Building a Regulated AI Law Firm: The Garfield Story

When the Human in the Loop Becomes the Weakest Link

If you're a lawyer, you've been told AI won't replace you - that the human in the loop is the safeguard. In this inaugural episode of Re-Leveraged, we explore why that model could be inverting for certain kinds of legal work, and what happens to the profession if the human stops being the protection and starts being the risk.

In this conversation, Philip Young - the founder of the UK's first SRA-regulated AI-native law firm - walks through how Garfield handles small debt claims through the courts end-to-end, why the £70bn UK unpaid-debt market has been broken for decades, and the moment he watched the system pull a critical fact from a font-size-six footnote that he, a senior litigator, could easily have missed.

He also gets candid on why he's spent a year fending off VCs, what he really thinks about Harvey and Legora's $11bn expectations, and the six-month window in which he believes lawyers could stop being a safeguard and start being a liability.

What's in the episode

00:00 - Introduction to Garfield and AI in Law Philip introduces Garfield: the first SRA-regulated law firm in England and Wales built to run on AI, recovering debts under £10,000 through the Small Claims Court. After 25 years as a commercial litigator, he frames himself firmly as an AI bull and explains the core idea - software that does everything a traditional firm would do for a client, only faster, cheaper, and available 24/7. Crucially, Garfield is regulated in exactly the same way as any other firm; the difference is that the service is delivered by software rather than people. Debt is deliberately the first product on a modular platform, with more to follow.

06:29 - The Journey of Building Garfield The origin story starts with ChatGPT-4 landing in 2023 and Philip realising the technology had crossed the threshold from interesting to useful. He wanted to build something to help his brother-in-law, "Andy the plumber," and businesses like his. Knowing he could code but not build it alone, he "beauty paraded" candidates and found his co-founder, Dan, later adding Pablo on the front end. He's honest about the false starts too - the team began in Flutter assuming a mobile app, before realising that a document-heavy legal product is fundamentally a desktop and laptop experience, and switching to React.

11:38 - Market Research and Commercial Viability Before writing serious code, Philip did the commercial homework. Ministry of Justice data showed roughly 1.7 million small debt claims issued a year - but the far bigger insight was that the majority of small debts are never pursued at all, because using lawyers is uneconomic and going it alone is too daunting. That meant the real total addressable market was many times the visible one. Sizing the broader unpaid-debt problem surfaced figures ranging from the Small Business Commissioner's ~£26bn to Hiscox's ~£70bn. The takeaway: a vast, deeply inefficient market that was effectively creating demand nobody was serving.

17:01 - Garfield's Commercial Model and Pricing Strategy Philip opens up on pricing without opening the books. A Letter Before Action costs £7.50 plus VAT, with optional postage, and later court stages are priced at or below what the court allows to be recovered - so the cost is largely passed through to the debtor. It's a volume model, with subscriptions likely once there's enough platform data to price them well. On the ideal client, he's candid about the split: his heart is with the sole traders and plumbers Garfield was built for, but the economics live with high-volume corporates - hence different routes to market, from brand advertising for the small end to a traditional sales motion for the large.

19:27 - AI Oversight and the Future of Legal Tech This is the philosophical core. Philip distinguishes deterministic processes (no review needed) from probabilistic, LLM-driven ones, which he still checks - while building an AI overseer to take on first-level review. Then comes the footnote story: Garfield surfaced a critical fact buried in a font-size-six footnote that he, a careful senior litigator, would probably have missed, because to an AI the size of the text is irrelevant. That's his evidence for a provocative claim - that the human in the loop is approaching the point of becoming a risk rather than a benefit. He covers his quarterly SRA meetings, the relevance of the Mazza case (which framed this as a supervision question), and why accountability stays with him as the firm's principal regardless.

27:09 - Scaling and Future Plans for Garfield Philip explains why, despite a year of inbound VC interest, he's funded Garfield through angel investors chosen as much for their expertise and judgement as their cheques. He watches Harvey and Legora with genuine admiration but no envy, reflecting on the astronomical expectations that come with raising at an $11bn altitude - and why moderate expectations suit him better. He's not anti-VC, but the money has to be aligned with the access-to-justice mission. With the first salesperson just hired and inbound demand from large companies and even a regulator, the business is moving from early stage into a deliberate scaling phase.

35:50 - Expanding into New Legal Areas Here the modular thesis gets tested. Philip argues the model generalises to any process that looks like litigation - parties set out the issues, evidence is produced, and someone decides who's right - including non-court processes that government departments have already approached him about. On complexity, he describes civil litigation as an S-curve: small claims are comparatively simple, fast-track is clearly within reach now that the system writes strong witness statements, intermediate-track feels achievable with another year of development, and the very largest multi-track claims remain out of scope for now.

