The California Appellate Law Podcast
The California Appellate Law Podcast
Legal-tech guru Ernie Svenson on how attorneys should use AI
Just a couple years ago when we talked with Ernie Svenson, the attorney who talks tech fluently, AI was not even a thing. Now in late 2025, it’s the only thing. Ernie joins Tim and Jeff to discuss the rapidly evolving landscape of AI in legal practice, why AI gives small firms an advantage, and how attorneys can safely leverage these tools without falling victim to “hallucinations.”
We discuss how to embrace AI tools without anxiety (or with the appropriate amount of anxiety), starting with inconsequential applications before moving to more consequential legal work.
- Pattern Recognition on Steroids: AI excels at pattern recognition and language expression, ideal for first drafts and oral argument prep.
- Not an AI Problem: Recent sanctions for citing hallucinated cases reflect a longstanding due diligence issue. AI just exposes attorneys who don't verify sources.
- Small Firm Advantage: AI works best as a force multiplier for individual cognitive ability, giving solo practitioners and small firms who master these tools an edge over larger organizations.
- Agentic AI on the Horizon: While fully autonomous AI agents need careful supervision, basic applications like data entry are already available, with complex applications developing rapidly for case prioritization and KPI extraction.
Jeff Lewis
Welcome everyone, I am Jeff Lewis.
Tim Kowal
And I'm Tim Kowal. Jeff and I are certified appellate specialists and as uncertified
we try to bring our audience of trial and appellate attorneys some legal news and can use in their practice. As always, if you find this podcast useful, please recommend it to a colleague.
Jeff Lewis
Yeah, if you don't find it useful, run it through Chat GPT generate a summary and see if you like that instead.
Tim Kowal
That's right. Apropos of today's topic, Jeff, we are pleased to welcome Ernie Svensson back to the program. Ernie, sometimes known as Ernie the attorney, is a former commercial litigator and a nationally recognized legal technology consultant based in New Orleans. He is the founder of law firm Autopilot and the 8020 lawyer dot com.
Ernie Svenson
Thank
Tim Kowal
websites that help solo and small firm lawyers use technology to improve their efficiency and quality of life. And that is just the topic that we want to discuss with Ernie today. Welcome back to the program, Ernie.
Ernie Svenson
Hey, it's great to be here and see both of you guys together.
Tim Kowal
Yeah, last time we spoke, it's been a few years. And I think in just that short amount of time, I don't even think we even covered AI during that podcast. I don't think it was even a thing. So that'll just show how far we have come in just a few short years.
Ernie Svenson
No. Yeah, it's technology has taken a new level of increase.
Tim Kowal
Yeah, well, let's get right into that. I wonder if you can give us just your view on the current state of play of legal technology and you all fit in the room in that space What is actually working for attorneys? What isn't? Last time I remember we talked about some of the gateway drugs for legal tech. What sorts of things should we should attorneys who are not using them start to use? We talked about things like Sane Box and TextExpander.
Tim Kowal
I still use text expander. know Jeff still uses text expander, but the big elephant in the room is ⁓ AI and AI tools. so just give us your roadmap of where have we come and where are we now?
Ernie Svenson
Yeah, we've come a long way really quickly because chat GPT is only, it's not even three years old exactly. And a lot has happened in those three years and it's not just chat GPT, there's multiple AI things out there that people can pay attention to. I think the big thing that lawyers who are interested in this and everybody should be interested in it, the thing that they can wrap their head around, the thing that will help them wrap their head around it is I wouldn't...
Feel bad if you're if you have a sense of anxiety like I don't really understand that I tried it. I don't really know how it works. Even the people who created it don't exactly know how all it works. So there's a lot of uncertainty and a lot of vagary about certain aspects of it. So instead of trying to be feel bad and stay in a place with involves vagary and uncertainty just use it like make it as tangible as possible. That's what I try to tell lawyers to do like use it.
what kind of result you get, learn from that, iterate, improve, and the lawyers that I've seen working with it and improving because they're using it are getting more comfortable with it because they understand how they can use it in a way that serves their interest. so reading about other people who made mistakes or had problems or who use it in a way you're not going to use it isn't going to help you. Make it concrete and eliminate the anxiety by using it in ways that serve you best.
and talk to other lawyers who are using it. And that's what I do in my community as you guys know, cause you're in it. And when lawyers are talking to other lawyers using it, that solves a lot of the uncertainty, anxiety and so forth. do that. That helps.
