337: Will AI automate legal work? The future of lawyers with Scott Stevenson, Spellbook CEO.
38:45 Share

distraught at the reality of the job when they graduated and just the amount of drudgery. But now, you know, what we hear from our customers are like, I love practicing law again. I wouldn't practice law without AI, but now I feel I'm getting so much more done. I enjoy my job so much more. So I actually think it's a great time to become a lawyer.

Good morning, good afternoon, or good evening, depending on where you're listening. Welcome to AI and the Future of Work. I'm your host, Dan Turchin, CEO of PeopleRain, the AI platform for IT and HR employee service. As you probably know by now, we started a newsletter. It's great. We share information.

fun facts that don't always make the regular show and some clips with our guests that we think you'll enjoy. Go ahead and subscribe to that newsletter in the link in the show notes. If you like what we do, of course, please tell a friend.

And give us a like and a rating on Apple Podcasts, Spotify, or wherever you listen. If you leave a comment, I may share it in an upcoming episode like this one from Gloria in Mexico City. Wonderful city. Who's in marketing for a software company and listens while walking the dog. Gloria's favorite episode is that great one with Dimitri Shapiro, CEO of Mind Studio, about how he built the AI app platform. It's growing faster than ChatTPT. We will link to that one in the show notes.

We learned from AI thought leaders weekly. Of course, the added bonus, you get one AI fun fact. Here's this week's fun fact. Hacker Noon published an article with the clickbaity title, AI replacing humans with sci-fi. Now it can do 43% of modern jobs, according to Anthropic.

The article is broken into three sections. Which jobs are most at risk? Why is the Turing test a trap? Which loosely says we shouldn't strive to have AI pass the Turing test. And the third section, how do we escape the Turing trap?

The article concludes with the thought-provoking question, is work in fact necessary? And if AI can truly replace human labor on a large scale in the future, do we need to rethink social structure? Whether you agree or disagree with the thesis, it's an article worth reading. We will link to it in the show notes. My commentary, let's be clear about what it means to replace human jobs. And let's stop hyperventilating about why that should even be a way to measure progress.

AI trained on massive corpuses of data can seem human and of course do human tasks by pattern matching what smart humans have done in the past. AI only knows what humans know and what humans know and how we experience the world changes continuously. AI is an incredible partner that is perfectly designed to assist humans.

The very nature of what it means to exist in a civil society requires humans with novel ideas, ambitions, and passions to define how we engage AI, not the other way around. Let's stop reducing this amazing technology,

inappropriately called artificial intelligence, down to how many jobs it can automate. Of course, we will link to that full article. Now shifting to this week's great conversation, Scott Stevenson is the outspoken CEO of Spellbook, a pioneer in the field of legal tech. The Spellbook AI co-pilot for attorneys is used by more than 1,700 legal teams to review more than a million contracts per year.

Scott and his team started spell book in 2018 and most recently raised the $20 million series a, from an incredible set of investors, including an Ovia capital Thompson Reuters ventures where great former guest Tamara Steffens is a managing director.

The Legal Tech Fund, Bling Capital, Moxie Ventures, Concrete Ventures, Path Ventures, N49P, and Good News Ventures. Full disclosure, Moxie Ventures, led by the incomparable Katie Stanton and Alex Redder, are also investors in PeopleRain. Thanks to both of them for prep for today's episode.

Scott has an awesomely unique background. Prior to founding Spellbook, he founded MUNE, a pioneer in electronic music, and received his BS in computer engineering from Memorial University in Newfoundland and Labrador in the great

proud independent country of Canada. Without further ado, Scott, it's my pleasure to welcome you to AI and the Future of Work. Let's kick things off by having you share a bit more about your background and how you got into the space. Thanks for having me, Dan. I'm really looking forward to it.

Yeah, I'm co-founder and CEO here at Spellbook. I come from a product and engineering background. Yeah, my journey into legal tech came from the fact that I had a small business building musical instruments. And then one day I got a legal bill that took almost half the cash out of our bank account. So that really frustrated me. And I thought as a software developer a lot about how we might be able to make legal services more efficient.

