The AI Toolbox for Lawyers, with Zach Blattner and Berkeley Almand-Hunter

Great lawyers don't always know

how to build great law firms.

Let's change that.

Join Ben Gideon as he shares hard won

lessons from building his own financially

successful law firm and practical

insights from top law firm entrepreneurs,

business consultants, and more.

This is a podcast for lawyers by lawyers.

Welcome to Elawvate: Build

and Grow Your Law Firm.

Produced and powered by LawPods.

Today's episode of the Elawvate:

Build and Grow Your Law Firm

podcast is brought to you by Supio.

I just spent an hour doing a webinar

with the Supio folks. It was great.

We had the head of business development

and also one of the lawyers from

Texas who's been an early adopter of

Supio and walked through some case

models of how you use Supio to

put a case together. It was very,

very impressive.

I would encourage everyone who doesn't

yet have an AI platform for their law

firm. If they're a plaintiff's

side, personal injury,

medical malpractice type of law firm,

check out Supio and get the demo.

Our podcast today is also

brought to you by VisionSpark.

VisionSpark is the company that

helps law firms and other businesses

find seconds in command. You know

any seconds in command, Jeff?

That would be me.

Now.

It's exactly where I want to be.

Yeah. When I think of number

two, I always think of Jeff.

Thank you, Ben. I appreciate that.

We have vision spark to credit for

the search process that led us to Jeff

and that Jeff has led us to the holy

grail of it appears to doubling our

revenues in less than a year

of your time at our firm.

So it's all paid off very well.

Anybody who needs a second in command,

like Jeff, I would encourage

you to talk to Vision Spark.

Our podcast is also brought

to you by 1% for Planet.

1% for Planet folks is the

organization that allows you to

contribute 1% of your gross

revenues to environmentally

conscious nonprofits.

Our firm has done that.

It's a great organization that's allowed

us to partner with some really terrific

nonprofits in Maine and

throughout Northern New England,

and we really would love to have

other firms join us in that.

And if you do join us in 1%,

please reach out to us and we'll

give you a plug on the show.

Hello, everyone, and welcome

to the Elevate Build and Grow,

your law firm podcast.

This is Jeff Wright,

chief operating officer at

Gideon Asen. And as always,

our host, owner, partner,

Ben Gideon. Hey, Ben.

How are you today?

Hey, Jeff. Seems like you've

already had a long morning.

It was a long night. I was here

for about 14 hours yesterday,

and then I was in at like 6:45 this

morning because we have a very busy day.

And technology has not been my friend

this morning, so I'm having a good time.

But what kept you here

for 14 hours yesterday?

Well, last night I wanted to work on our

focus group, which is coming tomorrow.

Gustavo will be making

his second appearance.

Gustavo for the listeners and

for Berkeley and Zach is Jeff's

lawyer alter ego. When he

plays a lawyer in focus groups,

he appears under the name Gustavo.

And last time he did a focus group,

he managed to win on behalf of the

defense against two seasoned trial

attorneys. So now he's going

up against the ... When's that?

Is that tomorrow night?

It's tomorrow night.

Yeah.

Yeah. So well, I'll be tuned in to

see how Gustavo performs this time.

I'm thinking I'm going to go

20. So I'm pretty excited.

I like our chances on

this one, but we'll see.

Yeah. All.

Right. Well, you want

to introduce the guests?

Yeah. We're very,

very excited today to be joined

by Zach and Berkeley from the

Ru Institute is something that is very

well known in our area up here in Maine

and their affiliation with Northeastern

University in Massachusetts,

but a lot of our listeners

aren't from the area,

so I think it would be good for Zach and

Berkeley to talk a little bit about the

Ru Institute and obviously what

your roles are there. So welcome.

Thanks so much, Jeff Bennett.

We're excited to be here. So yeah,

let me start by saying a little

bit about the Ru Institute,

and then Berkeley and I

can introduce ourselves.

The Ru is coming into its

sixth year, and like you said,

part of Northeastern gift

from David Rue made it happen,

who's a main billionaire, and wanted

to help drive the main economy.

And so the Ru Institute serves as an

opportunity engine for graduate education,

applied research, entrepreneurship,

and there's a number of startups

that are wanted out of here directly.

And then custom learning, which is

the work that Berkeley and I do,

non-degree courses for businesses

focused on AI analytics and

project management. So the Ruben

Institute is one of Northeastern.

Now it's at 14 or 15 campuses, Berkeley.

Do you remember how many we're

at across Canada and the US?

There's tons of them.

Something in that ballpark.

Yeah. I didn't know that about

Northeastern. When I joined that,

there were so many campuses,

but Northeastern has developed this very

entrepreneurial model where campuses

pop up in areas of need and

kind of business relationship.

There's a Northern Virginia one that

has lots of military and government

connections in Silicon

Valley, Vancouver, et cetera.