Jeff Lewis
Yeah, let me let me add to that Tim using Chat GPT could be scary or using any of the eyes could be scary at first use it for inconsequential things like I use it to help you with RV repairs or things around the house And if you get comfortable using it for inconsequential things then we get a little more consequential, you know preparing your Supreme Court briefs with it
Ernie Svenson
Mm-hmm. Yeah, that's it.
Tim Kowal
Yeah.
Sometimes I have to remind myself to still use Google for certain things. You know, sometimes I think that because I have started to curate my chat GPT so much, you know, use for different specified purposes, and I've trained it for certain things that I don't want to gunk it up with like random things like what TV show should I watch tonight? What's a good movie night for my kids?
Ernie Svenson
You still use Google for certain things?
Tim Kowal
I'll ask
the chat GPT AI or a different AI tool because I kind of want to keep my chat GPT siloed off for my, for the many business purposes that I'm using it for.
Ernie Svenson
Mm.
very sensible.
Jeff Lewis
Yeah.
Tim Kowal
I don't that that's being paranoid, but there certainly are lot of things to be paranoid about when using chat GPT. Of course, it seems like weekly we are getting some new case where a lawyer has gotten sanctioned for using AI improperly, citing a hallucinated case, not being careful about checking the citations that were assisted by chat GPT. I wonder what you feel about, know, what is your reaction Ernie, when you see these cases regularly coming down the pike, do you think, well, this is just something else that's going to
scare attorneys off from using chat GPT and another reason that's going to cement them into, well, I'm never going to use it. You know, that that's just solidifies it.
Ernie Svenson (05:52)
Yeah, let me get up on my soapbox. I'm ready for this because so I, after law school, I clerked for a federal judge for two years. And, you know, what I did was read the briefs, the memos, you know, look at the citations, write a memo for the judge. And this is federal court, which was reputed to be better than state court when I was getting out of law school. And in fact, that generalization was true. And so I expected a higher level of lawyering and generally that was the case.
But there were lawyers and some of them in reputable firms that would submit briefs that where the case didn't say what they said it said. And I was shocked because I was told in law school that that was like verboten and very bad and you would get sanctioned. And this was so often happening that you couldn't sanction everybody who was doing it because it was just a common, not a common practice, but it wasn't unusual. Let's put it that way.
To me, that's pretty much the same as hallucination. You worked harder to find a case that doesn't say anything like what you said it said. So this problem has been going on for long time. So when people frame it as an AI problem, I think that's complete mischaracterization. Yes, it is true that AI use is factored into what happened. It's part of the sequence of events. But that doesn't mean that AI was the root cause of the problem. The root cause of the problem is complacency.
ignorance, don't know, lot of things, incompetence. And that's going to go on whether you're using AI or not. the lawyers who, and now what I think where I think it happens that it's a new, okay, it's new, we got to adjust for it, is in big firms. Again, I have experience in this because I worked in a big firm. I'm sure you guys know this too, that you'll be assigned to do it if you're the grunt lawyer on the case, the newbie.
And you're the one who does most of the writing. You're the one who found all the cases. The partners who come later, I'm going to site check and see if what you found is really saying what it needs to say and all that kind of thing. So yeah, the younger lawyers are the ones who are at the bottom of the production chain. They might be using AI in the wrong way and they need to be told, look, you you're to need to make sure that that case actually says what it says. But that's always been true. you know, numbers get transposed and you you don't submit stuff.
without being rigorous about what's in there. So that's shame on the lawyer who did that. And AI is just a new way of doing something that's been going on forever.
Tim Kowal (08:22)
Right, lawyers have always cited cases improvidently or incautiously. They've just looked at the headnote and then looked no further or found a quote snippet and looked no further and cited to the quote, cited to the case. And then when the judge asks, well, what about this case? It stands opposite for the proposition you cited it for. Well, Your Honor, just, I didn't read that case. I just found the words in it that I thought sounded good.
Ernie Svenson
Mm-hmm. Yep.
Right. Right. Right.
Right.
Tim Kowal
That's just a difference in degree, not in kind with missiding from chat GPT.
Ernie Svenson
Yep. Right.
