back during discovery and requested the production of all prompts and outputs from that testing, including any negative results. So the court found that the council, the plaintiff's counsel's prompts and outputs were protected as opinion work product, right? Because they reflect counsel's mental impressions, legal theories.

Great, you know, that's our position as well. But that said, the court did say that anything that was disclosed in the complaint was waived, right? So therefore there was no claim for work product. So again, like, you know, courts are applying standardized doctrines to this, um, which I think, you know, again, you know, is why we think that opinion of work product is privileged, and that's our position.

So um, you know, again, like the idea that creating a policy is something that can improve defensibility. You can always submit a declaration saying you have a copy of a policy, the policy maintains everything is work product, here is why, right? One of the things that I recommend is there's a practice for sharing outputs within a policy, right? So outputs containing privilege analysis um should be marked.

You know, you instruct the tool to mark everything as an output. Privilege and confidential, attorney work product, because again, like these, while the a moniker, something that's stamped attorney work product, isn't again dispositive, it is something that courts look to. Traditionally, courts and judges will look to to determine whether or not something's work product. So it's just it's about make building the case slowly to make sure that at every step um, you know, there is a case for privilege.

And I think that that is, you know, courts are gonna be more receptive to you know, the argument that AI functions as an extension of counsel if an organization can point to a defined policy saying that at each stage of our workflow, we take precautions that this to ensure that this material is treated as privilege, it incorporates mental processes and therefore it is in fact work product. Can you give us an example of a couple of steps that a law firm would or should take to ensure that they retain that protection?

Yeah, I mean, I think that one of the things I would do is um when building AI agents, anything that's workflows, everyone has different names for them, but it's essentially a kind of a custom-built, reproducible uh prompt for those of who may be unfamiliar with the workflow agent terminology. But uh when you build that out, say uh instruct the the prompt saying, please treat this information as you know, attorney work product and privileged and mark it as such when the output.

So that's one thing at the output stage or the input stage. Then at the output stage, as I said, mandatory human review. Another thing I would say is don't give unsolicited legal advice. So, because the argument is that, okay, I'm instructing the AI tool to provide a legal analysis.

And, you know, under a COVID and progeny and other cases, directing a, you know, doing something at the direction of an attorney, there's a strong argument that that's opinion work product. Now, if someone just, you know, without direction, provides, you know, some sort of legal determination or evaluation, that is arguably may not be an opinion work product. So instructing the tool to not do that is very important because I've had instances in the past where certain tools think they're helping and providing more analysis, and they provide legal advice and or legal analysis or legal determination.

And you last thing you want to do is have you know some document sitting around in your stack of documents saying, oh, your client's liable, and here's all the reasons why. And then it's not it's not protected from discovery. So that's a very interesting uh discussion about work product privilege. Are there any other uh privilege doctrines that are implicated by using AI?

Yeah, I mean, I think that it kind of depends what how you use it for. I mean, if you're using maybe co-pilots, write emails or other things like that, I think maybe you could fit in like attorney client privilege. I think that if you're sending material, obviously to clients are repairing it. For clients, I would make the argument that some of it is attorney-client privilege.

It's a little bit more difficult if you're doing a client client attorney-client communication if there's not actually a communication at issue. So um, you know, I'm not saying that that there the argument wouldn't be there, and you know, we would again establish that all this material is attorney-client and work product, but when you're you know facing a court, I think court's maybe a little bit more skeptical of that argument. So I'm gonna take the opportunity to grill you on a controversial matter that I posed at a CLE event last week.

I had a hot take. And my hot take was record everything. Lawyers, you know, unless somebody's telling you, oh yeah, I murdered that guy, you know, record conferences within the firm, recur record telephone calls. Presuming you have consent, of course, to record everything, which is required in many jurisdictions, because I believe that there's value in AI software that will transcribe the calls, will summarize the calls, will order the in-person conversation, will have uh to-do lists that it will generate from it, will assign tasks to the various participants in the conversation, and will be a searchable artifact that you can put in a file that you can look up later if you, you know, two, three, four years later, when nobody's gonna remember that phone call or that conference, and nobody was taking.

detailed notes of it and this could be extremely helpful. What's what's your uh AI expert, criminal defense attorney response to my hot take? So I would say that for public conferences, absolutely anything that's kind of in um notes with that wouldn't arguably be between a client or would be a sensitive matter, I'd say yeah, feel free to use it. Just I would be wary a little bit of two-party consent states, right?