42:04 - Advice for Aspiring Legal Entrepreneurs Philip's closing message is direct: go for it. Start from a real problem you're passionate about solving, because passion is what carries you through the unglamorous 2am problems no founder escapes. He offers a pointed warning on vibe coding - building a prototype is one thing, shipping something production-ready in a regulated field is another entirely. And he's reflective on funding: the type of money you take shapes the incentives and behaviours of the business, so choose carefully. His one regret? Probably building out a sales function several months sooner than he did.

https://www.youtube.com/watch?v=cgJwh4G324w

---

Full Transcript:

Alex (00:10)

Hi everybody, welcome to Re-Leveraged, conversations with founders building interesting businesses in the legal sector. Each episode explores how lawyers and legal entrepreneurs are creating value and servicing clients in completely new ways, with a focus on the commercial realities of what they're building. Today's guest is Philip Young. Philip spent 25 years as a commercial litigator and in 2022 started building Garfield. The first SRA regulated

law firm in England and Wales and across the UK that runs on artificial intelligence and it helps organizations to recover debts of less than £10,000 through the Small Claims Court. In the last 12 months he's been on Channel 4 dispatches, Bloomberg Law, Forbes, The Economist, The Times, more mainstream press than almost any other UK legal tech founder that I know, certainly in my recent memory. Philip, thank you for joining us. Welcome to Releveraged.

Really excited to welcome you as our first guest.

Philip (01:11)

Thank you very much Alex and thank you for that very kind introduction. I appreciate it.

Alex (01:16)

Great, thank you. So before we get into it, I've got a question for you. Are you an AI bull or AI bear?

Philip (01:20)

Okay.

I think I would go for AI Ball because I think that AI is absolutely transformational for law and legal services and probably for very many other professions and markets as well, although I obviously haven't focused on them as much as I have done on my own.

Alex (01:43)

Excellent, I thought you might say that given the nature of what you do. So great, let's hear a bit more about Garfield. You launched in 2022, if I got that right?

Philip (01:46)

Heh.

Well, we started building it in 2023. I think 2022 was when ChatGPT 3.5 came out and then that was when in sort of October, November time. And that was when I first had the sort of realization in playing with it that this technology was going to be transformational. But it actually didn't occur to me that I was going to do anything with that until 2023 when ChatGPT 4 came out. And then I thought, gosh, well, maybe I could build something.

with obviously other people. And in terms of when we actually launched the product, it was really when we were regulated, which was last, it was March last year, though we were going to do a stealth launch for the first year, but that was then, we were then unstealthed, if that is a word, in May by the Financial Times discovering us and then writing about us. So that's been our of, our sort of, our story so far, so to speak.

Alex (02:27)

Okay, amazing. I actually wrote about you around about that time as well. And I think it's still the most engaged piece of content that I've ever put out ⁓ on LinkedIn. So it just goes to show that there's huge interest in what you've done and what you're continuing to do. And you have got immense publicity and coverage. So congratulations on that. I personally have been singing your praises to... ⁓

anyone that will listen really about the model that you've developed and the customer experience that you've created. But can you tell us a little bit more about the business and actually what it does?

Philip (03:30)

Yeah, sure. And well, thank you very much, Alex. mean, all publicity is good publicity, as they say, and I appreciate you saying nice things about us. ⁓ Yeah, so what is Garfield? Garfield is essentially a software application that does everything that a traditional law firm would do for a ⁓ client in pursuing a small debt claim through the English and Welsh courts. ⁓

it is regulated in exactly the same way as any other English and Welsh law firm. The difference fundamentally is that instead of providing a service by humans, it provides a service powered by software and AI. It's a lot faster than humans, inevitably, and arguably it's a lot more convenient because it's available 24-7, whereas, of course, for a lawyer, you've got to make an appointment and then talk to them when they're available. And it's also a lot cheaper.

Inevitably, because it's a software product, we don't have to worry about human costs. And so we're able to price it in a very different way. the it's no secret that the debt product is a first of many products, which we're going to be rolling out whole series of further products in the course of the next few months, because we built we've always said we built Garford in a modular way. And the idea is that that it's a platform. And so

people with other sorts of claims will be able to drop into Garfield and Garfield will help them in exactly the same way that it helps people with small debts. I should say that in terms of small debts, we already cover the entire range of businesses in England and Wales. So we have everyone from sole traders all the way up to as of the last month, a regulator, not the SRA, but I won't say who. And that's amazing to have the credibility of a regulator on the platform.