Yeah, so I think it's sad that we spend so much energy talking about that because what that does for some lawyers, hopefully not too many, but it does, I'm sure do it for some, is say, oh my gosh, it can go wrong. It's not safe for me to use it. It's like, well, now here's how you will know if it's safe for you to use it. Do you have a functioning brain? Are you capable of thinking critically? Are you skeptical about propositions?
Do you vet things? Do you check them out? Are you cautious? Okay, well then it's not gonna be a threat to you. Like that's, it's that simple.
Tim Kowal
Yeah. Do you have any must-dos or must-not-dos when it comes to AI and ChatGPT or any of the lawyers that you consult with on a regular basis at all levels of ⁓ ChatGPT embracing some who are just putting their toe into the water and some are looking to make more aggressive use of it? Have you spotted any trends that AI is just not suited for this type of application?
Ernie Svenson
Mm-hmm. Mm-hmm.
Tim Kowal
But AI is definitely suited for certain other applications. I remember one thing that I believe we talked about last time you were on is workflows, standard operating procedures, and you used to be big on documenting them and how important that was to bringing efficiency to your law firm by documenting your SOPs. And now what has ChatGPT and AI tools done with that bedrock advice? I think that's turned it almost upside down.
Ernie Svenson
Mm-hmm.
Yeah, I think there's three areas that are fertile. There may be more, but let's just pick three. Document drafting, first draft, going from the crummy first draft, which is hard to create and requires a lot of cognitive energy. And then once you've spent that cognitive energy, you don't have much left to do the next level. So chat GPT or any of these tools are really good at cranking out a first draft.
And first drafts are meant to be not that great. So the fact that it's not that great doesn't mean anything. It's inconsequential. So that's a good inconsequential or minimally consequential thing to do. And then from there, when you have the draft, whether it's in an outline form or narrative, doesn't matter. Your brain, if it's functioning properly and has critical thinking capability,
We'll say, you know, I need to change this. I don't like that. You'll jump into action. So it'll give you a lot of momentum quickly. That's a no brainer. The other one is summarizing. So note taking is one area where you need summaries of notes. So if we're meeting, we're talking about a bunch of different things, you each person can be devoting some part of their cognitive energy to try to capture notes. Or they can spend more of their time talking and thinking and then let the AI capture the notes and maybe
You you supplement the notes here and there with like in the case of granola, which I know you all know about and use granola has that feature where you can type in your own notes and then you let it do its thing and it captures your notes and the notes from the conversation. That's a no brainer. mean, like who doesn't want that? Everybody wants that. And it's really good at that. And if it doesn't capture it perfectly, which it won't, it'll miss, it'll misspell somebody's name. It'll misunderstand something, but you can clean that up. And that's way less effort.
to do it that way than to use your brain to try to do all that node capture in real time when you're having a conversation. So that's one. And then the one that I think people don't think of, but once you start using it more and feel comfortable and you think, okay, what else can I do? You can use it to coach you. And coaching can be any one of a number of things. But one example, and I forget who this person is and we can search and find it, but there's...
There's a woman who used to be a federal judge and then left that world and started working in private practice for a big New York firm. And she appeared on a Bloomberg law podcast. That's probably two years ago. And she was saying AI can work and here's how it worked for me. And so she was going to give a ninth circuit argument. And once the panel had been picked, she's like, well, I want to know how these judges are, you know, what kind of questions are likely to ask me. And she fed everything she needed to feed into chat GPT and said,
tell me what the kind of questions are that I might be expected to face. And it told her and she was shocked at how good they were and how much better she was prepared when she went to the oral argument. she, her explanation was, said, this is something that would have taken a senior level associate at least a couple of weeks to do. And chat GPT did it in a matter of minutes. So those are the kinds of things where you consult it and have it help you think better about how to do things. Like if you're going to do a direct examination, here's what I'm going to ask.
Give me all the questions I need to ask formulated in a non ⁓ objectionable way. But if I get an objection, tell me how to respond to the objection. It can do that again. Is it going to be perfect? No. But if you're using your brain, it'll get you ⁓ going and you'll know what to do next. So there's a lot that it can do, but those are three examples.
Tim Kowal
Ernie, when you mentioned the example just a minute ago of asking ChatGPT, I assume it's a tool, something like ChatGPT, how, you know, here are the members of my panel, how is this panel likely to rule on my issue or what questions am I likely going to get from this panel? It reminded me of a tool that was available a few years ago. I'm not sure if it's still available, but it boasted this ability. They created a database of all of the...
Ernie Svenson
Mm-hmm.
