S2|E2: Building AI-Native Law Firms: Insights from Helen Fan

Speaker A: Nobody was prepared for AI agents because we don't know it. It's a brand new, you have to learn it. So there's no playbook. Like I had a question. How is AI going to fundamentally change my work? And I couldn't sit still until I found out. So I started building and I kept learning and I think that's the.

Speaker B: How do you stay up to date with AI news?

Speaker C: Welcome back to Future Ready Lawyer AI and the Evolution of Legal Practice. My name is Mark Bennett and I'm um, here with Amin Ali Madani. Just a little bit of news. As you can imagine, doing a podcast is quite a lot of work and Alex has a very large workload in her job doing all the amazing work that she does and so she doesn't have capacity to continue as a co host. So we'll just continue with me and Amin this episode. We are really delighted to welcome Helen Fan onto the podcast. Helen, could you introduce yourself to our listeners? Yeah.

Speaker D: Hi everyone. So I'm Helen and I'm a California lawyer working at a uh, boutique law firm in Silicon Valley. And uh, and I'm Also uh, legal AI enthusiastic. So I'm keep posting some experiments about OpenCloud recently and I'm also a community builder. So my community uh, is called the Silicon Valley Legal Tech Frontier Community. And it's a cross border community and we already have 1,000 members across the US and China.

Speaker C: Yeah. And you're very active on LinkedIn, sharing some of these insights and showing people how you're building that community. And it's, it's really inspiring to see this experiment you're doing with your uh, legal AI lab. So we'll get onto that in a moment. But first of all we just wanted to get a little bit more about your background. So you're a California lawyer based in San Jose, uh, practicing cross border transactional work. But you started off in some of China's largest law firms. So what was that like and how did you make that shift across the Atlantic?

Speaker D: Yeah, very glad to share my own experience. So I started my legal career in China. I studied law at Qinghuang, Uh, it's one of the top two universities in China. So after that I worked at Founda Partners, which is one of the top law firms in China. I was doing MA and uh, financial regulatory work. So many of my clients are large international financial institutions such as Morgan Stanley. And I worked as a secondee in Citibank for half a year. Um, and after that I came to the US for my LLM M at Columbia and the timing was perfect. Uh, ChatGPT had just come out and I started using it heavily. So I entered two legal tech hackathons and LED bosms. Uh, one hackathon was hosted by Tsinghua University. I built a cross border data compliance tool. I joined the hackathon remotely. Uh, that tool used AI to help clients generate binding documents for China's government on cross border data transfer between China and U.S. uh, and at uh, Columbia's hackathon which I think were their first legal type hackathon ever. So I built a ah, legal assistant for minority groups and I got second place in both hackathons. So um, these two projects were a turning point for me. I wasn't just reading about AI anymore, I was building with it, solving real legal questions. Uh, and I could feel how fast the things working. Uh, well the curiosity already started this feeling that AI is going to like fundamentally change how legal work is done and it's happening now. So um, yeah and after Columbia I moved to Silicon Valley to get reunioned with my husband and I started doing cross border transaction work and rental capital things, stock financing fund function, uh, very typical Silicon Valley legal services.

Speaker A: And I realized uh, a huge chunk

Speaker D: of my clients are AI founders and builders.

Speaker A: So they have a very typical uh,

Speaker D: thing they trust AI deeply and here moment that uh, I think changed everything for me. Um, I was on a zoom call with a client and discussing a uh, vesting schedule on the uh, California law. Uh, and I told him like a four year vesting with one year clip is definitely a very standard practice. Uh, and he was like, he didn't believe me. So he shared his screen and opened ChatGPT, typed the same question there just on the call just to cross check my answer. And then he said something I've never forgot. He said um, even if you are right, my employees are going to ask ChatGPT anyway and um, they will trust that answer more than yours. So that was the first time I felt maybe one day AI is a real threat to my authority as a lawyer. Not because I smarter but because you

Speaker A: know, people just trust it.