And a surprise to me because you think regulated individuals would pay their debts to their regulator, but apparently not all the time. ⁓ And also a giant utility ⁓ and some other companies as well that are very big are coming on the platform. we're seeing our original thesis that there would be a demand across the whole business spectrum being proven. And so what we're now doing is moving from ⁓ the early stage into the next stage of our evolution, which is a scaling stage.

and we've just taken on our first salesperson and we're starting to do outbound conferences and outbound marketing. So we're sort of interesting inflection point where we're now going to be beginning to quite a lot.

Alex (06:02)

Okay, amazing. I'm very excited to hear more about what you're going to be doing moving forwards. And I've got a couple of questions for you on that later on. But can we go back to right where this all began? How did you get started? What did the first six months look like for you? And I think, you know, for me, the beginning isn't necessarily when you incorporate your company is when you're thinking around the business starts to

crystallized. So what did that like first six months look like from from that point?

Philip (06:38)

Yeah, so this was 2023. basically, Chat GPT-4 came out, I thought this is a really big jump in capability. And now it's possible to build something useful with this. And then I knew what I wanted to build. I wanted to build something that would help my brother-in-law, Andy the plumber, who I've spoken about elsewhere, and businesses, other businesses that have small debt claims. So

I knew what I wanted to do. And I also knew that I wanted to do it in a modular way so that we could expand it to other claims in the future. But I wanted to start with debts. And then what I thought to myself was, OK, I am a nerd. I can code. ⁓ It's not my sort of core skill set. So although, as I'm talking to you at the moment, I'm very conscious that on my left out of camera shot, I have a huge wall of programming books. And there's some more recent than others, because some are sort of Python and

got help me TypeScript, not a fan of TypeScript, ⁓ but some sort of older languages like C and C++ from when I was much younger. So that's sort of one of my hobbies. And so I knew I could do some coding, but I also knew that my skills were not extensive enough that I could build the product by myself. So I knew I needed a co-founder and somebody more technical than me. So I went out and looked for somebody and I found Dan. And essentially the way I did that is I... ⁓

I beauty paraded people, know, it's that phrase that solicitors like me use when we refer to instructing barristers in traditional litigation or experts, people like that. I went off and I found people that had what I thought were the right skill sets, a passion for something, doing something like this. And of all the people I met, Dan was the was absolutely clearly the standout guy. He we just got on really well. And I think we've

with any sort of new business, you need the founder, whether they be partners in a law firm or shareholders in a traditional company or founders in a tech company, you've got to have a really good personal relationship between the founders. I always think it's a bit like a commercial marriage. I don't know if Dan would appreciate that description of our relationship, but you know what I mean. And yeah, and he was

Alex (08:51)

Yes.

course.

Philip (08:59)

He's such a clever chap and he got it and he was passionate about the project. And so between the two of us, we started building a prototype and we, although I've spoken about Python and TypeScript and that sort of tells you that the front end is built in React, we actually started off with Flutter because we started by thinking, it would probably be a mobile app. And so we built some of the backend in Python and then did a lot of the

front end in Flutter. I actually really like Flutter and I particularly like the language that it uses Dart. I think Dart is a very elegant language, programming language, but it became apparent to us pretty quickly that this was really a desktop laptop product, not a mobile product. So we decided not to bother with Flutter going forward and switched to React. And then

Alex (09:48)

And what

told you that you should make that switch?

Philip (09:53)

So basically trying to look at PDFs on a mobile phone. Because I mean, essentially, any sort of product that is law related is document powered, document heavy. And we just thought, well, OK, you can look at PDFs on a phone and you can sort of turn the phone sideways and all of that to read it. But it's quite painful. although Garfield is mobile ready and works on mobile and has its own mobile application,

Alex (09:57)

that's better,

Philip (10:22)

In fact, it's really fundamentally a desktop and laptop product so that you can see the documents and do the dialogue with Garford effectively. So yeah, so then what we did is we realized we needed more help on the front end because Dan is a full stack guy, but he's probably more backend than full stack. And so we went off and found Pablo, who Dan knew, because I've always believed that if you're recruiting, recruiting is fundamental to any business. You've got to find the right people.

⁓ And it's so much easier if people already known quantities, you're not taking such a big risk. so Dan had worked with Pablo at a different startup before. so Pablo came on board and then ⁓ that was probably the first six months. And then later on, of course, we expanded the team more and then we moved to a world where we also obtained some external funding to expand the team and progress.