Tim Kowal
opinions
that all these various federal judges had created and then you subscribe to it and you tell it what jurisdiction you're in, what judges are on your panel, and it would predict for you what type of questions you're going to get. it going to be, are they going to be on an originalist bent or a purposivist bent or a textualist bent? ⁓ What type of public policy are they going to be concerned with and all the rest? It was a very impressive looking tool, but now here we are two, three years later.
Ernie Svenson
Mm-hmm.
Tim Kowal
And I think Chat GPT-5 can do all of that on its own without any shell overlay to it, pretty much, probably get 90 % there. And I wonder, something that you mentioned, I think it's something that you passed along an article about, where is AI going? And we almost don't know. And so I wonder if it would be nice if there were nice polished AI-based apps that we could, that were ready to use for...
Ernie Svenson
Mm-hmm.
Hmm.
Tim Kowal
for legal practice, but I don't know that anyone's going to want to develop AI apps until we know that the underlying AI engine is not going to eat their own lunch, take away their own functionality in a year or two hence, because with Chat GPT 5 that just came out, I think two months ago, mean, it's head and shoulders above Chat GPT 4, which was amazing. So I wonder what you can, do you have any comments on that and what we can, yeah.
Ernie Svenson
Yeah. Yeah.
Fine.
Yeah, yeah, wouldn't,
I wouldn't, I wouldn't, I hear people tell me, cause they'll say, oh, you know, you're a lawyer and you know, but AI and I've got this brilliant idea how lawyers can use it. I'm going to build this application. I'm like, really? I don't, I don't know if I would do that. I were you, because everything is becoming commoditized and made, you know, more, the information is becoming freer of previously more manageable constraints that you could charge people money to get. Right.
And so when it comes to something like what you're talking about, and I think I forget the name of that service, but I remember it coming up and looking at it and thinking, that's genius. But when you break it down into its component, what you have is a database, which I'll call a rag. You know, that's like in the AI world, there's like, Hey, I go out there with your training and just search the entire ocean of your information that you know, and come back with a great answer. And sometimes, you know, a lot of times they can do it that way.
But sometimes it's going to actually not work so well because what you want it to do is be focused on a certain database. So like if you want to know about judges, you you don't want it ranging around the whole internet because there's a lot of noise there. You want it to go, okay, here are the opinions of these judges and only look at that and then tell me what questions they're going to ask. That's, what's called retrieval augmented generation. So that's how you go about that. Well, how hard it. Yeah.
Tim Kowal
That's the acronym RAG that you mentioned, Retrieval
Augmented Generation.
Ernie Svenson
Generation.
Yeah. So like how hard is it to create the database to create the database is the challenging part, but it's been created. It's been created by Lexis Westlaw, whoever all else case text, fast case, and even Google scholar has a lot of those cases in there. So if you can just get, you know, and you don't the database for the questions, for the example with the questions to the judges, you don't need every case that every one of those judges.
has been involved with, you need a good sampling of them and every case would be nice, but you don't need every case. 80 % of them would be probably good enough or whatever. Some number is going to give you enough to discern or for the AI to discern what kind of questions they would ask. That's out there. That's increasingly acquirable. So there's not much of a barrier there to entry. The barrier to entry for using AI is, I can do that.
Okay, well, that's an awareness thing. And then, okay, how do I do it? That's, you know, execution part. And then discovery like that it works or you need to adjust for it. That's it, you know, a little experimentation adjustment. That's where it's all going to be happening. And people who try to build the grand thing that solves the problem for a lot of people and remains that solution to that problem for a long time. I don't know, man, this stuff is.
Developing really quickly, as you pointed out, like, don't know that there's a lot of longevity and thinking you have a moat around some brilliant scheme you have to use AI in a certain way as a company. I mean, as an individual, you know, we have wonderful things, opportunities.
Tim Kowal
Now Ernie, you're a big law expat and I wonder what are your insights on how big law will embrace AI? Is AI going to be ⁓ more of a force multiplier for solo and small firms than it is for big law? Because I know that one challenge with the law business model is to scale. You want...
Ernie Svenson
Yes.
Tim Kowal
junior attorneys doing all the grunt work and then senior attorneys above them refining it and then the partner's taking credit and the big bonuses for all of it and AI is going to make it more difficult to scale especially when you look at it through the eyes of the client who's gonna start wondering, why do I have to pay you for doing all this grunt work again at $600 an hour when look, I could do it with ChatGPT.