Although I have a co-founder, Daniel, who is a lawyer and came at it from the other side of going to law school, practicing law, being very frustrated at the amount of drudgery and minutiae he was dealing with on a day-to-day basis when he wanted to be doing more kind of meaningful legal work. So that's how we ended up starting Spellbook and launched it as the first generative AI co-pilot for lawyers in the world back in 2022.

There's so many problems that AI in legal can solve. Talk us through how you, I mean, you got a big bill, you run a different company. Makes sense that you try to automate some of that and reduce a small company's bill like what you're facing. I just pick a problem to solve. I know you're now kind of using AI to automate legal review of, I've shared in the open, more than a million contracts a year. Why that problem?

Yeah, I think legal work can broadly be separated into litigation and transactional work. So litigation would be like courtroom cases and things like that. Transactional is based around commercial agreements, agreements between people, sales agreements, real estate agreements and leases.

And even going up to mergers and acquisitions and much larger multi-document transactions. So yeah, we focused on that space because it was the space I knew the best and the one that I was frustrated with as a business owner. And I think contracts at the time, GPT-2 and early GPT-3 were around. And contracts were a really

good place for AI to add value because there's so much public contract information available online that can be ingested by these models, which enables them to be pretty good at manipulating contracts and understanding contracts. Whereas if you look at litigation and if you see any of these high-profile cases about AI hallucination, it's usually on the litigation side because the litigation side is...

There's a huge amount of case law out there, but you have to be really factually correct. You can't just be directionally correct when you reference it. And I think it took a lot longer for AI to kind of catch up to that side of law. So that's why we focused on contracts. I shared that you have an awesomely unique background.

I've watched videos of you playing the Mune, an interesting musical instrument to produce electronic music that you created. So now talking to Scott the human. So that was clearly a passion. You're a musician and it was quite an accomplishment that you created that device and brought it to market.

Talk us through the, call it like the psychological and emotional pivot of doing something that, look, I'm going to say is maybe a little bit mundane, automating the drudgery of a lawyer's job. Yeah. Yeah, it's a big pivot in my career for sure. But

But maybe not so much. I think I always wanted to make things. So I always considered myself like a maker or creative first. Even before I considered myself an entrepreneur, I consider myself a creative. So growing up, I really wanted to make video games at first. So that was how I learned to program because I just really wanted to make games and then

I got a little more into music and music production as I got older. Built the Mune instrument with a team of great people, which was a lot of fun. But what I found I was looking for eventually was like, how can the creations I have have a real deep impact? And I

I realized that whatever I'm going to do, whether it's making a small business that produces this electronic music instrument or I'm starting a coffee shop, all of these things are actually really hard. It's really hard to run and build any kind of business. And the reality of that kind of set in. And I was like, okay, if I'm going to put in this amount of effort into anything, our chief strategy officer, Chris, has this saying. He says, it doesn't actually cost more to think bigger.

Any kind of business is kind of excruciatingly difficult to run. So if we're going to run a business, if I'm going to be dedicating my life to this, I really wanted to find a problem that was impactful enough to solve. I would feel comfortable dedicating maybe 10 or 20 years of my life to solving that problem. So that's what really caught my eye with the legal issue. And I think when I looked at it, I was like,

I think a lot of people have that experience of getting their first legal bill and being like, holy cow, it doesn't feel like any other bill that I get. It's probably one of the scariest bills you can receive for the first time. And it just felt like it was an industry that hadn't managed to capture this sort of efficiency that most other industries had during the computer and internet revolution. And so

I liked that there was sort of this big problem that no one had really figured out yet. The legal fees still seem really high to the average person or average business. And I love that it was sort of this uncracked nut. One of the things I learned is you want to look for spiky opportunities because it was easy to solve. Someone solved it already 100 times. And so I liked that it seemed hard to solve.