Alex (11:17)

Okay, excellent. And I mean, you talked a lot there about the product side. How much commercial thinking did you do before getting into the weeds of building an actual application? Did you go out and do some customer research beyond speaking to your brother-in-law? ⁓ And how big do you think this market is? Did you look at that before you started building?

Philip (11:46)

Yeah, I mean, you've got to do that. I think there's no point building a product if if no one's going to use a product or if if you're building something that either there's no market for or you're building the wrong sort of product. So there is a demand, but you're not hitting the demand. You've also got to think a bit about pricing and stuff like that. So I knew from my days as a practicing lawyer that there are plenty of debt claims. And I did a bit research and the Ministry of Justice publishes statistics quarterly on

on county court claim issuance. And so I could see from that that there are an average about 1.7 million debt claims, ⁓ small debt claims issued a year. So that gave me an idea of the existing market. I was also very conscious that for my time in practice, that the over one majority of small debt claims are not pursued at all because using lawyers is just uneconomic and too time intensive and doing it yourself is just too time intensive ⁓ and daunting.

for a lot of businesses. So I knew that the total addressable market would be many, many times the current issuance. And actually, when I look at what we are experiencing, I think I'm talking to some of our users that it's anywhere between five to 10 to more times than that, because we're seeing people that say, I might have issued one claim form and I'm issuing 10 or 15. So I knew that if done right, the technology would

would open up a market that at the moment is very inefficient. And then I also did a bit of research on the total size of the, what you might call the debt market, the unpaid debt market in Wales. And there's many different statistics out there, many different surveys. The biggest figure I've seen was a survey done by Hiscox, the insurance company, and they said 70 billion pounds a year, which is a huge, huge figure. The small business commissioner,

did a survey just solely aimed at small businesses and she came up with a figure of think about 26 billion pounds a year. So I've seen figures of sort of six and a half billion. So there's, there's many different figures. What is very clear is that it is a huge market and it's what you'd expect because anyone who's been in business knows that every week or month, depending on how often they do it.

has to be people sat down and look at an aged debtors report and the aged debtors report has always got lots of aged debtors on it that people would prefer to be cash rather than unpaid invoices. I knew whatever, however you decided to try and analyse it, it was an enormous market and it was very inefficient and right for some improvement.

Alex (14:25)

Amazing. And what I really love about what you've done is you've in essence created a new market that previously wasn't being serviced, if I've understood you correctly. So yeah, that's really exciting. ⁓ And on that note, who is your ideal client? From what I understand, you have a couple of different ⁓ service tiers as part of the Garfield offering. So how do you think about the different ⁓

Philip (14:34)

Yeah, we think we have, yeah.

Alex (14:55)

ideal client profiles if you like ⁓ and how are you tapping into each each part of those different market segments?

Philip (15:04)

So yeah, so I mean, we're not a very narrow product. We're broad because as long as you've got a debt up to 10,000 pounds and your debtor is in England or Wales, you can use Garfield. So therefore pretty much any professional industry vertical can use Garfield. I've not come across any that can't so far. And so we've got a very broad market and a very broad sort of ICP.

In terms of where my heart, where our heart is, obviously our heart is the smaller end of the business spectrum. So we're thinking about the Andes who are the plumbers of this world because we started building something to help them. But the economics of the business are of course in the big companies with ⁓ a large volume of debts and large volume of claims because you need volume to support this sort of business model. And so we have sort of

different routes to market. ⁓ And for the larger volume users, it's a traditional sort of route with sales people and going out to conferences where they're going to be at and things like that. That's what we've decided we're going to do with those sorts of people. Obviously, because we have hit this inflection point where we're about to start scaling, we've been just dealing with inbound so far. And so quite a lot of big companies have found us just from all the press interest and the marketing we've been doing. But that's sort of brand marketing. It's not sort of

focused outreach marketing. So yeah, very much on the bigger companies, it's traditional routes. And then smaller companies, very much things like Google brand advertising and getting known in the right circles with influencers and people like that. That's very much our approach. And from time to time, of course, we do a bit of, you know, old fashioned sort of SEO. And we'll probably do

more of that as time goes on. But that needs to be quite focused, I think, to reach the right sets of eyes.

Alex (17:09)

Excellent. Would you mind sharing a little bit more about your commercial model? And I think from what I looked at previously, that your letter before action in terms of like the first step in this process is just £7.50 on Garfield. So how are you able to operate at such a low price point? without asking you to open the books, can you give us a sense of what the unit economics?

might look like.