Recording is it kind of depends. I'm this is not my expertise, but you know, generally speaking I wouldn't record an estate uh without asking the other party for consent. So let's just say you have a call with um you know an AUSA um you know about a proffer that you're making there might be some benefits to um asking if they can use AI to summarize the call just make sure you do that and then you have a clear notes and then you don't necessarily need to you know worry about taking notes and making sure you have everything down while also making your the best case for your client.

There's another role here where I would also say also like depositions, right? Like and if you want to turn it on during a deposition especially if you're listening in it's a great tool. However, I would say for other tasks I would be a little bit more reticent to using it. I would not use it in conversations with clients.

I would not use it in investigation interviews. I think it's a little too risky in terms of is it a transcript? Is it not a transcript? Is it what uh to the extent what mental processes are being used by an attorney and even if you're reviewing it summaries, you know last thing you want to do is have something as sensitive as an interview notes be considered like a compilation of facts and just have fact work product.

Again, like that's going to depend on the particular client's risk profile. Some some uh clients may be open to that risk profile in terms of efficiency the number of investigations they have they they can't um you know uh they have to use you know more automated tools to kind of streamline the investigations but you know I would I'd be very reluctant to use it for more sensitive matters. All right so Matt you're an incredibly interesting guest we could go on I have I have notes that go on for page after page after page with questions we're just not going to be able to get to them it means you may have to come back or maybe like so many metaphor metaphysical questions they will just have to go unanswered.

But this brings us to this week's practice signal this is a segment that we have every week where lawyers in the wild tell us what's really happening inside law firms, their frustrations, bottlenecks, operational headaches that don't always show up in CLE or AI demos. So this week's practice signal comes from a lawyer in a small firm asking a deceptively simple question which is how do I automate my intake process? And when you look at the workflow it's not just intake it's inquiry management, scheduling payment processing, matter creation, document handling, accounting coordination, calendar management all spread out across Outlook, PC Law, Adobe, intake forms, local file storage.

So first of all, Matt, you do intake intake is standard in every business around the world whatever your business is there's some form of intake and if it's a service business there's massive intake does Denton have like a firm-wide protocol or do they leave it to the attorneys and the departments? How does that work inside your firm? I mean I think it's pretty standard uh at most uh large law firms you know you get a matter run it through conflicts send down an engagement letter open a matter I think a lot of that's automated already on our side uh whether or not you can have some ideas for doing that there's you know maybe some co-pilot features you could set up that if you get an email from a client, you know, once you've had that initial discussion with them, it automatically generates a file.

There's a little there's a little bit more sophisticated ways you could do that. I mean maybe if you connect uh um you know certain AI tools through the API and to your Outlook platform or whatever platform you can maybe it'll generate automated file or you could just build an agent custom drop everything in and it generates you know an engagement letter relatively simply you know you obviously have most of the time you have a standard template anyway it fills in the details you send it out and then starts building the file um I would say generally speaking within my practice anytime I get materials from a client everything goes straight into my vault in Harvey I'm building out anything there is asking preliminary questions, reviewing the materials myself.

Harvey's really great I think most AI tools do this now you run a prompt and it gives you an answer but then it gives you the source. So with a there's a little number next to it you click on the number it pulls up the deposition or the document and sometimes you know it's inaccurately analyzes a document or extrapolates really you know too much of a an assumption that's just not supported by the document. But I think that it's great at pointing in the right direction you get you know hundreds of documents right off the bat you need to get through it quickly you need to get up to speed um you know that's it's a great tool to do that especially at the outset of a case a lot of my crypto cases involve lots and lots of WhatsApp conversations right there's not necessarily like a written written contract sometimes and you know using AI to parse those communications is is something that is really helpful especially when you have like you know a hundred page WhatsApp conversation you need to parse and get through the entirety of it.