Speaker D: And that's one curiosity turned into anxiety like kind of the work I'm doing now to the actual sustain. So um, yeah that mix of curiosity and anxiety is what pushed me to act. So I started my legal app. Um, it's not a real company, it's just a um, personal brand or personal studio, whatever. Um, I just you know I want to put my name on it because I want to make it as a serious thing. I committed to doing some real exploration about AI. So um, yeah, out of that legal AI lab so came the community, the meetup, the podcast and the experiments I'm doing now. So everything I do starts from a question I can answer on my own kind of work I'm doing to the accurate. Yeah, that's pretty much about my background.

Speaker C: I really love that idea of curiosity and anxiety. I think it's exactly what most lawyers are uh, thinking. Uh, obviously in different measures some have got more anxiety than curiosity and vice versa. But yeah, I think it's a great state to operate in and want to see what's happening out there with AI and see its capabilities. So you mentioned that you've got this Silicon Valley legal tech frontier community and having these meetups in Silicon Valley and also on the other side of the Atlantic. What kind of conversations are you having there and how are they changing how you're thinking about AI and law?

Speaker D: Yeah, so I think big one AI adoption in legal work is much more uneven um, than big and about uh, structural things, not cultural things. Like first I think um, like big law firms still run on um, billable hours and when AI makes you faster that actually a threat to their revenue. So what I say is they hourly raise to compensate. But they are not fundamentally changing the model. They are not embracing threat fee, they are protecting the old structure. But for smaller firms, middle sized firms, political firms, um, that's where things are actually moving. Like fixed fees or flat fees are already becoming the trend. And these firms don't have the luxury of raising rates so they have to get more efficiency. So uh, I think they adopt AI more and more. Uh, we just saw this with a clock called native law firm polls. Uh, I think both of you, I've read it through the viral post about the crowd code. Native law firm. Yeah, so that's a small law firm with only two people. Um, not a big law firm. And about in health legal, I also have my observations. I think um, because they are a ah, co center so they need to do more with less. Uh, big tech companies build internally. Like Google has their dedicated engineering team just to optimize their legal department. So sometimes they don't need vendors. So the real opportunity for legal AI vendors I think probably lie in middle sized companies or middle sized firms. The ones that need the efficiency but don't have the engineering resources to build it themselves. Yeah, that uh, I'm able to see these patterns because our committee has a lot of uh, lawyers, CDR councils and of course CD engineer managers in big Tech companies in Finca Valley. So these observations come from real conversation, um, not from reading reports. Yeah, oh one thing else because my community is a cross border community so I have like the opposition of both US market and China market. So uh, I've seen different things in Beijing. Like China's nickel type market runs on a uh, different end entirely. Um, the main players are ah, the governments and the court. So I wrote a full article about this online substack where legal tech in China evolves differently from the U.S. um, I'd recommend m it to anyone who wants to study AI adoption structurally. Not uh, just to compare two markets, but to understand how different systems shape adoption in fundamentally different ways.

Speaker C: So we'll put that in the show notes and I'm sure many of our listeners will want to, want to check that out to uh, to get some of those insights.

Speaker B: Helen, so you have published an AI native law firm roadmap. And um, can you tell us about the origin story of openclaw Law llp? But before that we just want to make sure everybody is on the same page. Would you tell us what does it mean when we refer to AI agent? And what is OpenClaw in particular?

Speaker D: Yeah, so um, OpenCloud is an open source platform that you can run your AI agents. And what makes it different from ChatGPT or Cloud? Is that so ChatGPT is more about uh, you ask a question and you get an answer. It's very simple. But agents or openclouds, um, there's something like uh, your own personal assistant that can really uh, do something for you. Like it can connect with your files, your web, uh, your, the website, your emails, your calendars. It can book a meeting for you, it can buy tickets for you. So um, it's totally different things and I think it can act on your behalf. It's not a chat window you open when you need something. Yeah, yeah.