I happen to know there's one

in London because- There is.

My daughter's boyfriend is

a freshman at Northeastern,

and he's spending the year at the

Northeastern campus in London.

And my daughter actually just got back

from visiting. It was a great trip.

Oh, cool. Yeah. Yeah.

It's really changed as university

in the last bunch of years,

and it's continued to grow in this

way. To step back, introduce myself.

I come to this work from K-12

education. I was a teacher, a principal,

then a teacher of teachers

for a number of years.

I moved into virtual teaching pre- COVID.

So I learned a lot about how to drive

engagement and learning in the virtual

space that wasn't just kind of passive

and having people just sit and listen,

trying to think about ways

for ... This was for teachers,

but how to engage them in the learning

so that yes, they were virtual,

but they were actually walking away from

virtual sessions with real learning and

real outcomes in mind. And I

joined The Ru about two and a half,

three years ago when this custom learning

program that Berkeley and I now lead

up was kind of nascent and looking for

direction. And so I came on without any

real AI learning or experience. And

my boss, when she was hiring me, said,

"I don't know anything about AI.

I just know about learning and

teaching." And she said, "That's okay.

I have someone for you to meet and you

two are going to work together to run

this.

" And so that's by means of

having Berkeley introduced

herself and her side of

the thing,

I kind of bring the pedagogy and learning

and I'll let Berkeley share more about

herself.

Yeah, sure. So I originally

started as a mechanical engineer.

I was doing my PhD in

mechanical engineering and I

had some really hard sensor

calibrations to do.

So I started getting interested in

data science and how you could use

statistical modeling for

these sensor calibrations.

And then I ended up moving out to San

Francisco and becoming a data scientist.

I worked at Square, the payments

company, building underwriting models,

and then moved back to Maine when

I had my second kid to be close to

grandparents and was recruited to

teach in the graduate school at the Ru

Institute, which I really enjoy teaching,

but I also like to be

working with industry.

And so when I started teaching

some industry classes,

that was when I got really excited about

staying at the Ru long term because I

never saw myself being

a long-term academic.

And I was really excited when Zach

got brought on because like he said,

I have some graduate teaching experience,

but Zach's depth of teaching knowledge

has been great. And I think we're a

really good team because we

have complimentary skills.

So let's just start with

the very basics here.

So for lawyers and those

who own and run law

firms, I'm going to start with the primer.

What are the core things you think they

should be looking at and what do they

need to know now with the

development of particularly new AI

platforms and technologies?

Yeah. So I think the most important

things to know are everybody should

familiarize themselves with AI and

understand the risks and benefits of

these different tools, sort

of what they're best at.

I think there are a lot of

legal specific AI tools.

There are also some general purpose

tools that work well for attorneys,

but sort of understanding what

the best use cases for AI are.

AI is great for summarization,

it's great for research with the

caveat that it does hallucinate,

which means that it makes things up.

Sometimes if it doesn't know the answer,

we like to say that you have to think

of it as basically a really smart intern

and think about the fact

that with an intern,

they can bring you really good work.

But if you just give that work

to your client, if it's not good,

they're not the ones who

are going to get blamed,

you're going to get blamed.

And to top that off,

that really smart intern is a

little bit of a pathological liar.

So they'll tell you that they

know things when they don't.

So you got to keep that in mind and

remember that you are always sort of the

final check for it, but it can

do a lot of really great work.

And I thinking about how these models

work can also be helpful when you're using

them.

You guys have engaged with some lawyers

and law firms and helped to teach

certain principles. And can you

just tell us what in that role,

how you've approached

it, what you've done,

what takeaways do you have

from those experiences?

I think a big one as we started to

think about law firms and lawyers in

particular is that the learning

about AI is different than a lot

of the ways that lawyers

have learned traditionally.

And this is true across,

not just in law firms,

but because it's this new

skill that's out there,

everybody's having to learn it

on the job for the most part.

It's definitely not taught in

law school in the same way.

And so there's all these people in the

middle of their careers having to learn

how to integrate this.

So that's a challenge for everybody

across every business right now.

In firms in particular,

the kind of apprentice model of learning

and how that functions where people

sort of have a mentor and learn under

them and the lack of sort of formal

structure makes that even more complicated

because now how is this learning

going to be delivered? And like Berkeley

said, because of the risks involved,

because of the hallucinations, because

of being strategic with it is so crucial,

there's a challenge really with like 15

lawyers approaching it 15 different ways

within a firm and then teaching the

people under them 15 other ways and really

lacking coherence in strategy.

We try to bring a more

coherent framework to that.

And I think rather than just

focus on the tools themselves,

we believe that to understand

the tool and use it well,

you need to understand a little bit of

the concepts behind it, not a full PhD,

but like we don't just, and at

least in the learning that we do,

it's not just about the

product and the tool.