I would also say just as generally if you get a complaint throw it into Harvey ask it to ask it to analyze generate a chronology based on events in there right off the bat gives you a great understanding and no matter what it will streamline your review of the case because uh you go into the complaint after you that after you get a chance to read through it yourself, you'll have uh an idea rather than having to read through it and develop the idea of what the case is like as you read through it, which obviously you know you have to be a little bit more thoughtful and contemplative and it's more of a slow read you immediately can go through it and point out the key parts and knowing exactly what the the evidence is that the other side's pointing to or the allocation the other side's pointing to to support their claim.

So Heather you're dealing with law firms and their intake bottlenecks all the time are you seeing any movements towards more automated intake workflows definitely yes uh people are utilizing the tools that are now being baked into CRMs case management retention softwares but uh Clio has uh multiple AI tools now available to help automate intake my case and practice Panther all have those available and I've seen even small firms building out their own AI intake processes with their own enterprise tools which I think we're going to be seeing more and more of um and I have two questions that you could touch on Matt because listening to you is fascinating.

Do you think it's fair to say though that for courts and and people but especially for courts and judges the advancement is so fast that it is very hard to keep up with any sort of regulation or case precedent at this point because even the ones you've mentioned in three more months may be completely useless because AI will have it may have advanced that far. And also do you think that the legal profession overall is underestimating how much AI is changing confidentiality itself and privilege.

You know I think that you know in the first point yes I definitely think every I think most people are underestimating how fast it's moving. I certainly underestimated how fast it's moving mean back in like early last year I actually I think I started learning about it right uh praying about it took some like just in my own spare time started well you know watching YouTube videos reading about how to build a prompt and initially I was kind of unimpressed it wasn't able to really do very much in terms of February 2025.

Fast forward six months it's still limited to ministerial tasks. Now I'm writing briefs with it. I throw in you know a Dalbert motion I had recently I throw in the copy of the report I throw in the case law that I've you know have have uh written I give it basic instructions of what I want it to focus on and it writes me out what I think is a great first draft. I'm not saying the whole thing's usable you know 60% though is quite well written and I need to kind of retool it a little bit obviously but um and obviously I put my mental impressions and thought processes into it that it's all opinion work product until the final draft that gets filed.

Yeah no I think it's it's certainly moving really fast. You know the fact that technology companies are going to maybe even consider changing their standard work product from Word or PDFs to readme is one one indication of that. And then secondarily yeah I think absolutely think that to answer your second question I absolutely think that lawyers are underestimating the privilege and confidentiality concerns. And I've seen this at every level in-house counsel outside counsel there's a lot of you know there's a lot of depends on your risk profile again but there is some risk here and there in terms of how you want to maintain that defensibility argument.

And until there's a really strong body of case law that maybe you know goes which way I think you have to take a little bit more of a uh precautionary approach uh in terms of again very low burden building in certain aspects in the tools mandating human review are just a couple and there's obviously a huge amount of things you could build into a policy a broad range based on what you're doing exactly but I definitely think that um there is an underestimation of that and you're gonna see maybe some some mistakes that are made.

I mean have you seen REC mistakes that are made within ethics people are just generating these hallucination cases and no it's shocking to me nobody actually just looks at the case in general and I wouldn't you know I wouldn't have you know you always always look at the case I mean the AI tool is great at analyzing you throw the case in there it analyzes the case but you obviously want to make sure that the case stands for what it just helps you kind of direct you to the right part and make sure that you can analyze a little faster.

I was just going to say most uh technology historically the big technology we have lived through the internet the cell phone we've seen gradual advancement AI is is just going at such an exponential pace yeah that it is very hard for people to keep up with even the courts. It's absolutely the case once you're established and you have a workflow it kind of it's kind of like relearning how to do something right at the at first you know I was um I had to kind of retool the way I do things and the way I think about things.