Speaker B: So technically it would give a language model access to like your files and Internet to take actions on your behalf. And OpenClaw has made it easier for people to arrange that. All right, so would you tell us about the origin story of the openclaw llp?

Speaker D: Yeah, very happy to share. So uh, it started because uh, I was using color code, uh, which worked fine. Very good tool. The OpenCloud hit number one on GitHub, uh, just within two mantras and I think, okay, I had to try it. So at first I set up a clean mark, installed OpenCloud with a uh, Vue Apple account. So there's no my personal data Yet

Speaker A: M. And um, then I connected two

Speaker D: agents through this code. I give them roles. I have one called Morgan. Morgan M is my senior associate and Creel is my first year associate. So I have two agents and I gave them a real, real question. I said, how can I set up an um, AI law firm in America? Lawyers can hold equity, so this is a hot topic. And I figured I'd guide them step by step. But that's not what happened. Like within minutes they were scoping the research, dividing the work, setting deadlines, and just talking to each other like, like a team. And what is most surprising about Morgan told Cleo, like, Kyla is building a business. She doesn't just need to know, uh, what's possible, she needs to know what's prudent. So that's one. It stopped feeling like a chatbot. That's when I thought, okay, that's a law firm in a group chat.

Speaker B: That's so interesting that you're saying that you use two AI agents and they work together. So these are kind of the new trends at the moment, when people use the models at higher level and allow the models not only plan but also communicate with each other and then take actions. You said you described watching Morgan assign tasks and then clear flag issues and the two of them built together. How did that change how you think about, um, supervision and responsibility in a genetic environment? Like because they're taking action, right? And sometimes they may send an email or invite someone to a meeting or something that you didn't mean to. How that made you think differently about supervising and responsible responsibility in terms of

Speaker D: agentic AIs, honestly, it's about panic. Then it came with curiosity and then more panic. I will take uh, sub agent as an example. Like initially I didn't notice my agent cans for more sub agents. Clio found it by itself. Clio did a limitation in well paired tools. So instead of stopping, she diagnosed a problem, listed three options and proposed boring.

Speaker A: A, uh, sub agent.

Speaker D: And I didn't even know that was possible. As a builder. Yeah, it's definitely an incredible, um, moment. Uh, you have parallel research shared complex questions running simultaneously. Um, but as a lawyer I will think, okay, who's responsible when an um, agent creates another agent? Uh, where does the accountability end? So this one I realized the lawyers now have a completely, uh, new job. You have to write security rules for your AI. What permissions does the agent have? Can it read buyers outside its sandbox? Can it write real data? So I had to write rules, uh, write something like only if it's necessary. You can uh, spawn more agents and uh, subagent cannot run more subagents because only one layer. And um, when you decide to generate sub agents, no more than three. So uh, I just set up a lot of guardrails to control it in my scope. And I think these security decisions players have never had to make before.

Speaker C: Yeah, it's such a great way of describing that. We started with anxiety and curiosity. Now we've got panic, curiosity and panic again. And I think any, any lawyer who maybe has just been playing with chat, GPT, Claude, possibly even deep research with uh, it, which is a kind of agent to ki. So you're getting a little bit of a sense of okay. Now when I spin up a Claude Deep Research, it tells me, okay, I need some sub agents to, to go away and research the different things. And maybe that's fine, but this idea of, okay, I've got to buy a whole new computer because who knows what's going to happen with this thing and then I've got to see what it does and on the fly come up with further guardrails. You can see why it's such an important thing for someone like you to be setting this up as an experiment in sort of seeing what can go wrong. And we've seen a lot of people talking about what can go wrong with openclaw. Uh, deleting all of your family, family photos was one classic example. Uh, so yeah, that whole Mac Mini idea where you just get a whole new computer, whole new Apple account, no personal data, that seems to be something that reduces the panic somewhat. But just it's quite interesting to think about how this technically works in a little bit more detail. So these agents are being spun up their brain sort of is a LLM, right? Like, so you're connecting in with one of the main providers. Like how do you figure out how to do that? You install OpenClaw, but how do you decide how is it going to run exactly? And how do you pay for that as well? Right.