Philip (17:41)

Yeah, so our economic model is, at the moment at least, sort of follows what the court allows recovery of. So for Letters for Action, we charged £7.50, obviously plus VAT, and there's also an optional postage charge as well. If the user wants Garfield to hard copy, post a letter for them, which most users do, because it's so much more convenient for them than them having to sort of download it and print it and post it themselves.

And with LBAs, the pre-action protocol says that you have to post those. So someone has to post them. So it's easy to get Garfield to do it. And then for later stages of the court process, we charge per stage and we charge a fee that is at or below the sum of money the court allows recovery of for those stages where there's a couple of fee. So for example, for ⁓ a small claim of ⁓

think it's five to ten thousand pounds. From memory the court allows recovery of a hundred pounds so we charge hundred pounds for that. So from our perspective the economic model is both in volume and also primarily from court claims because as you go through different stages of a court claim you can charge different fees. So if you issue a claim form then apply for default judgment for the default judgment you can

For memory, the fees are either 22 or 30 pounds, depending on how big the claim is. And of course, that's recoverable from the debt as well. At some point, we will ⁓ offer subscriptions as an option, I'm sure. But we need enough data on the platform to be able to price that effectively in a way that people would use. And obviously, as we expand the product offering, then there are more things that users can, if they want, do and pay for.

So that's the current economic model. So it's very much a sort of volume model.

Alex (19:35)

Yeah, yeah, I hear you. hear you. And are you still involved in reviewing every single output that the system produces?

Philip (19:47)

Less so. some of our processes do not involve any form of probabilistic ⁓ LLM process. And so for those, they are deterministic. And it's not an essay for me to review documents that come out of those anymore. I do because I'm, ⁓ like all lawyers, I'm a little bit OCD. When I say a little bit, I mean, very much OCD. And so I do sort of sample those and drop in and drop out.

But for the probabilistic documents, I'm still checking those. The volume is not such that that's difficult or impossible. And we've also built and are constantly improving an AI overseer itself that I've also been keeping an eye on to make sure of its accuracy. So in due course, we will have that as a first level review ⁓ solely. And then we'll only escalate to humans when that thinks something needs to be looked at by humans.

It is an interesting process, because as I've been sort of jovially saying in other places, ⁓ I can see the point in time is rapidly approaching where the human in the loop will be a risk rather than a benefit. So I give as an example something that happened last month, where I looked at a document that Garfield had produced, I thought, where does that fact come from? I remember seeing that fact. And I went back through the claim and had a look.

And what Garfield had done is it spotted a footnote in a document the user had uploaded and the footnote was in a very tiny font size. But of course to an AI, doesn't matter if it's font size six or font size 40, it's exactly the same text. And then Garfield had found an important fact buried in this font and pulled it out and applied it. And I as a human probably would have missed that. I probably would not have spotted the importance of the tiny footnote. And I just thought to myself,

You know, this is a demonstration of when we will get to a point where technology is more reliable than humans and then the humans in the loop will be a negative rather than a positive.

Alex (21:53)

Yeah, I hear you. I couldn't agree more. And you've obviously got a very close relationship with the SRA. What is their view on that? Is there a point in time at which you can remain a member of the SRA and regulated but actually not have a human in the loop? Is there a certain threshold in terms of accuracy that they might be looking for?

And I think this is a question that's on a lot of people's minds, right? It's like, at what point does the human in the loop no longer become ⁓ a mandatory component of delivering the service?

Philip (22:38)

Yeah, so I mean, yeah, we've been meeting with the SRA every quarter since we were initially authorised. Originally, they preferred to meet every half year, but we proposed every quarter because we thought that would be more useful to them and also to us. And they agreed and they've been very interesting meetings. We've had a very sort of candid discussion every time about what's happening in the market, what's happening at Garfield. ⁓

what we're seeing. And we've been trying to pass back know how as well, because we are the guinea pig for the SRA on this and also arguably the guinea pig for the legal community globally, given the way the first people to be doing litigation using AI in a regulated way. So we've been very conscious that we owe a sort of citizen's duty to try to help the regulator sort of prepare for this new world.

And also to a slightly lesser extent, we've been talking to the legal services board, the legal super regulator who also are very interested and also parts of the government. So in terms of the SRA's view, as I understand it, of the future, they've always been willing in principle to accept the idea of not having to check every document. very much as long as.

As long as what's going on is clearly prudent and appropriate in accordance with the regulatory rules and also your duties to the court, then they've been comfortable in principle, the idea of that. I've not yet reached a point where I want to move to a purely sampling approach for things that involve LLMs. that's, because I'm a very, like all lawyers, I'm very cautious and very OCD and I'm probably a bit different. I probably struggle personally to let go.