I have to write a draft an email to a client that's you know more than more substantive with containing analysis explaining a litigation hold to business people. Usually historically I've sent it through waterfall method right you send it through in-house counsel in-house counsel uh disseminates across the company that in this particular company I wanted to make it very understandable as a bunch of software engineers so obviously step by step step one this is what you do step two this is what you do AI did a great first draft got great ideas for marketing so it makes you less mental mentally fatigued throughout the day right I felt certainly felt more you know mentally fresh later in the day because what I'm doing now is much more advanced um you know more legal analysis strategy exercising my judgment and I think that's the real value to it it frees you up to do more I find interesting tasks that are more a little bit more thoughtful and high conceptual that it's the exciting part of legal practice for me.

See my feeling is that it's using AI as a primary part of your workflow is not likely to save you a lot of time. There's a lot of discussion about that but I think it's gonna really elevate your work product. It saves me so much time so much time. I I I know we agree to disagree on that point but it saves me the two hours it would have taken me to to do that first draft for instance it's gonna take me 45 minutes to to repeat and that's mostly just writing the prompts right and three hours right that's writing the prompts right and I'm talk to texting those I'm talk to texting those and that's a great point also a lot of attorneys are very time sensitive from the big law of private practice our time is most of the time billed out in an increments and you want to be sure you have almost an internal clock if you've been working in private practice for a while how much time you're spending on something.

So say telling somebody to sit down and spend 20 to 30 minutes on writing a prompt it's almost like you feel like you're wasting time. That was initially my mental block where I was I feel like I'm wasting time this is taking too long but then you know it ends up generating something that as Heather said saves you multiple hours. So you know I think it does both right I think it increases efficiency at the margins absolutely I I don't think there's ever a case where it takes just as long with maybe some some exceptions here and there.

But I also think that makes the work product Ron I think you're also right where the work product is you're starting with a work product that is really sophisticated really great you don't have to always worry about training and this is not maybe not great the best for junior associates you always have to worry about training junior associates about you know getting um them up to speed and obviously there's a lot of challenges for people entering the legal market for that reason right now and I you know I'm not I wouldn't envy envy I I graduated during the recession wasn't a great legal market to to uh to uh graduate into but I think right now there's gonna be similar challenges for different reasons in terms of getting up to speed as an attorney is a lot more challenging um right now all right so now we're gonna turn to our normal closing segment which is the Flintstone Simpsons Jetson segment and I had something prepared but Matt I've been so impressed with your presentation here today I'm gonna turn it around I'm gonna throw it at you so the first question I'm gonna ask you is a Flintstones level lawyer we know they're they're the ones that are they're afraid to touch the AI stove because they don't want their hand to get burned.

How would you advise them to get started in bringing AI tools into their workflows? Honestly start watching YouTube videos about AI, different AI products. I think it's a great source I mean you know doing things around the house it teaches you how to do that right it's it's just as effective as learning anything really for that matter and honestly you can always ask an AI tool about how the best place to start with AI, right? And you know there's different market variants Heather talked about a couple of ones actually I'm I'm less familiar with in terms of more administrative tools but there's so many different ways it can add value to your practice and there's so many what ways I mean you're seeing this across a lot of different things where you're seeing with small companies they're sharing HR kind of functions and this is something that's been going on for quite some time.

But I think I'm sure that there's some sort of AI tool that offers cheaper rates for smaller companies or firms it's just gonna be I'm not I'm not I'm not as familiar with those because it's not nature of my practice but I'm sure there's stuff out there. It's just everybody's getting in uh at the ground floor there's so much competition. But yeah no I think the best thing place to start learning about tools prompt structure is very important and then just honestly like listening to podcasts like this Ron, I don't think I told you this but when I was re when I was starting to do research over the last couple of months about AI I think I came across this podcast back a couple of months ago and I think you had just just started it.