It's about what's going on behind the

scenes with this particular tool because

that helps you understand the risk of

hallucination, how to prompt it better,

which out of a few options can use it.

So those are all pieces we try to bring

into our learning to help make it more

outcomes focused and actually bring some

value to folks.

So I know AI's been around

for a little while now.

I wonder if you could tell me,

because we struggle because it seems

like every third party or every vendor or

every software that we have to engage

with is rolling out some semblance of an

AI platform to various levels of

success. Some of them are terrible.

Other ones are doing one

part of what we needed to do,

but not all the parts. So I'm curious,

are we just on the cusp

of beginning of seeing the

true value of AI or are

we at a midway point?

I feel like we're just

at the beginning of this,

but I'd be interested in what you think.

Yeah. I mean,

I think that's the million dollar

question and nobody knows for sure.

I think we're at a point though

where there's the massive hype cycle.

There's a lot of potential

use cases that are great.

There's also potential use

cases that aren't great.

And I think not everyone

has really learned how it

works best for their job and

their field. So I think

we're sort of in the middle.

I think some of the promises aren't going

to pan out, but I think you also have,

we've talked with a lot of firms that

say that they've bought all these tools,

no one's using them, no

one knows how to use them.

So I think as you get

those people trained up,

they're going to realize

a ton of productivity that

they haven't realized yet.

And I think in a few years,

we'll be at a point where we really

understand what it's good for,

what it's not good for, et cetera.

It's also interesting,

this is something Berkeley and I

have talked a lot about recently.

There's a lot of claims made in

all these tools and softwares.

I think right now it's agents, right?

This is Agentic, agents work with this.

And that's like a newer thing within

the last year and a half or so,

agents have become big.

But even within that,

there's a lot of gray and

things that claim to be agents,

but they actually aren't

operating independently.

Can you explain what an

agent is for the uninitiated?

Yeah. I'll let Berkeley give

us breakdown for this group.

Yeah. So I mean, the word agent

is thrown around pretty loosely.

I think I don't know that there's really

an agreement on a true definition,

but I think what I would say the

definition of a true agent is,

it's a large language model that has the

ability to make autonomous decisions.

So you can give it different tools.

Most of the typical chatbots

can behave in this way now.

So say you're on ChatGPT

and you ask it a question.

If you ask it a complicated enough

question and put it on research mode or

something,

then it will actually make a research

plan as to how it's going to figure out.

And it's going to step-by-step say,

"Okay, first I'm going to look at this,

then I'm going to look at this,

then I'm going to look at this,

and then it can access different tools,

whether that's the internet or

APIs or whatever." So I think

the true definition of an agent is that

sort of autonomous decision making. Now,

for every use case,

you don't necessarily want or need

that autonomous decision making.

And if you have something where you have

a workflow that you want to automate,

say you get an email from someone,

you want to summarize that email,

and this isn't a great use case,

but summarize that email and

tell you about it in Teams and

tell you how urgent it is that you don't

necessarily need an agent for because

it's the same steps every time,

but that could be like a workflow or

some people might call that an agent.

I think a lot of the tools

are calling that an agent.

So I think finding the

solution that works the best,

if it's something where there's

really well defined steps,

you don't necessarily want to give it

that autonomy to make its own decisions,

but there are use cases where you

do need to give it that autonomy.

So stepping back, I mean, to Jeff's point,

there's such a proliferation of different

platforms and products right now.

It's kind of mind numbing

and hard to navigate.

If you were approached by a

small, mid-sized law firm,

they don't currently have any AI

platforms that they're using on an

enterprise level.

Maybe some of the lawyers individually

do things at home in their personal life,

but as a business, they haven't

adopted any approaches yet.

What would you recommend

that they start with?

What are the kind of core platforms

that you're thinking would add

most value and how would you recommend

starting to roll out that process?

So I think it depends what

kind of attorneys they are

and what they spend most

of their time doing,

whether that's research or

drafting things or litigating.

There are definitely legal

tools like co-counsel,

and I think LexisNexis makes a version

that are research-based and look at legal

cases. There are also more general

purpose legal tools like Harvey.

And then also you can just look at the

AI tools that are general purpose for the

public like ChatGPT, Claude, Copilot.

And we always tell businesses when

they're thinking about using AI to think

about what you spend your time doing.

What are your business goals and what

are the things that you either want to

make faster or better and then

base your decisions on that?

And I think that's really crucial because

there is this hype and pressure for

people to sign up and buy it.

And these tools are expensive.

So what is the goal of

bringing on AI? And yes,

it's to be competitive with other firms,

but that can't be the only reason.

And this is all intention depending on

how firms bill because those that have

billable hour constraints versus ones

that are operating in a different model,

I assume like you guys are, have

different tensions there with like,

what does it mean to save time and is

that something that is being passed on to

clients or not? That kind

of a major impact as well.