And it seemed like if we could make all the legal transactions around the world go, I thought maybe we could make them all go 90% faster, more efficiently, then that would actually have a really big impact on the GDP of the globe and just create more wealth and less frustration for everyone. So that's why I went into legal. Every entrepreneur has experienced that sticker shock of receiving that legal bill.

Now, you're selling Spellbook to attorneys, right? And the hope is that they'll pass along the cost savings from automation to the client. Is that right? Yeah, yeah, exactly. I mean, I think the cost savings will be shared both ways. I think lawyers are able to make more money because it takes less hours to produce the work. But also, they're able to take on clients that are more price sensitive that they wouldn't have been able to take on before.

There's a quote I love, I think Jack Newton from Clio, another legal tech company wrote a book. And he talks about how 70% of potential legal clients out there just don't bring legal matters to lawyers when they come up. 70% of the world does not feel that lawyers are affordable. So I think by making legal services more efficient, we can actually service that 70% of the market that's just not serviced at all today.

And so I think it expands the pie for everyone. I think legal services become less expensive. I think, yeah, individual lawyers can maybe make more money with less work. I think the wealth can be shared in both directions. Human lawyers always take on the risk and understand, obviously, the framework of the law. And there are a lot of things they do. But taking Spellbook one step further,

Do you envision a time where AI could automate the lawyer as opposed to augmenting the productivity of a lawyer? Yeah, so our first product is a co-pilot for lawyers that kind of sits in Microsoft Word, which is where lawyers spend a huge amount of their time on the side, kind of like an electric bicycle or a wing at their back that helps them do the sorts of surgical contract edits that they do on a daily basis. So very much a co-pilot that helps them do their job.

Last August, we launched our second product called Spellbook Associate, which is, you know, the goal is to actually build sort of an AI junior associate and a junior associate level AI tool that, again, lawyers can use to get work done because lawyers don't really like doing a lot of that junior associate level drudgery, the kind of things that is typically assigned to a junior associate today. So I do think that's something

achievable um within a legal team um and we only sell the lawyers today we do not think that ai tech is anywhere close to replacing lawyers for the most part today um i think it can provide you know chat gbd can provide decent answers to like basic questions but anything substantial you know i still i still don't think like a consumer should go direct to ai for because of

There was just so much judgment and information. And one of the things we learned with Spellbook, actually, before we had Spellbook, we actually had another product called Rally. And we actually tried selling

automating work and actually bypassing the lawyer. It was the very first thing we tried a long time ago. And what we found is even if we can automate all the legal work for someone, they still need that human trust element that like, I trust you and you're going to walk me through this and explain the risks to me. And I think a lot of people just don't really trust a machine to do that. So I think we're a long ways from actually

replacing lawyers will it happen someday in 10 or 20 years i mean it's it's hard to say no no one knows really what's coming um maybe my jobs of ceo will also be automated in 20 years it's it's hard to know um or maybe not i don't know the fun fact article obviously we're in agreement i don't think that should ever be the objective um and yet let's say you're talking to an audience of uh

kids graduating undergrad and going into law school, and their ambition is to land a high-priced job as an associate in a prestigious law firm or elsewhere in a company, whatever the case may be. And here we are talking about, you know, an AI associate potentially doing a lot of the things that an associate does. And I get it, how that benefits the senior partners and it benefits the client, obviously.

What's your pitch to kids who are considering entering law school?

My pitch is it's probably the best time to become a lawyer ever because through this journey, I think you mentioned we had 1,700 customers at the beginning. We've actually grown, and that's probably from our own materials. We've actually grown really fast. We have around 3,000 customers today, law firms, in-house teams. And we have routinely seen the stories of lawyers or law students going into law school, graduating, and actually just like

needing their job because of the sheer amount of drudgery. You're up until 2 a.m. with 10-word documents on your screen, and you're copying and pasting from one to another. And if you make one little mistake, it blows up the whole deal. That is not fun for most people. And I've met a ton of lawyers who were just completely...