⁓ But I will eventually, I will eventually. ⁓ also it's just so interesting reading what Garfield produces and how it's produced it. mean, so you have these, even though it's a product I've been working on for three years now, you I still have ⁓ every week one or two, you know, crikey moments where I look at something, that's so good. Like last couple of weeks ago, Garfield did a set of witness statements and they would just, I mean, I read them and I thought I...

I changed like one sentence in four witness statements and that was it. And then I only did that because I thought I should do something. ⁓ That's how good they were. It's amazing. yeah, so the other aspect of this, it's obviously also been relevant, has been the Mazza case ⁓ and what that might have meant. And actually, I think that has opened up ⁓ or at least not created a blockage ⁓ for sort of sampling as an approach because the Court of Appeal was really clear that

Alex (25:01)

Amazing.

Philip (25:23)

⁓ it's fundamentally a question of supervision. ⁓ And as long as you've got regulated, authorised people doing an appropriate level of supervision, the regulator's happy with, ⁓ then that's sufficient. So I think that's also helpful too, because it means there's not some sort of difficulty created by the regulatory rules. ⁓ so I think that's where we are. We're going to end up in a world where

I'm still responsible for everything. I'm accountable. I am the culp of the law firm and I think the SRA would expect that. They're not going to allow a regulated service to be provided without all of the accountability and responsibility that comes with that.

Alex (26:07)

And if you had to put a timeline on it today.

Philip (26:13)

⁓ OK, go on then. But this is going to be tricky, Alex. ⁓

Alex (26:15)

Yeah,

at what point do you think it will be possible for you to extract yourself from being involved in the process?

Philip (26:27)

I reckon in the next six months, I mean, I'm moving more towards a sampling approach and focusing on particularly high risk areas, which really means claims that have got defenses and they're going being defended in court. ⁓ But that said, ⁓ I think I think the question of when I could is slightly different from when I will just because as I say, I'm a little bit too hands on. It's all just too darned interesting, Alex. ⁓

Alex (26:53)

Hahaha

Philip (26:54)

But

yeah, yeah, I mean, think sometime in that period, I would be very much beginning to reduce my checking. I'd also want to talk to the SRA in advance and make sure that they're comfortable with what was being planned. That would be important for me as well, because they're on this journey too, and I don't want to do anything that causes anyone any sort of angst.

Alex (27:17)

Amazing. What, so I just changed tact here a little bit and we were talking earlier about some of your plans to expand the business, new service offerings, et cetera, new parts of the market. What is your ultimate goal here? Like how big can this thing be? And I think you've been ⁓ quite vocal before about the fact that you didn't think that this was a... ⁓

a you were skeptical about whether or not Garfield was a VC backable business. ⁓ I think that was well when we first met you were talking about that but you publicly talked about it on the non billables podcast in May of last year. Has your thinking changed on that since then?

Philip (28:05)

So yeah, mean, this has always been a business that VCs could back and God knows I've spent a year of my life trying to field VCs trying to talk to me ⁓ of all shapes and sizes and all sorts of different countries. ⁓

Alex (28:20)

I think a few

people would say that that's a good problem to have in some respects.

Philip (28:24)

Well, yeah,

maybe, yeah, if you want VC money. ⁓ It is a good problem to have. Yeah, I mean, we've taken the approach today ⁓ that we haven't sought VC money. We've got ⁓ a group of angel investors who are brilliant and they've not just sort of supported us financially, but even more importantly, they've supported us with expertise and know-how and wisdom and generally being just helpful people to have around and they've all got

Alex (28:28)

Indeed.

Philip (28:52)

very ⁓ different sets of skills and I did pick them, ⁓ they were all friends of mine, but I also picked them because I knew they'd all be bringing something very different to the table that would also help the business and help get it to where it will need to go. And that said, mean, I can see that VC money, if from the right VCs who understand the access to justice side of the mission, could be potentially very useful in terms of

accelerating growth and moving into other markets and things. So yes, it's possible that VC money will happen, but depends a bit upon who they are and what they want to achieve and how aligned they are with what we're up to. I do watch, I mean, watch LaGora and Harvey with a certain amount of wry amusement. I think, know, Max and Winston and Gabe have done

Alex (29:37)

Indeed.

Philip (29:48)

brilliant job in what they're doing, but I also sort of feel for them because the amount of money they've raised, the expectations on them as founders must be astronomical. I'm not sure I want astronomical expectations on me. Moderate expectations might be fine, not $11 billion or whatever expectation.