Speaker B: So just to put it in more context, when we are uh, using Open Claw, we don't have the kind of restrictions that we usually have in normal generative AI models like ChatGPT or Cloud, because they predicted that the models may just uh, talk too much and they put kind of safeguards to stop it when, especially when there are two agents, which means a model is taking two different Persona. Uh, it seems it's talking to someone else so that conversation can go forever. And as these things are being automated and the whole purpose Is so they. You put it on and they. They go and do the task and we are not observing all of them. That can be quite dangerous and costly.

Speaker C: So the way you've set up these two agents interacting with each other like how do you set that up? Because I know that within we have these ideas of skills or soul MD instructions that the LLM will be applying as it you know does its agentic functions. Is that how you put those guardrails in, put those Personas in and set it up the way you want it to work?

Speaker D: Yes, because uh first I, I uh make Morgan thinking like more like a business advisor, not just a senior uh associate because sometimes uh loan litigation lawyer. So most of my job is to find like the commonplace between the business side and legal side. So I like to know why Morgan and Creel. Let's just get an argument so I can see well the business side ask for and where the legal side ask for so I can see the trace back their arguments and to find the I think the most uh, uh. The most uh suitable solutions for my clients. Uh Morgan can set risk flags and rules thought about how to present option before ask it that pretty much what a senior lawyer can do. But I just put it uh a step further so it can work like a business advisor. It can search on some like business news and uh analytic resources to help me to like get um, like a more comprehensive picture of a legal question. And that is uh why I want to split one agent to two agents. And the other point is that uh you have multi agent you might reduce errors because it makes errors traceable. So when two agents disagree each other you know exactly where to look and main job shift from review everything to review the disagreements first. Uh, I will give you one example. Like during the first assignment I asked them about how to like up um an AI native law firm. And they just got an argument about uh. I think it's a model of legal zoom. It's uh native law firm and thinks uh yes but clio just push it back and when I trace back to it and I thought okay maybe there's no clear answer in the like on their research. And based on my research I also uh. Like doing that is a very like a vague point. Uh the practicing and the academic or whatever like not totally resolved. It's not clear. So I can see, I can tell the blind spots the AI may have. And it's also may become uh something that uh not being really resolved in practical world.

Speaker C: It's really interesting and I'm interested In how you work as part of that team. Like you can see what they've got up to, what their debates were. And then obviously you go back and look yourself at the materials they're mentioning and come to a human perspective. Do you then get back into the debate and point them in another direction? And also like another question related to that is as you are in this kind of dialogue or observation of what your agents ah are doing, how much do you go back in and change the instructions from when you first wrote them? Like how much of a uh, iterative process has that been or did you get it right the first time?

Speaker D: I will not change my instructions often, but I will change the sole MD often because every test you can find it cannot work as you expect. And also I will ask them to generate more skills. Like I will say okay, so what do you learn today? And can you tell me do you think what your, what your thought about uh, my input? So in this way I can get a uh, self learning data like the skills and the other day updated skills so the skills can be updated by the agent itself. And um, yeah, that's pretty much about

Speaker C: it just in terms of those two concepts and we've mentioned it before but the sole md, what role does that play?

Speaker D: Not a tech expert. I have no tech background at all. And for me I will take it as something that you ask your agents to do. Some behavior rules and yeah, like about the uh, age of agent rules, what I just described uh before. So I just set it right, these words into the sole md but someone will write those words in Agent md. I don't know but I'm not a tech person. I have to say that's what I'm doing now.