I'm distraught at the reality of the job when they graduated and just the amount of drudgery. And there's a lot of, there has in the past been a lot of dissatisfaction. But now, what we hear from our customers are like, I love practicing law again. I wouldn't practice law without AI, but now I feel I'm getting so much more done. I enjoy my job so much more.

So I actually think it's a great time to become a lawyer because you get to actually spend more time with the client, more time on strategy, more time on negotiation, the kinds of things that lawyers were excited to do when they went to law school. And yeah, is there a risk of the market shrinking or needing less lawyers? I don't know. I think AI is going to create a lot of growth. I think we're going to have more companies, more stuff going on, like

The barrier to entry for entrepreneurs is going to get really low. So there's just going to be more transactions, more stuff happening in the world. Those 70% of customers that can't afford legal services are going to start to be able to afford them. So law firms in the legal segment have generally been growing year over year, every year. So I don't see signs of any kind of slowdown. I think AI is going to be a big growth driver in the economy.

And I think it's going to be more fun to practice law now than it ever was. Great answer. Maybe I won't discourage my kids from going to law school after all. So I read an interesting interview with you online where you said building MUNE, you didn't really experience what it meant to have product market fit.

And then you start Spellbook or Rally and then immersion to Spellbook. And here you are, you know, eclipsing 3000 customers. I didn't know that apparently what's online is a little dated, but it's phenomenal growth. And I think you're experiencing what, you know, a lot of entrepreneurs who are listening are trying to find that, you

that market pull or that, uh, where the market's pulling you as opposed to having to push. Um, talk to us about when you felt like you achieved product market fit and how do you kind of lean into that once you achieve it? Yeah. Um,

Yeah, so yeah, PMF is everything. And it's what every entrepreneur is chasing, especially in tech. It's very hard to achieve. Yeah, we kind of didn't really speak to it, but just to simplify the story. But before we had Spellbook, we actually had another legal tech product called Rally. And it was kind of doing the same thing, except it was a little pre-AI and we were using templates to automate work. It was basically like a templating engine rather than an AI engine.

And we sold it to around 100 firms with a really lean team, a small amount of capital. And we were able to sell it, but it just wasn't sticking. The retention wasn't that great. The thing wasn't being pulled out of our hands. And I think the gold standard for tech product market fit is getting to the point where your product's basically being yanked out of your hands faster than you can keep up with. And that's a really, really high bar.

The way we went about it is we just said, we're not scaling the business. We're not hiring a bunch of people until we're really, really confident that we have that kind of runaway product market fit. And for us, that meant on the path to launching the product that did hit, we actually launched over 100 different landing pages with different twists on what we thought might work. There was a huge, literally over 100 failed hypotheses on the way to

um launching a product that really did work and a lot of learning along the way um when it did hit it was unquestionable um within uh three months we had more customers and revenue than we had generated in the last three years um it was it was like a lightning strike it doesn't always happen like that i think enterprise software is usually going to be slow no matter what um but for us um

It was very much like a sudden lightning strike. We had caught lightning in a bottle and completely different. Completely different. When we showed our product to customers, their pupils were dilating, they were leaning in. I've never seen anything like this before.

And it made everything in the company just work in a way that it never had before. And then once we had found it, yeah, it totally changes your approach. You think a lot more about hiring and scaling up the company and you think about district, we have a great product, how do we distribute it, firm wide.

And a lot of thought goes into that question. When you think you have an incredible product, it's like, okay, we just need to get it in front of more eyeballs. How do we do that? And how do we do that efficiently and not too expensively? So that was a big mindset shift. But also we had to keep developing the product because it's a fast-moving market, brand new. So I think in AI, you have a good product and can just sell it. You have to be improving that product.

almost every week if you're going to keep up. Note to all entrepreneurs listening, it's always a long road to an overnight success, as you heard Scott describe. There's a reason why the P is ahead of the M in PMF, the product first, and then it takes building the right product, the MVP, the minimum viable product to be able to achieve product market fit. You said something interesting about how

The first version of Spellbook was kind of a plug-in for Microsoft Word. That's the tool that lawyers use. It makes so much sense. I'm sure there were a lot of naysayers who said that's too small a slice of value. Nobody's going to pay for that. How did you get conviction? And kind of when you were deciding what to take to market as your MVP, you must have discarded a lot of other ideas and extensions. How did you know what was MVP'd?