Alex (30:06)

you

Indeed, yeah, it's certainly a lot of pressure and a lot of thinking that might need to be done in order to achieve a lot of work that will need to be done in order to achieve that valuation. So I appreciate your considered approach about when was the right time to take some money. I think the other

Philip (30:29)

Yeah, considered is a

good word. I try to be considered.

Alex (30:37)

In terms of where you're taking this and where you're going with this, you've mentioned some new markets, you've mentioned some new products. What can you share about what's coming in the next six months?

Philip (30:50)

Well, I mean, I can say a few things because a few things have been said sort of semi publicly, so I can talk a bit about that. So, I mean, from our perspective, we are now aiming at completing essentially the automation of the small claims track and we are on the edge of having completely automated that. So we're about to roll out some functionality that deals with enforcement of judgments. And we're also going to do some other types of.

of small debt claims. So at the moment we don't handle landlord and tenant for example, but that's coming very, very soon. And then we've got other sorts of claims that are not debt claims that are on the way as well. And then once we've finished with small claims, I think we'll look closely at doing fast track claims. So, know, debt claims from 10 to 25,000 pounds. Once you can do witness statements effectively, there's absolutely no reason why you can't do fast track.

So we will we will look at those ⁓ and then build accordingly. It's since day one, we have been thinking not just one thing, we've been thinking further ahead and how else we can build things to help people. And essentially, the fundamental thinking is always where is there an access to justice gap? Where is there a problem here? Where where is it that companies cannot and individuals, individuals very important to cannot effectively get legal representation and, you know,

What could we therefore do to solve that?

Alex (32:20)

Yeah, and I love that word problem as well. think it gets bypassed quite quickly when some people get excited about the solution, particularly with everything that we've seen come through with vibe coding and all the applications that you can now use to create software and prototypes very, very quickly, which is hugely addictive. If you've ever, if you've ever ⁓

Philip (32:37)

Yeah.

Alex (32:48)

ever use those tools as a non-software developer in terms of somebody that doesn't have those capabilities. But what I think it amplifies is the ⁓ lack of due diligence that gets done on the commercial viability of a venture. If you're building something to turn this into a business, really everything is embedded in where is there a problem?

and how much value can we create by solving it? And you're evidently doing it that for a section of the market that clearly needed some help because the alternatives weren't fit for purpose. ⁓

Philip (33:19)

Yes, absolutely.

And

you can see that too in the number of people that are trying to use chat GPT to work their way through a court process, wherever it might be, whether it be in civil or family or crime. And that just shows you that there's a problem because you want to be able to provide regulated products to take people through basically a regulated process. And I agree with you. think anyone who's thinking about founding a business or starting a business, they should start with what is a problem.

I'm trying to solve and be passionate about wanting to solve it because, ⁓ you know, anyone can identify a problem, but you will have your own problems as you build the business. And there'll be those times when you're up at 2am trying to deal with something that's naughty and difficult and not very exciting, frankly, those sort of problems are never very exciting. And they'd like, why am I doing this? And you really want you need to be passionate about what you're doing in order to get through those sorts of times. And, and

Alex (34:25)

Yes.

Philip (34:27)

Yeah, on the vibe coding front. I mean, I've loved that. I've loved watching all these lawyers ⁓ discovering coding and using Claude code. I think it's great. a fair play to them all. Obviously, what I have gently said to some of them who have talked to me about this is, ⁓ it's one thing to build a prototype. It's another thing to have something that's production ready, particularly in a regulated field, and be a little bit careful about that. And that's basically the same advice that a lawyer would give to somebody who is, I don't know,

gone off and asked ChatGPT to draft them a contract. You know, the one thing to do a prototype is another thing to a properly ⁓ solid thing that survives careful scrutiny. yeah.

Alex (35:08)

Completely, completely.

And I think about that whole problem piece in a slightly different way. I gravitate naturally straight to the business end of the spectrum, which is, we can solve this problem, how much value do we recreate and therefore how much can we charge? And how much of this problem exists within the marketplace in which we want to operate? You're clearly starting to expand into some closely adjacent areas.

Philip (35:24)

Yeah.

Alex (35:37)

But do you think your model, this way of working can be applied into completely different areas of legal services? And how far up the complexity spectrum do you think it's possible to go with this approach?