Speaker C: Yeah. And then the skills are uh, particular.

Speaker D: The skills is more like uh, yeah, how to like how to write a um, memo, how to have a better research. Um, because I remembered the first draft, they delivered a very bad memo, like a very messy, I have to say and then I ask them, okay, you have to like have a bullet point, you have a very real citation. And yeah, so I just list the rules and it follows. And uh, because after that I, I keep continue lettering, I continue educating it

Speaker A: how to perform better.

Speaker D: So I after like several rounds and it learned how to write a better memo. And I will say okay, now you can update the skill so you can use it uh, anytime.

Speaker C: Such a interesting iterative process where you're giving feedback basically in a way like it's not reinforcement learning but you are uh, figuring out ways to make the agents do better work and give them better instructions for those subtasks they're doing. Like, it's just an amazing new way to work that we're just trying to figure out in Git.

Speaker D: If you get a chance to look at my five step map for a native law firm, what I'm doing now is definitely the uh, layer three self learning data layer, I think. Yeah. Um, what most people are doing now are still like using AI as a chatbot. They cannot imagine like in an agentic way, ah, in an identical world than that you can see all those things will be changed by agents. So only if you try it, otherwise you cannot know how it goes.

Speaker B: So just to do a quick recap to make things clear for our uh, listeners, when it comes to agentic AI, it's technically a normal generative AI with a specific prompt to take some actions and access to various things. Like we can create an agent which is technically, for instance ChatGPT. And we explain how it's supposed to access our calendar and set up our meetings and send us notification. We can create another agent to do research and we say first you got to go, um, on Google Scholar and then here and there and then collect them. So it's just we provide them with a clear prompt and access to whatever we want them to access and that becomes an agent. And when it comes to some rules that we want to define for agents or restrictions, we can save them in files, but these are again very much technical programming. So yeah, in programming you can save files. We call it, let's say SOL md, which means the uh, the agents know where to go and get the rules. Whenever we run them, they look at the rules first and we say go and check SOL md. So that's how we refer the name. M could be anything else. So it's just a file name. And when it comes to skills, it's technically it is a perfect prompt to do a task. You can ask it to draft a contract and you can write this very long and clear prompt, that how it's supposed to do it step by step and you save it as a skill. So when I say save it, it means next time you don't need to write it again. We call this skill the drafting Contract agent or something like that. And some models automatically identify when you're asking, I want to draft a contract and this is the basic information you need for it. You, you don't need to say that, go and use that skill. The models know this person want to do this and I need this skill to complete the task. So they just go and pull out the file and read the whole prompt that you have written and execute for you. So Helen's work shows us language models. They are not limited to chatbots as uh, we usually use them. We can go way further. But it comes with also risks and understanding those risks is very important to help preventing disasters. As Mark said, we have had cases like the models deleting files. Like you say I want to clean my computer because things are not organized and the model thinks the best way to do it is to delete everything because that's the cleanest way I can make this computer. So sometimes there are these type of misunderstandings and in the last few months you've seen some cases like disaster like is actually happening. So it's not like a sci fi anymore. So having experience and yeah understanding the risk, I think the way Helen explained is extremely important.

Speaker C: So Helen, you mentioned your five level value stack and this is uh, another amazing insight that you've gotten from your experience with using generative AI and using agentic AI. Could you walk us through what are these levels from 1 to 5?

Speaker D: Yeah.