Ernie Svenson
Right.
Right. Yeah. Well, I think I have X-ray vision into large firms and large organizations because anybody who's worked there, there's a certain dynamic that's persistent. It's not too variable. The size makes certain things happen in a more pronounced way or whatnot, but it's the same dynamic. You have the power structure and all that kind of stuff. And there's a lot of reports about, like, for example,
Big firms that use, example, Harvey, which is one of the shadowy, expensive, filled with promise, but nobody really understands what's going on. Harvey is only available to big firms. It costs a lot of money. Nobody knows how much it gets used, yada, yada, yada. Well, we can kind of figure out mostly what's happening without paying for Harvey. Here's what's happening. It's expensive. It was rolled out at scale. Some partners at the meetings,
We're reluctant said fine, you know, go ahead and spend the money but I object some said, you know, it's a great idea and these young associates will have it and they'll be geniuses and we'll have less of them. We won't have to pay as many of them and it'll be wonderful for us. So a lot of people at the top don't really know how any of this stuff works. And they're just thinking we'll buy the solution will come in and the younger associates of which there will be fewer and more profits for us. But if you're not paying attention to the people using it and you don't really know.
how it's being used or what is good use or bad use, how much traction is there really in those firms? And the answer I think is not much. And what people report is they'll say, there's only 30 % use in not just law firms, but organizations. Well, is that 30 % use, which is self-reported higher or lower? Is it real higher or lower? Like are there some people are going, I'm gonna use it, but I'm not telling people I'm using it I wanna take credit for being a genius.
Jeff Lewis
Hello.
Ernie Svenson
Or are there some people who are claiming they're using it because they feel like they're supposed to say that, but not really using it because they don't care for it. Maybe those two balance out and, but there's some mishmash of all that stuff going on. So the real, the answer to your question, is this a force multiplier and is it more beneficial to smaller firms? I think the answer is yes, but not because of all those reasons alone or even at all, but because AI is a force multiplier for a single cognitive entity.
A person who's using it and thinking, a second, it didn't give me the answer I wanted. Now I need to divert and do this. And that's been true for law all along. you know, every one of these things we do, it's an individual solving the problem and working through it. Now you go report to somebody else and they pick up the ball from, where you carried it and they move on. But all of this is cognitive critical thinking stuff, right? You don't force multiply cognitive critical thinking.
with groups of humans, okay? Just take a look at what happens in committee meetings. you know, committee meetings are where great ideas go to die quickly. And, you know, so it's not...
Tim Kowal
That's the expression,
designed by committee, as a default negative connotation.
Ernie Svenson
Yeah, yeah. So I think it's
small firm, solo lawyers. This is your time, but you got to go use it and learn how to use it. And then you will be better than any other individual lawyer who doesn't know how to use it. And if there's groups of them and there's group cluelessness, that's even better hunting for you. That's how I think it works out. I could be wrong. I don't know. I hope I'm not.
Tim Kowal
And while you were talking Ernie about, ⁓ you know, how, how a big law might be throwing some money at something like ⁓ the Harvey example, which I've never used it by heard it described something kind of like a chat GPT for lawyers, basically had some, some guard rails into it. And, you know, the, partners don't know what it does or how it's going to return the money, but here we are in the AI world and we got to start throwing some money at it. Reminded me of my brother works for a tech company that, ⁓ that places ads on.
Ernie Svenson
Mm-hmm.
Mm-hmm.
Tim Kowal
on videos on the internet. And in the early days, big companies knew that's where advertising was gonna be, but there was no way to, there's no transparency. You could just dump money at these companies and they'll come back with a report at the end of the month and says, we placed your ad at a thousand websites. But no one's going around and personally checking them and they had problems with fraud early on. Videos would play in like a one by one pixel frame in the website and.
Ernie Svenson
Right. Right.
Right.
Tim Kowal
The algorithms would report it as played, but there are lots of examples like that. But in the early days,
Ernie Svenson
Right. Right.
Tim Kowal
the corporations knew that, okay,
this is brave new world. We're just going to have to throw money at it and just pray that it comes back to us. But we know it basically is because this is where advertising happens now. But with an AI world, yeah, like you said, I think the most of the leverage is going to be at the smaller unit, at the individual lawyer unit, a force multiplier per brain.
Ernie Svenson
Bye.