In terms of the intuition of what to build, should it be in Word or not, we notice a lot of patterns over time. And when we think about product, we actually don't think about... We don't think that much about what problem is the customer solving or...

Think about the customer stories. What we think about a lot is pattern matching. It's like, what sorts of patterns or shapes of things have we launched that have been successful? And what sort of patterns or shapes of things have we launched that haven't been successful? And since we iterated so much, literally over 100 times, we got a really good sense of what seemed to be working and what wasn't. And

One of the things we found is just getting lawyers to go outside of their normal daily workflow to even get them to log into a new web app in their web browser was extraordinarily hard. And people don't think about, I think a lot of people don't think about this enough. Getting someone to go to a website and put in their email and their password to log in is a crazy hard thing in any product. It's one of the hardest steps to get anyone past.

And so the thing we love about being in Word is it's just there. They log in once in the onboarding call and it just uses, actually they don't have to log in at all because it connects to their Microsoft account. So it's just there and they're logged in by default all the time and they don't have to go to some web app or some new place. And that was something we had learned over time that just friction reduction and

being in the lawyer's workflow, meeting them where they are is as important as anything that we actually build. Because changing people's habits, especially lawyers in 2025 when there's so much noise is just really, really difficult. So that was part of what drove us to go there. The other thing is, over time, we definitely learned to lean into the contrary a bit more. Like,

When we first started thinking about building the Word add-in, some of our development team was like, oh, that sounds like it's going to suck trying to figure out how to integrate with Word and all these Microsoft, figure out all these Microsoft APIs. Is a Word add-in really that big? When you hear that kind of pushback, but it makes rational sense in your head, but you get this sort of emotional pullback or emotional pushback.

It's often a sign of a really good opportunity because that emotional prickliness is what is going to keep your competitors away. And that's what's going to mean the stone is unturned. If something sounds like a good idea and it also feels like a good idea and it's obviously a good idea on the surface, then 100 people have already done it. So what you start to look for is like, what's a good idea that sounds initially bad? And that's how you find something that someone hasn't actually done before.

So that's why we had no hesitation to kind of dive into the word add-in approach. And actually, we're very confident that the prickliness of it is actually what made it attractive to us. Great answer. So we can't be having this discussion about AI and legal without thinking through what it means to exercise AI responsibly. And I'll just tip the topic. I'd like your thoughts on

Who is responsible if Spellbook AI inserts something that it made up? That's probably too extreme of an example. But does something, makes some automated decision that's accepted by your customer that ends up leading to some kind of damages, financial or otherwise? Who's liable?

The lawyer is, and it is our responsibility as a company to build a product that does not put them in a position where they feel like they're going to blindly trust this thing. And so everything we build is built to create sort of what I would call an electric bicycle for lawyers. So it's not, they're still in the driver's seat, they're still steering, they're still choosing what to accept, what to reject.

And we try to create a user interface that is not just blindly telling them things or asserting things confidently that might not be true. So an example I'll give is in our chat, when you ask a question, you're going to get citations back with that. And it's going to be like, hey, to answer your question about this 200 page document, here's my answer. And here's where I found the information that I used to give you the answer.

And it's the lawyer's responsibility to click on that citation or to look at that citation and be like, yeah, that makes sense. And it's our job to build a tool that can be used in that way where it's very easy to verify answers. I think the challenge with a lot of the early language model products is that they couldn't go and look up information. It was all based on memory, basically. It's like you asking me now, hey, Scott, answer this really complicated question about

legislation in the US and I'm not allowed to use the internet. I'm not allowed to look anything up. Well, I'm gonna give you a crappy answer. And so what really changed is providing AI models, the ability to look up information and cite information that allows them to be a lot more accurate. So I think lawyers should use tools that do that. The other thing we do in this public interface is we have a contract review UI.