I mean, it seems to me that

every business probably,

and certainly every law firm will

need to have some large language model

platform, whether that's Claude, ChatGPT,

Gemini or whatever, Perplexity. I

currently subscribe to all of them.

And so again, putting it back to you,

it's not practical or economical

to subscribe to all of them.

So which one should someone subscribe to

if they don't know better and they need

to have access to an enterprise level,

large language model

platform? Because again,

I think every business

will need that. I mean,

there's not really an argument

against that at this phase,

even in this early phase

of AI progression, right?

It's an absolutely essential tool

that every business will need.

Ben, I'd say yes, and this sort of

brings us to what we do, right? Yes,

with Training, with proper training

and guidance around it. Personally,

I really like Claude. I think it does a

number of things well in the workplace.

I haven't even started to

play with its cowork function.

We're trying to still get access to

at Northeastern in our enterprise

subscription,

but I've heard from other folks who use

the cowork function that it's great,

but even without that.

And cowork is the concept that it can

automatically access all of your other

platform documents for your ... So I mean,

we're on Microsoft SharePoint or you

have Dropbox, it can integrate with that.

So you don't have to upload

documents as you use them.

It will have access to that automatically.

Berkeley and I use the Claude

projects a ton. They're really useful.

I could see them in law

being useful for case.

And what is a Claude project?

A Claude project is its own

mini RAG system that allows you

to upload sources of knowledge and give

specific constraints and instructions

around it so that your chats within that

project or questions you ask are all

referencing just a specific kind of zoomed

in set of files and knowledge so that

you can really target the discussion

and it kind of persists over time rather

than one-off chats, which

allow for any sort of-.

Is that kind of Claude's version of

Notebook LLM? Yeah. I mean, I have to say,

having spent a lot of time

with ChatGPT and Claude,

I'm now growing to prefer Claude to

ChatGPT. I find it's less conversational,

but it's more business-like and

less seemingly susceptible to

influence in terms of wanting to please.

Yeah.

And I think talking about your

preference for not having it just

agree with everything you say,

there are some things you can do within

each of these tools to customize them to

how you want. So they call

them different things,

but they each have instructions that you

can add in your user preferences where

you give it overarching

things that you want it to do.

So you can tell it like

your job. And I tell mine,

"Don't be overly agreeable. Only

agree with me if I'm correct.

Tell me if you're not sure about an

answer. Always give me references.

Don't use a bunch of business jargon.

Just be concise." And I actually

have mine set up. I say,

"Don't answer in more than a hundred

words unless I explicitly ask you to.

" That's one of my AI tips is if

I'm brainstorming or looking for

ideas,

I often want it to give me very

short ideas and then I ask it to

elaborate on the ones that I like.

So I'm not waiting for a long time and

wasting a bunch of my time and

energy. And when I say energy,

I mean the compute energy that drives

the cost up and causes environmental

impacts. Having it respond in five

pages when I just want to say,

"Give me five activity ideas for this

class." Four of them are probably going to

be bad. So I just say, "Give me

five," and then I move on. Sorry,

that was a little bit of a tangent, but-.

I actually think that's important because

a lot of listeners here have probably

taken that first step of using

a large open model or chat or

Claude,

but I think they're using it

like a Google search and it

sounds basic,

but if they're not setting their

parameters like you talked about correctly

or prompting it in a way

to get the most out of it,

it's really not going to work for

them the way it's designed to.

And I think that's where you

guys kind of come in to ...

Anyone can sign up for chat, but

in order to get its best use case,

it has to be done appropriately.

And my guess is the majority of

people are not using it correctly.

Have you seen that?

Yeah, I think that can be true.

And we have several courses

on this at different levels.

And our most sort of beginner

course, we started several years ago.

And to be honest,

we thought it was going to sort of age

out very quickly because everyone was

going to figure this out,

but people have other stuff to do and

not everybody is interested in this.

So not everybody has really

taken the time to learn about it.

And we found that it continues

to be a really popular class.

Sometimes you see these articles

saying prompt engineering is dead,

the models are too smart now, but

no matter how smart the model is,

it can't read your mind. So it can't

know who you're doing something for,

what your role is,

what your preferences are and the style

that you want your answer in, et cetera.

So you still have to give

it that information. And

if you're new to prompting,

one thing you can do is ask it, like

say, "This is what I want to do.

Write a good prompt for

me, or what questions,

what do you need to know more about to

answer this question better?" And it will

actually ask you questions

and give you examples.

There's a great book that I read

called The AI Driven Leader.

I don't know if you've seen or read that,

but it has a lot of useful

guidance in terms of how to best

use a large language model, whether

it's not a platform specific,

but they advocate something

called the CRIT model.

Have you heard of that where you define

the context, you define the role,

you then have the AI inquire

so that it's actually asking

you a series of questions to better

refine what you're looking for,

and then you clearly define the

task. And I've found using that,

it just frames it in a way that

it's a much more robust and

disciplined set of prompts that

achieve a much better outcome

generally.