Tim Kowal
not per organization.
Ernie Svenson
Yeah, because
an individual lawyer can't help but know what's going on or not going on. They either know like, I don't know how this works. I need to learn it. Or I do know. But at a big corporation, you've got this diffused cluelessness. But I'll say about reports because this is important to get out. Bernie Madoff put out a lot of reports. Reports don't mean squat. But if you're there seeing the data collected, you're like, this isn't working the way I expected it to be working.
Tim Kowal
What are some of the knowns and unknowns with AI? There's certain things that we know it does pretty well and certain things that it seems to struggle at, but maybe it's getting better or maybe, and some of us are desperate to find what is that one thing that only humans can do and will ever be able to do? And AI will never replace us when it comes to X. So now we just have to solve for X. What is, in your opinion, at this point in October, 2025, knowing that everything could be ⁓
Ernie Svenson
Yeah.
Tim Kowal
could be upside down by the end of October 2025 even. What's AI really good at and what's it really bad at?
Ernie Svenson
Yeah.
It's really good at generic kind of stuff. Like it's pattern recognition on steroids. That's basically what it is. And that alone is magical, right? Because we, humans, we like to think we're good at picking up on things happening and we see patterns and you know, we believe that that's true of us, but we're not as good at pattern recognition as computers are. mean, please, that's what they do. So.
If they can do pattern recognition at scale, which is what AI is, and if they can create language expression and explain things in the way that they do, which now they do, and they can receive our questioning and use that to go in different directions to look at different patterns, then that's what's happening. What's not happening inside of AI is any kind of thinking. That's not pattern recognition can help thinking, and it's maybe part of thinking, but it's not
the highest level of thinking. The higher level of thinking is having brilliant insight. Like, wait a second, this doesn't fit here, that fits over there. So AI is not doing that now and probably won't be doing that anytime soon. And coming up with a good solution that, know, when you've practiced, it's that thing of the 10,000 rule, 10,000 hours, it's the beginning of this explanation. Like when you've practiced law for a really long time, you've seen a, you have a bunch of reps.
on a bunch of different scenarios and yeah, you know, can go look things up, but a lot of times you already know the answer before you even go look it up. You just know, right? You're like, ⁓ you know, the answer is going to be this. Then you go look, you confirm, you go, okay, that's what I thought. I have support. Great. AI can't do that. AI is working off of actual tangible stuff. And a lot of times if you look at things, the answer as, you know, explored by a young associate who gathered all the information would seem to be X.
when in fact the person who's wizened with the experience goes, and it's really more like why, you know, I just, and maybe they can't explain exactly why, but you know, that instinct, that domain expertise is really, really valuable because that domain expertise can tap into the AI and use that to get their work done faster. But that domain expertise is what's going to be more valuable to clients is going to create better outcomes.
And AI can't do that. I don't know that it'll be doing it anytime soon. Maybe it will, but for the foreseeable future it won't. So if you build your castle on the foundation of I'm looking at things and getting context and insight and optimizing for that, assuming that you have the cognitive ability to do that, which sadly not all humans do. But if you have that,
you're safe. you don't, make plans for universal basic income.
Jeff Lewis
Okay.
make plans. Hey, Ernie, let me ask you this. You know, I think as of October of 2025, most listeners this podcast understand the value of chat GPT in terms of working with a lot of evidence or working with the law. The next big thing is agents. You know, my favorite program notion has agents coming out. What do you predict? How do you have a prediction about how lawyers can use agents as the next big thing in their law practice?
Ernie Svenson
Mm-hmm.
Jeff Lewis
⁓ to make them more competitive and better lawyers.
Ernie Svenson
Yes. So agentic AI, which we hear a lot about and it's, it's coming, it's here, whatever it's, it's here for people who are programmers who have a high degree of expertise in automating things and understanding that in part of automating is you pay very careful attention when you first trigger the automations and then you continue to pay close attention because unsupervised automation can go out of control very quickly. And
⁓ most lawyers don't have that expertise and what they do is they come in with the enthusiastic expectation and they hear only the words around, you don't even need to supervise it. It'll do all this stuff on its own. Yeah, that could be true, but you don't start with that. You start with being very careful about unleashing the agent. And so I don't think it's going to apply to most lawyers. think it's, you know, just know that it's coming. and what lawyers should do.