And everything that we suggest is called a suggestion. And there's a button you can either accept or you can reject. And if you reject it, you can leave feedback on why you thought it was a bad suggestion. And I think getting lawyers in that mindset of accept, reject, this is a suggestion rather than this is what you have to do. I think that's our responsibility to create a user experience that's like that. You shared in the fun fact about how increasingly...

as we talk about agency, agentic AI, and giving more prominence to the I than the A, the intelligence part than the artificial part, we start to insert, again, I have an opinion here, kind of damaging language about how the AI is like a human and it's doing a job like a human and

There are companies that are putting the bots on the org chart and, you know, there's talk about, you know, should they pay employee taxes on bots? It's a slippery slope.

So it seems like you've gone out of your way to say it's a co-pilot. Here's where it's getting its information. It's just phenomenally, phenomenally good at pattern matching at scale. And it sounds super credible, but that's what it does. And you're always responsible for it.

what decisions you accept and reject from the AI. Play this scenario out. Do you think that that paradigm will persist or can we ratchet back the temperature when it comes to assigning agency or sentience to bots?

The ideal interface to work with these tools, even if it's agentic, is still sort of like, here's my suggestions. You still have to manually accept or reject them. And even if you think about how a senior lawyer works with a junior lawyer, the senior lawyer is actually liable at the end of the day, not the junior lawyer. The senior lawyer is the one who has to review and accept or reject the tracked changes that the junior lawyer has submitted.

And so the first thing I'd say is, regardless of whether it's human-like or not, you still have to kind of treat it like a very junior employee who's like, they are not the ones liable if something goes wrong. It's actually, it's the manager or the senior person who's sort of the one who's saying, yeah, putting the stamp on the work or reviewing the work and putting a stamp on it. Yeah, I think it's confusing. I think there's

the word human like or the compound word human like is like very um has a lot of meaning baked into it and i think it means a lot of different things to different people i don't like the idea of sentience like i don't think these systems are sentient or that we should treat them as sentient um i think that's definitely misguided and that's a risk of using terminology like that um

Yeah, I think, but I think what people are trying to convey when they say human-like is like, you know, it can, you can assign it a little project and it can pull it apart into a plan and accomplish those steps and review its own work and give you something, you know, from a small prompt and a little bit of input, it can give you something back that feels almost like maybe close to what like a very junior project.

a lawyer might give you. And I think that's what it's trying to convey. Is it the best way to convey it? Probably not, because it's such a loaded term that, yeah, conveys sentience and there's other things. And I don't think that's

the best thing that can be. We mainly use the word agentic. And what we mean by that, we just define it as like an agentic system is a system that can take a broad objective. It can break it down into a plan. It can do the work or the steps in the plan. It can check its own work and then it can go back and fix its work if it finds a problem or it can adjust the plan as needed so it doesn't get kind of stuck. And we say if it can do these four things, it's agentic. And

Mainly leave it at that. That's a satisfying answer. Thank you. Hey, Scott, we're about out of time. Unbelievably, this one flew by. But you're not getting off the hot seat without answering one last important question for me.

You've had a non-traditional entrepreneurial path, and I know how intense the journey is, and it can teach you so much about yourself, about humanity. It goes way beyond the tech or the product that you bring to market. Talk us through your personal journey, and what's the single lesson that you've learned most about yourself over the years? That's a big question. The biggest lesson I've learned about myself is

Yeah, this is maybe esoteric, but there's a blog post that is probably the most impactful blog post that feeds into everything we do at Spellbook. Actually, everything I do in life now, it's on a site called Ribbon Firm. It's written in 2010. It's called The Big Little Idea Called Legibility. And it talks about this concept called legibility and about how humans have this bias towards like,

really nicely structured system. So it gives this example of these scientific forests they had a long time ago compared to an organic forest. And these scientific forests look really nice and orderly, and they were supposed to grow trees more efficiently than the traditional forests, normal forests. But they actually just totally failed.