We have a similar kind of resource

that has all those same topics.

Going back to this idea of projects,

one of the things you can put in the

instructions of a given project is ask me

five questions to clarify where

I'm going. And then that way,

that's embedded in the project.

And every time you start a conversation

in that given project, it's going to,

whatever you prompt, it's going

to ask you questions about it.

So cut down on having to do that in

every individual prompt and instead drive

efficiency and make sure that

you don't forget to do that.

And it's always pushing you

and building those pieces in.

So you could even upload into the project

files like a structure that you want

your prompts to go and force it to push

you if you don't properly prompt it and

to equip the project with that knowledge

and understanding in particular.

And you can even define exactly which

questions you want it to ask you.

So we use it a lot for like, we'll

have a template for something,

like a one-page summary for a

course that we do. And we'll say,

"Before you make this, ask the user,

what are the learning objectives?

Who's the audience?" Whatever.

So then you give it all the information

that it needs to give you a first draft

of that thing.

Yeah, it's really quite remarkable

what you can ask it to do. I mean,

Jeff and I did an hour long

live podcast covering 10 fairly

complex business topics and

just to prepare for that

and make sure we could get

things in and the timeframe

I role played it with,

I think it was ChatGPT and it said,

"We're going to do this hour long."

It knew what the topics were. It said,

"You play Jeff and we're going to

do the podcast." I played around

with that for five or 10 minutes,

but I can't say it replaced

Jeff in his role, but-.

You can say it.

It's fine. But you can do, I'm

cross-examining this witness.

You play the role of the witness, be

aggressive, be defensive and difficult.

There's lots of roles, but I think a

lot of times people don't think about,

well, you're asking it a question or

putting in a prompt, but what is its role?

Because if its role is

the judge or the jury,

that's very different from if it's

opposing counsel or if it's just a neutral

arbiter who's trying to,

or a consultant who's giving you advice

or a friendly person that's aligned with

your interest.

So it's going to be very different

depending on what role you give it,

just as any person would be if

you were interacting with them.

You mentioned at the beginning Jeff's

role is Gustavo in his focus groups. Well,

I just was reading about a new AI

firm called Simile that does these

digital twins and creates kind

of archetype AI avatars across

different ...

It's four focus groups that all these

companies are already using to test

against different regional and ethnic

backgrounds and all kinds of different

pieces, age, gender.

What's it called? This is interesting.

It's called Simili.

Simile. Okay. Oh.

Boy. We're going to sign

up for something else now.

Yeah.

And I assume it's going to change a lot

of the way focus groups work because of

that, because at least it purports to

be able to allow tons of questions and

permutations of those questions across

different groups to learn and explore

what a given person would say.

And this simulation stuff

around deposition prep,

that stuff too is all coming online with

more realistic simulation so people can

practice what it will sound

like in advance, get feedback.

There's a bunch of

different tools out there.

One called AltaClaro has

something called DepoSim,

and I think it's like a new

launch product of theirs,

but that does this live deposition

experience that ... I mean,

I haven't tested it,

but I think more and more of that stuff

will come online and be higher and

higher quality across fields in law

and also for teachers and doctors,

anybody who has that sort of performative

aspect of their job and needs to do

something in front of

a customer or a client.

Yeah. I mean, building on

that role play thing, I mean,

a friend who's an attorney I was talking

to was saying that he'll set up role

play to poke holes in his arguments and

he'll chat with it while he's driving to

court so he can practice his arguments.

And I think even beyond

an official role play,

asking it to poke holes in any of your

arguments or asking it whether something

you've written is correct is a

great way to use it. And I think,

especially attorneys maybe who like to

think about different sides of arguments,

it's a really cool use case.

And when I'm learning new stuff,

I use it that way a lot.

I'll say, "I wrote this,

is this correct?" And I don't always

trust what it tells me, but I'll just say,

give me a reference and it really helps

me make what I'm writing better. I think

that's one of the best ways to use it.

Yeah, that's so critical.

And I completely agree because it can

be really hard when you're invested in

something or your own work to honestly

critique it, but AI doesn't have

that problem with the feelings.

It'll give you honest feedback if you

ask it to do that in a way that may be

hard to get from others.

And I find that always leads

to improvements. In fact,

I think the concept of stress testing

anything using AI is going to become

mandatory. I mean, at our firm, we've

already said when we work up a case,

you're required to

stress test that with AI.

You're required to ask the AI to play

the role of the defense lawyer or the

judge and say,

"What are all the challenges and problems

that you're seeing with this case and

how do I address those?" Because

why wouldn't you do that,

right?

If you have that at your disposal,

you're going to learn it later when

they actually do attack your case.

So why not learn it now and preempt that?

Need help on a complex personal

injury or medical malpractice case?