Speaker A: So that's pretty much a uh, theory about real value live in legal AI world and I think about in layers and I call it the legal AI value stack. We have five layers here. If you sketching it in a whiteboard you have five layers and there's an arrow points up. So in the button level, the level one, I call it the raw AI capability. So there's nothing unique left and it's just pretty much you use ChatGPT Cloud like raw AI capability. Everyone use it, right. Level two, I call it AI plus workflow AI ah, plus your like lawyers workflow. And you have a very pretty like UI UX and you have no um, templates in it. Some fantastic arms already saved in it to the level three, call it a proprietary data layer. Um there is some real defensibility because that's not something about your client's data, it's about uh, the data that how lawyers use your software like uh negotiation patterns, the preference about the clause that'll be a real defensibility but uh, it's still transitional and which means you can switch it to another very quickly without much cost. And so we come to the level four I call the system of record. I believe it's the strongest defensibility today. Example I would say like Clio and FireOne and PracticePanser so like some management platforms. So the switching cost is very high and lawyers and the law firms will operate on it every day. And I think levels one through level four are uh, all pre AGI strategies. They slow down the raise to the button but they don't stop it. So I think the only real mode is level 5. I call it a hybrid model or someone will call it like a, a native law firm, um AI, native legal team, whatever.

Speaker D: The principle is that stop selling software

Speaker A: to law firms, become one law firm. So because AI handles everything that will scale up but humans will always handle the trust issue, the accountability, the judgment. So uh, the scarce resource isn't intelligence, it's about maybe it's about accountability I believe because AI cannot take the liability when something goes wrong someone needs to be accountable. But now AI companies ship with very perfect disclaimers. But uh, lawyers can carry malpractice insurance software alone.

Speaker C: Can, there's a lot of ideas in there uh to help the listeners understand can completely. When we're talking about value, who's capturing that value? Is it the same each level of the stack or does is it different?

Speaker A: Uh I think when I talk about value it's more like how different legal AI companies can resist feel the uh, like competition with uh LLM companies like Claude chatgpt and how they can survive during this computation. And I think because um M most of these series are uh like developed after the anthropic ego plugin at that moment many people kind of panic uh about the thing because they say okay no SaaS at all because AI the LLM providers will take off the market. And I don't know if you remember at that day we have very you know big drop down in the US capital market. Most of the SaaS companies are influenced. So yeah at that moment I just posted the first draft of the theory and then got I think like M30K impressions on one post and I think okay, wow it's so many people feel impressed. So I just continue to iterate it and I like I have a full analysis on it and I just posted it in my substack it's a full analysis.

Speaker C: So how does that analysis relate to what a lawyer uh should be thinking?

Speaker A: Yeah so that pretty much the theory behind the open claw law and uh, uh yeah I have explained it in my inkling so um, I just flipped this theory so each layer can uh flipped into a uh specific step if just a regular law firm wants to transform into a native law firm. So the first step is based on basically a raw AI capability. So I think it's a table stake. Every law firm I believe most law firm are using AI every day. So for that uh, the second layer, the workflow. So I think the law firm should use AI agents to transform its workflows. And the transform doesn't mean you just use AI like what they are doing now they are still in the chat boss era. And now I think like I said uh, now agents can spawn more sub agents. You can do the workflow in a parallel ways and you can gather results of different issues simultaneously. And that's something will never happen before. So it's a different, totally different theory and you should to figure it out. Um, and that's just one example because I think one agent developed along the day and you will see I think it can do more magical works than about. Yeah. And so we come about the third layer about the appropriate data just to talk about the way I train my agents to generate or to update the skills by itself. I call it a self learning data layer. As uh, all of your like knowledge can be updated by AI and you just, you have to transform all your knowledge into something that AI can eat and then it can just digest it and get something new and the things will be recorded in the system and it become a cycle of your data. So that would be a ah, very big transformation and I think uh, it's a very key point which is different from what you see in the era. And so about the level 4 in the legal AI value stack is about uh, um, system record. And I think yeah when, when you flip into the step four my roadmap to AI law firm um, it becomes something I call it like you just you know make AI, make agent as your fundamental system. You do like your IT stuff, your tech stuff, your HR stuff, your marketing stuff. All the things you will rely on agents to help you like you have different and currently I have Morgan and I have Cleo. They are senior lawyer associate and junior associate. But I mean gradually I can have a team like M someone who are dedicated to security issues in the law firm and I can have some agents who can help me to deal with the immigration issues for my employees. So yeah, so basically all the things you can based on agents and finally we get into the level five. I think it's you cannot just uh go directly into a level five but you can go through from step one, two, three and uh, naturally you will getting into level five and it's also the step five to become a real

Speaker D: AI native law firm.