is prepare for that by using AI. And the more you use it, the more instincts you'll have for how to manage it in an agentic way. The first place where I see it happening in a kind of a primitive way is in email drafting. like if, and this is gonna be in Copilot for Microsoft, it's already in Gmail. It's coming, you know, the thing where it looks at all of your emails, it knows how you write, it knows what kind of emails you tend to reply to, and it can then start to say,
OK, I'm going to take a crack at creating drafts of emails that Ernie, Jeff, and Tim haven't looked at. And I'm just going to put it in the drafts folder, and you're to go look at it and say, ⁓ OK, that's pretty good. That's great. I'm off to the races. And so you didn't have to tell it, hey, I need a draft email for this. It's already prepared. That's agentic, right? It took action. It was unsupervised. But it didn't send the email unsupervised. That's no bueno.
Tim Kowal
Yeah.
Ernie Svenson
You know, like,
and maybe there'll become a day where for certain kinds of emails, it'll do it and you'll know that's fine. That's a couple of years off, but I think in email drafting is where we're going to see it first.
Tim Kowal
I'll share two quick examples of how I'm experimenting with agentic AI in Notion. Notion just a week or two ago in September 2025 released their big upgrade that includes agents. And I clip a lot of recent news articles, legal news articles and cases that I might want to work up into a blog post or ⁓ a case summary that we talk about on this podcast. Sometimes that list grows and grows and grows and I don't have time to read all of it. So I have an agent that will
I tell it to look through all of my database of the cases that I've clipped and give me a salience rating, one, two or three stars, depending on how relevant it would be to my audience of trial and appellate attorneys covering a unique or a trap door in California or Ninth Circuit appellate procedure. And it will do that. It'll give me one, two or three star. here, don't miss, your audience should not miss this. And that'll help me to prioritize it. Another way is capturing my KPIs.
which like many lawyers, all my billing information is contained in Clio. And I like to get that out into Airtable or even Notion has ways of being able to pivot your tables and everything so you can see your quick charts, customized charts of things. through Comet, which is a perplexity based browser, I'm having my assistant,
who would manually do it, he's now experimenting with using the browser, the perplexity browser, which has an agent built in to take the information from Clio and put it into my Airtable forms, so as to take away that brute force task of having to manually look up every data ⁓ point. Now, it hasn't been 100%, in fact, we have a lot of errors in the first pass, but we're gonna experiment with it and report back and see if we can get the agent to...
be able to manage our KPI data in a reliable way.
Ernie Svenson
Yeah, that comet browser is great and they just released it so that everyone can use it. Even people are using the free version of perplexity, so if you want to check it out, go check it out folks. It's worth it.
Tim Kowal
Yeah. Okay, last question. Ernie, is there a tech habit that you have adopted this year that now you can't live without?
Ernie Svenson
Yeah. Well, okay. I'll call this a habit. The use of Wispr Flow, which is a tool. And that requires a habit of saying, now I'm going to dictate or I'm going to say something. And I wanted to turn to text and I don't want to have to type it because that's annoying. I just hold down a button on my computer. I blabber and the text comes out. And because of the AI use it, it captures the essence of what I'm saying. Or if I change my mind in the middle, it doesn't.
Verbatim transcribe exactly what I'm saying that captures what I'm trying to say Wispr Flow is like game changer and I just showed it to a friend of mine who was staying with me who has Parkinson's and He therefore has trouble typing and he's a doctor and I said look, know, check out this Wispr Flow things like my god This is a game changer, it's a game changer for everybody because who wants to type when you can speak and have You know the words appear
Magically wherever you need them to appear so Wispr Flow check it out. It's not spelled with the H and Wispr. It's w I s P R No, either SPR Flow and check it out and my god, so good
Tim Kowal
Okay, Jeff, unless you have any last questions, I'll take us out.
Jeff Lewis
No, thanks so much for joining us, Ernie.
Ernie Svenson
No, thanks for having me. I always love talking to you guys, especially when we all talk together. All of our beards, I don't know people can see this, but we got three beards lined up.
Tim Kowal
Yeah, yeah, the bearded episode. Okay,
Jeff Lewis
Fantastic beards.
Tim Kowal
that'll wrap us up. If you have suggestions for future episodes, topics or guests to bring on, please email us at info at calpodcast.com. In our upcoming episodes, look for tips on how to lay the groundwork for an appeal when preparing for trial. See you next time.
Jeff Lewis
See you next time.