Gideon Asen accepts case referrals

and regularly co-counsels with lawyers

nationwide on high value claims.

The firm has recovered millions of dollars

in cases that competitors turned away

because they dig deeper.

Ready to learn more?

Email begideon@gideonasenlaw.com

to start the conversation.

Gideon Asen shares fees as

permitted by the laws of each state.

Don't let complex cases

overwhelm your practice,

elevate justice together with Gideon Asen.

So moving on from the large language

models to the kind of more industry

specific things,

this is where I think there's such a

proliferation of platforms. I mean,

you just mentioned a

couple I'd never heard of.

We've experimented as just said with

a couple. How do law firms ... I mean,

I know Harvey's a big one that said

billions of dollars of investment capital.

We actually haven't used that one.

I'm curious if you can speak to that,

but we have tried

co-counsel, didn't like it.

We're using a different

platform now called Paramble

that's specific for medical

malpractice. We love that.

We're familiar with SupiO.

I've heard really good things

about that, but there's so many.

So how does one start to

navigate that universe?

I think your points about trying

them and liking them or not liking

them are good.

I think starting with what your actual

use case is and finding the one that

matches the type of work that you do and

then testing a bunch of them and come

up with a use case and say like, "What

is it that I'm trying to accomplish here?

Is it making me faster?

Is it making me better?

Is it poking holes in my arguments?

What's the goal?" And then test a bunch

of different platforms and see which one

helps you achieve your goals the best.

Do you think we're ever

going to get to ...

One of the things we're struggling with

is we have to use multiple platforms to

get what we want. So in an

ideal world, as this evolves,

we're going We would like to

have, I guess, one stop shopping,

for lack of a better term. Right

now we're using the open models,

we're using Chat or Claude, but

then we have medical malpractice.

We also have an AI that's

within our CRM system,

and then we're using a

separate AI for intakes.

Do you think we'll ever get to a point

where there's a one-stop shopping that's

going to be able to do everything

within one simple model?

Probably not as well as all of

the individual things do it.

So I think that every product is going

to incorporate AI and try to get you to

pay extra for it. At some point,

we might get to a point where that's

just part of the product for each

individual product you use.

But I think what makes some of these

really specialized tools good is that they

spend a lot of time fine tuning

the model and getting it to work

really well for that specific use case.

So maybe for law,

they would come up with one that's good

for all the different legal things,

but it does seem like probably smaller

firms that fine-tune for different

specific use cases will win,

unless they get bought by a bigger

one and they incorporate all of that.

But I think you'll still have something

legal specific probably. And then you'll

probably want a ChatGPT

or a Claude is my guess.

Ben, in terms of your question, well, one

other one that I've heard folks using,

it's hard to get the trial

for these. So for us,

I can only look on the websites and try

to explore the trainings that they have,

but LaGoura is another. I've heard good

things about it and seems like it has,

just based on what I look,

I think it's similar to Harvey or any of

these that allows some basic prompting.

It can do tabular review of pulling in

all the documents on a case and sort of

allowing analysis on it. It can do

track changes in Word and help to draft.

And which I think a number of them

kind of purport to do the same things.

I think in terms of how

to select, of course,

I think Berkeley's right thinking

about the specific use case,

what's highest priority. But I also

think this is where in law in particular,

there's a challenge because firms

are run by and large by lawyers,

attorneys,

different from other businesses.

They haven't necessarily incorporated and

privileged technical people to help

with this kind of decision making.

And so who is that person at a firm

who has the authority and vision and

alignment with leadership, but also

the technical skillset? Now it's AI,

but it could have been something else

in previous decades to help make some of

these decisions.

And so this is where I think the role of

paralegals could very well change to go

from people who are highly skilled

in supporting lawyers around case law

and work into this new brand of

paralegal potentially becomes more of a

technical expert along with those pieces

and is able to understand the use cases

and make recommendations or make

decisions or demo or pilot pieces to help

move the firm forward.

And where it's actually something that's

kind of cool that we do it at part of

Northeastern is this co-op experience

where grads and undergrads can go and do a

co-op,

but it's like the perfect role for

someone with that who's data science or AI

graduate student.

And you can embed them for a little bit.

Tell them your challenges, have them look.

We're trying to work with a local firm

to do this, grad student in there,

have them just observe what the lawyers

are doing and make some recommendations

about what they should do because it's

not realistic to think that someone's

going to go to law school, be a strong,

experienced lawyer, lead a firm,

and be an expert on workflows and

AI.That's a lot to put on anybody

from a decision-making standpoint.

There's so many things that come out

of what you just said, Zach. I mean,

so one of the things that I think

about a lot is when Berkeley,

you use the term use

case a lot, I've noticed.

And when I think of use cases in a law

firm, they're really two different,

very distinct areas of that.

One is in the practice of law

itself and the legal services,

that tends to be a harder way to

use it because there is judgment,

individual, attorney discretion.