And he goes on to talk about how there's this kind of recipe for failure in the world where you look at a complex and confusing reality like the social dynamics of an old city and you fail to understand how all the subtleties of this complexity work. And then you kind of point out it and you say,

You come up with this idealized, simple, orderly vision, and you point out and you say, no, this is wrong. We need an orderly, simple vision, and this is the right way. We need strict city blocks. And I use this mathematical formula that's going to figure out how to build the perfect city or the perfect forest. And he talks about how people have this kind of

strong bias towards orderliness and create these rational utopias that just end up failing. And yeah, this is probably one of the biggest things I've learned about the most of what's changed in me the most. And also, I think as a CEO, I can go to my board and my team, and I can give them this sense of structure. And I can lay out these really nice

strategy and vision that makes everyone kind of have this very calm, legible and calming feeling like, yeah, we know what's going on. But I just don't think reality is like that. I have found that the more we have allowed our company to be a little bit more like controlled chaos, embracing the anxiety of chaos and reality is just really complex. AI is really complex.

no one really knows what the right thing to do. Like the more we've embraced that and leaned into kind of accepting chaos and the messiness of say product development or whatever, the better we've done. And so I think, you know, the biggest way I've grown and changed is, you know,

I would say when I started down this path, I was really attracted towards really nice project management systems and coming up with strategy slide decks that lay out a really concrete vision that everyone can buy into. And a lot of this stuff just didn't work for me at all. It gets everyone excited, feeling a sense of structure and direction, and then just doesn't work and it kind of falls flat. And yeah, again, the more we sort of embrace the messiness of the process and accepting chaos,

you know, the better our product development has been, the better our business development has been. So that's something that's the biggest thing that I think has changed in my perspective since becoming an entrepreneur. If you could share with me that link, I'll make sure it gets in the show notes, that Ribbon Farm blog post. That's very helpful. And that's a fascinatingly appropriate metaphor, particularly for life in an AI-first era, where as creatures, I think you're right, we seek...

orderliness and predictability and yet inherently to thrive in this and I'd argue many other times as well, learning to embrace unpredictability

and fit that into the structure that we expect the world to have is challenging, but it often is the thing that distinguishes the ones who end up doing the best achieving their goals. So very profound. And the other point I'll give is I think we're entering into an era that's more illegible. And the thing I would say is like, if you look at these really incredible chess AI engines, like the AI engines that play chess,

they can make completely illegible moves that are almost impossible for a human to understand or explain, but they're probably the optimal move. And we're going to live in a world where these AI systems are supporting us and

telling us to do things maybe or recommending things that actually eventually become kind of hard to parse why. And that's going to be very difficult and challenging thing to get to figure out. Like another example, an AI system just designed a new chip. There was a story about this a month or two ago. And the chip designers looked at this and they're like, we don't understand why this works. We have no idea, but it does. And it's better than anything we've designed ourselves.

So, yeah, we're going to be entering an age of even more sort of illegibility and we're going to have to figure out how to deal with it. Great example, right? Never doubt that machines are better than us at pattern matching at scale, but also never doubt that

The obligation is ours as humans to figure out what the chip should do, what problems it can solve. And good that AI invented it, but not going to be used to benefit other humans. I hope that's always the framework that we use. I agree. I agree with that. Brilliant. Scott, this is a great conversation. I'm glad Katie and Alex from Moxie had nothing but high praise for you. And I really enjoyed hanging out.

Thanks for having me, Dan. I really enjoyed it as well. You bet. Where can the audience learn more about you and the good work of Spellbook? Spellbook.legal. That's their website. So you can check out our product there, sign up for a demo. All the info is there. Well, gosh, that's all the time we have for this week on AI and the future of work. Of course, we're back next week with another fascinating guest.