So you know I think that you know podcasts like these are great great tools to learn more about it. Alright so how about a Simpsons level lawyer they they've got ai they've got their their Claude Pro uh account they're using it frequently what advice would you give to a Simpsons level lawyer or law firm that wants to upgrade their AI practice? Honestly the best thing is practice makes perfect. I mean you just gotta what I do is uh I want attorneys in my group and a lot of tech companies are like this and I'm like this set aside two hours one hour a week to just like test use cases, research the research ways to use it you know talk to people about use cases.

I've had some associates think of brilliant ways to use AI. I mean they came up with um and there's eDiscovery platforms that do this as well with the automated AI tools and they're quite great depending on the platform. But I've had somebody create a vault in Harvey and instruct Harvey to do kind of a tiered relevancy analysis based on you know a number of here are the issues, score them from like one to a hundred with 75 to 100% of relevancy being high, then there's medium there's low err on the side of being low, here are the issues and it produced something that was really great.

And I just actually I asked somebody they presented on a call and I asked them to send me the prompt and I collect those prompts that you know throughout the firm and share them with my team. Right so so Matt, you're obviously a Jetson's level lawyer what's the last thing you saw in the AI universe that made you go, wow. Yeah I think the readme files is something I just learned about that last week. Right.

So it's the fact that you know technology companies in-house counsel is gonna have to start drafting documents in a completely different format or not have to but they're after think about and learn how to do that. And it's not something I was super familiar with and I think that you know attorneys are they're very heavy word practice PDF practice it's very document heavy obviously you've probably been using Word for the last 20 years maybe it's word perfect Microsoft Word something along those lines longer yeah longer I'm I'm 67 yeah so there's some sort of word processor that's being used.

So you know having those documents make then transferring them in the README files that's going to be a challenge. I also think building custom tools right I'm working with a a client right now to help them build a custom tool and they have their own tech department and I'm kind of building out playing input logic output and essentially putting that together and helping them think about those types of things and what they can do is remarkable right there this is this is this is the most advanced practice I've come across in terms of the AI functionality.

A high level Jetson's lawyer is most impressed by the um introduction of these readme files which are MD files. Yeah.md files yeah custom any sort of custom tools right that are being built like enterprise versions I think that I've been talking about this for a while is um building out like firm specific tools I'd like to do that more but then you know also it takes away are we a technology company or a law firm right and there's a little bit of tension there right as well but and then like also building out custom agents everybody should be having agents so that you don't have to write a prompt every single time you just drop it in you write a quick prompt click enter produces the document and it saves you as Heather said it might save you like 30 45 minutes just in building the prompt and it's a lot more streamlined so once you have a couple of documents templates and you've gotten more familiar with the tool you can definitely do that.

That's another thing that you could do is like a your words of Jetson style attorney should should all firms and all attorneys and any legal professional you utilizing AI for any legal work in you know within that ecosystem be architecting building their their AI workflows assuming that every interaction they have with AI creates a discoverable operational artifact? Again it depends on like your risk profile what kind of work you're doing maybe certain types of work is there's a little bit of a less risk there's maybe an emphasis of speed over protection but my work is the risk of disclosure is so prominent that I would be a little bit more conservative in terms of you know I'd be more wary of that.

But I think it kind of it's gonna be practice specific it's gonna depend on again risk profile the speed you want to accomplish what's your volume right it's gonna help you do a tremendous more volume as like a maybe a smaller shop and maybe you uh you know if you're doing cuts or transactional work where the risk of damage from disclosure is not necessarily as as high as my practice maybe you want to make that decision that you can you should definitely consider it but you know maybe it's not worth retooling your entire practice to make those considerations.

Well the part that I really want listeners to understand is that the goal is not to become Jetsons overnight. The goal is to move intentionally up the spectrum because firms that try to skip operational maturity often create bigger messes just with more expensive software. Matt Lafferman thank you so much for being here my head is blown it's been an incredible episode thank you so much. Great thanks so much for having all right and we will all see you next time on AI Tools for Practicing Lawyers.

That's it for today's episode of AI Tools for Practicing Lawyers.