There's certainly client

confidentiality concerns.

You worry about hallucinations, obviously,

and problems that AI can create for you.

But then there's the whole host of other

things that you're required to do in a

law firm that fall really more under

administrative, answering the phones,

triaging new cases, logging them into

the system. There's a whole host,

and that's probably stuff that's common

to every business where the use case is

really just administrative efficiency,

where it feels like right now,

if AI can do it better than humans,

that is a win and it removes a lot

of the tasks for people that are

boring, that lead to high turnover,

high burnout. So I'd be very interested

in developing that within the

model better.

And I think most businesses haven't

really begun to do that much yet,

but that's the easy part. And

then the harder part is, well,

how would AI be used to deliver

legal services, drafting contracts,

drafting briefs with litigation,

triage, I mean, evaluating cases,

figuring out arguments, stuff like that.

That's a harder quote unquote use

case. And then to your point, Zach,

about the internal structures, I

mean, I've thought a lot about,

in the old world,

you'd have a chief technology officer

or an IT person in your office.

Don't really need that so much anymore

because everything's cloud-based and our

IT firm is located somewhere else

and they can manage all that.

They don't need to come here and tweak

our servers or we don't have servers,

stuff like that.

But it does seem like we would

benefit enormously from an in- house

chief AI manager who could look

at all of the systems and figure

out best cases and different

platforms and then work on

training. Are you seeing businesses,

whether it's law firms or not,

move to that kind of a model? And if

so, how are businesses structuring that?

Sure. Yeah. I mean, I

think at big companies,

you're definitely seeing

a chief AI officer.

I think everyone is just trying to

figure this out right now though,

because you have your ... In a bigger

company that has a bigger technical team,

you have your technical people, your

data scientists, your engineers,

and they might be working on

the more complicated AI stuff,

but then you also have the

whole rest of the company.

And a lot of these people have turned

on these AI tools and they're like, "Oh,

great. I have Copilot.

I can build agents now." But what

they're not thinking about is that just

because it's low code doesn't

necessarily mean it's low complexity.

So if you have an agent

that's running a workflow,

how much compute is it using and how

much is the company getting charged for

that computing power?

What's the security? Have you just

given information access to a bunch of

people? How much is it hallucinating?

If you're making a chatbot for somebody,

how reliable is it and how

reliable does it need to be?

Risk with hallucinations really depends

on the use case. And with attorneys,

we've talked to some of them where

with in- house counsel, they say,

"Sometimes we're less risky. We just

want to less worried about risk.

We just want to innovate.

We want to push fast." Whereas if

you're counsel for an external company,

obviously then not making

mistakes can be paramount.

Yeah. And I mean,

part of the problem is law firm owners

that if everybody's just doing it

freelance their own way,

you have eight lawyers in your firm all

doing something different and you don't

know what they're doing.

What we really want is to systematize

it in a way such that there's a minimum

standard that everyone's following and

there are some outside parameters that

people aren't going beyond.

But developing that requires understanding

the capacities and limits of the

different systems and then developing

that protocol and then training people on

it, all of which most people

don't have time to do.

And they haven't even gotten to step

one of that of understanding all of the

capabilities. We're just sort

of playing around with it now.

And without identifying it,

a single person with authority within

an organization to have to be able to do

that seems hard to get that done.

It is. And I think it also comes down,

and this is true in maybe law firms

even more so than other places,

but I think we've seen across the board

finding time to invest in learning,

whether it's about AI or effective

project management or matter management in

firms. And the time spent to

train people on that is time.

If you actually want people to learn

something new or introduce a framework,

it's not just a 15-minute

standup meeting or in our minds,

like an asynchronous video that

you say, watch this on your own,

you've got to actually invest the

time and bring people together,

explain the rationale, teach them the

ideas, have them practice it, discuss it,

give them feedback. Those things all have

to happen for learning to take place,

but that takes time to do. So

there is always a tension there.

Is it worth investing in?

I think Berkeley and I would say it is.

Spend for your line of business

people spending two hours on

a training like we have on how to prompt

effectively and understand the risks of

AI will result in far better work

outcomes beyond the two hours you

spent investing on this training. But

on the surface, it could feel like, oh,

we're putting 40 people

through two hours of training.

That's a ton of time that we're losing.

And so it depends on the vision of

your where they want to put that bet.

But when people use these tools

incorrectly, I'm sure you've all seen it,

but we've seen these emails or chats

that are just somebody obviously just is

pulling someone from AI, putting

it right in. And it's like,

we're just making everybody

dumber by doing this. It's like,

I don't want to read someone else's AI

response to my question. We could just

ask AI ourselves if I

was trying to do that.

It makes them look bad too when everyone

else knows it was AI and not them.

It's like, well,

what role are you serving at this

company if you're just asking AI? I.