Speaker C: Yeah, it's mind blowing as you talk about it. No, no, it's It's a vision of some of this is the present and some of it's the future. From that law firm perspective, do you see the most efficient or the most likely path of the law firms to be helped along the way by some of these major legal AI players? Do you see some of them maybe shifting to build their own or is that just too difficult even for the biggest law firms to, to start trying to do a uh, massively technical and complex security sensitive?

Speaker D: Yeah, I think I will definitely are supporters of the middle sized law firm

Speaker A: and booking law firm. I think they will. Resources of the innovation, the ideas, all the innovative ideas may come from these small law firms because in the big law firm I think they have very strong IT teams already. They have, they have legrand, they have the most fantastic legal AI products. But the point is that I think they are still struggling with billable hours. If this model will still exist, I do not think there will be a big change for them to have something really transformative. But for the smaller firms, I have my own observation. Like many partners in the middle sized firms and boutique firms, they tell me they wanted more AI in their daily work because first is hard to keep billable hours, they have to have more flat fees projects and second, they really think maybe AI can help them to reduce the hiring of the new grads. I think, I don't know is it the same in big law firms? But that's what I hear recently.

Speaker B: Okay, so following what you said that you heard that some law firms are looking into not hiring junior lawyers because like AI can do it, why they should pay uh, human technically so it would save them a lot of money. What would be your advice for the lawyer students who will graduate in few years or even earlier? What skills do you think matter the most for them when it comes to this new emerging agentic AI world?

Speaker A: Nobody was prepared for AI agents because we don't know it. It's a brand new. You have to learn it. So there's no playbook, there's no degree for it. So I have no tech background, I can't write a single line of code. And yet here I am running a multi agent law firm experiment in public. So I think the real driver every student should have is preparation is a curiosity. Like I had a question. How is AI going to fundamentally change my work? And I couldn't sit still until I found out. So I started building and I kept learning and I think that's the point. So in the AI era, your past experience, what you learned in law school Might not even count. So pre Alm. Um, legal technology may m. Not work too. I think a lot of it doesn't apply anymore. Everything's new. So the only thing that really matter is are you willing to keep learning? Yeah, that pretty much. That's also actually why I do the open call experiment in public. Because I want people to see I don't have all the answers. I have a lot of things I don't understand. But I show up every week and uh, I learn something new.

Speaker D: So.

Speaker C: Yeah, and back to that idea of the, the anxiety or the panic, plus the curiosity. And I know that a lot of law students are very, you know, they often want to know the right answer. Uh, and part of what law professors are trying to get them to see is you've got to be happy and comfortable and curious in the uncertainty. And now we're just transferring that over into these new technical tools.

Speaker B: Yeah. Just to add something to that, I think, uh, law students should understand the world is changing very fast and the type of knowledge or skills we traditionally thought lawsuits will need or they're invaluable are not as valuable anymore anymore. And they should be very open minded about learning skills that are not directly associated with black letter law. It has nothing with case law or legislation. It is about understanding how you can use these new skills and tools to get the knowledge about the law. And the process is becoming way more important than what it was before traditionally. And yeah, I do have a student. Some of them go to law firm and even when it comes to the using their digital technologies, like the messaging app they use is not something that we use every day and they struggle to use them and they realize that, oh, okay. When working in a law firm is not all about law. So we have come to the end of this episode and we would like to thank Helen for being very generous to spend this time with us and uh, share her experience with us.

Speaker A: I really enjoy our conversation. Thanks for having me.

Speaker D: Goodbye.

Speaker B: Thank you. Bye.