Just have my agent read

those and get back to them.

I don't personally look at it.

Exactly. Exactly.

I mean, before we wrap up,

if somebody was to hire you

guys to come do trainings,

what's the protocol?

You guys said you have an intro training

and then do you have more advanced

trainings and what are those involved?

Yeah, we have trainings

across three groups.

We have technical ones that go

deep, teaching people Python,

extended machine learning concepts

for attorneys in most firms,

those probably aren't that relevant.

But the other two that we

offer are leadership ones,

helping understand what's going on in AI

because we believe that having, again,

understanding the concept is

crucial to making decisions.

And then those trainings are focused

on strategic enterprise-wide decisions,

right? What is the problem

you're trying to solve?

How are you going to know if

this investment was successful?

And then thinking about what are the

next steps? How do you implement?

Who are the people you need on board?

Sort of all just technology change

management thinking with respect to AI.

So that's one piece of

training that we have.

And the other is more like applied line

of business stuff where it's like the

basics of prompting. Berkeley and I

just launched a more advanced prompting

class where we talk about personalization,

how to use notebooks or projects and to

do those more advanced pieces as well

as a matter management

class. Because again,

that's something we've heard

from attorneys is often

passed down in this kind of

apprentice style. But

for a firm to be like,

"Here's how we approach matter management

in a strategic way and that this is

how we want all of our staff to do it

so that it's aligned so that people are

more interchangeable and that we can

actually have a system that withstands

personnel change." So the other thing

we do in our stuff that I think is

different from asynchronous

learning is we customize everything.

So we meet with people before we teach

anything and talk about what their use

cases are and talk about their tools and

make sure that what we're talking about

in the learning directly relates to what

their people are doing rather than kind

of a generic training that

theoretically covers the stuff,

which I think makes a big difference again

when it comes to takeaways for folks.

Yeah.

And we do try to prioritize the things

that each role needs to know back

to the idea of the more expert AI person

or people within the company. I mean,

everybody should learn the basics of how

to prompt well and how to be safe about

it,

maybe how to use a legal specific

tool that's relevant for your job and

maybe go up to the level of

like a cloud project ChatGPT

notebook level,

but not everybody needs to know how

to get into Copilot Studio and build a

complicated agent.

And you have to be concerned about all

the things that I mentioned before.

So then you can identify a few people

in your firm who are going to be,

or one person who's going to

be the person who does that.

And we have more in-

depth training for that.

But we understand that attorneys are

busy, you already have a skillset,

you're not going to spend all of

your time on AI. So what is it that's

absolutely essential for you to know?

How are the classes delivered?

Are you doing these online?

Do you come to the firms?

Most of our classes

are online synchronous.

We really believe in these synchronous

courses because our courses have a lot of

interaction and we come up with practice

scenarios that are relevant to your

work,

which helps people sort of connect the

learning with what they do so that they

use it going forward. We

can do things in person,

but usually people don't want that because

their teams are usually spread out.

And how do folks get in touch with you

if they're interested in getting some

training?

They can email us or

check out our website.

Can we put that stuff into the show

notes? Do you guys have show notes?

Sure. Send it over. We'll

put it in the show notes.

And they find us at the Ru Custom

Learning website. But yeah,

we would love to talk more,

follow us on LinkedIn.

Berkeley speaking at a sailing

conference today in Newport.

So if this gets up soon, you can go.

She's a very in- demand speaker for

her ability to make AI translatable,

I think, to random other, to sailors.

So sailing is not my expertise.

What happened is I got invited to

talk about AI at this main outdoor

conference and somebody from the

sailing organization saw me speak and

said, "You have to come

talk at our conference in

Newport." So that's where I am

today. That's why I'm in a hotel.

Oh, well, Newport's lovely.

I love.

Newport. My in- laws live in Newport.

Oh, nice. This is my first trip to-.

That's where I got

married there. Yeah. Well,

it's so great to have you guys and we've

enjoyed our relationship with the Ru

Institute and thanks for joining

us and educating folks on AI.

I think this is probably the first of

many programs we'll need to have on this

as this continues to

evolve and change rapidly.

Yeah, it was great to come on.

Thanks for having us and let us know

how we can help you all and we're still

learning across industries

what to do here.

And the law industry in particular has

been really interesting to us because of

some of the stuff we've talked about

where we think a spot for training and

support and because the deluge of just

stuff that is being thrown on everyone's

face, which it is overwhelming

and hard to deal with.

And so we're kind of like a impartial as

much as it one can be in that space to

offer some guidance

outside of all the hype.

Appreciate it. Enjoy Newport Berkeley.

Thank you. Have a good day.

Thanks for listening to Elawvate:

Build and Grow Your Law Firm.

Share with colleagues if you

found it valuable. Remember,

building a successful law firm takes

discipline, strategy, and determination,

but you're not alone. Produced

and powered by LawPods.