Philip (35:58)

I don't see any reason why it can't be applied in other sectors because what we are doing conceptually is, well, it's not, let me approach it a different way. If you can provide a solution that takes someone through a litigation type process, there are many such processes that may be litigation and might not be litigation that ⁓ are conceptually similar.

where the parties have to explain what the issues are. There has to be production of evidence that relates to those issues. And then someone else, which in litigation context we call a judge, has to decide who's right and who's wrong. So conceptually, that process applies to many things. And if you can build it for one type of process, and there's no reason if you've done the construction in a modular way, you can't adapt it. And actually, to that note, we've been approached by

various sort of government departments who have situations where they have ⁓ basically conceptually similar problems. In other words, they have a ⁓ large volume of ⁓ things that are contentious in nature that are relatively small value that at the moment they try and deal with basically by humans at considerable cost. And they're then looking at things like Garfield and saying, well, you could build something for us that would dramatically accelerate all of this.

And my answer to that is, yes, we can. Absolutely, we can. And there's plenty of advantages for them in exploring that and also, for us in helping them explore it. So that's an example of sort of how you can adapt this sort of technology to things that are analogous to litigation but aren't. And then separately, to your point about complexity. So I think this is limited to an extent by

by the capability of the existing technology. And it's limited partly by the abilities of the founding team to build a product that can match the level of challenge required. I think those are the two fundamental issues. I've always described civil and commercial litigation, the complexity as being a sort of S curve like that, because small claims are in comparative terms, reasonably simple. Fast track claims more complicated and intermediate.

obviously a lot more complicated than Ditto multi-track. But when you get to very, very big claims, like the sort of claims that I used to do when I was in practice, actually the complexity then begins to flatten out again, because if you're doing claims for hundreds of millions and billions, then they're perhaps complicated. There's other claims for hundreds of millions of billions. So yeah, and I don't think at the moment that the existing technology could pursue a multi-track claim, for example.

Alex (38:43)

Thank

Philip (38:52)

I definitely think it could now do fast track. And I think with the right ⁓ structure and architecture and also with a bit of technological development over the next year or so, I think we might be able to get to a world where it could start to do intermediate track cases very effectively. ⁓ You may see a world where you need to have a lawyer on top of the technology for some time, but over time, I think that will drop off. ⁓ So yeah, that's where I am sort of in my head now, which is different from where I was a year ago.

because obviously things have moved on a lot in a year and goodness knows what will happen in another year's time. Sometimes I look on social media and the internet and things and I see sort of people doing charts of the current progress in LLM models and stuff and some look like that and some look like this.

And obviously a looks like that. Well, I mean, it goes exponential. It comes to a point where none of us are doing anything because we're all sitting around eating grapes because the computers are doing everything for us. And I'm a bit skeptical we'll get to that. So there have been any and then banks of a culture world in which minds do all the work and humans to sort of do cultural things.

Alex (40:02)

Yeah, indeed. think I share your level of the AI bull bear spectrum. I'm quite bullish on it, but I'm not yet at the point where I can see a world, certainly in the next couple of years, where anyone that's doing a desk job in front of a computer is removed from the situation.

I might be completely wrong. Things are moving very, very quickly.

Philip (40:35)

Yeah, I

like you. also think there's always I think what will happen is there'll just be different jobs, jobs will just evolve. rather like in the Industrial Revolution, the invention of what was it called the spinning Jenny or whatever it was that made the production of certain products far quicker and far more efficient, just meant you had less people actually sitting around sewing. I think it's similar for for professional services, we're going to be spending a lot less time doing

boring, repetitive grunt work. And also a lot of advisory work is is repetitive too. I mean, was very conscious of that in practice and even more conscious at Garfield that a lot of the standard recommendations you give at particular stages are just standard. You know, to explain, you've got to produce your documents. So you want to produce the ones that are in small claims track, the ones that are helpful to you because there isn't the extensive rules on disclosure there is at the higher tracks.

That's just the same recommendation every single case. There's no real reason why someone should bill an hour of their time to give that same recommendation to just a new client.

Alex (41:42)

Yeah, I hear you.

Philip, we're coming up to time. Can I ask you, what would you say to someone that's been thinking about embarking on a similar journey? Perhaps they're a lawyer in the traditional sense, they've been inspired by what you've shared here today, or they've been following you, or what you've been doing over the last couple of years. What would you say to that person?

Philip (42:11)

go for it basically. If they are passionate about what they want to do and this is what they want to do and they want to build something and what they want to build is worthwhile then we only get one life and there's no point looking back years later with regret because you didn't try. Obviously people should be responsible and if they've got obligations such as a mortgage and things they need to work all that out how are they going to sort of meet those responsibilities but

No, I encourage people to do entrepreneurial and interesting things. I think that's very important, not just for themselves, but also for the economy and for the country. And if they build something that helps in a really big way, arguably even the species, which would be pretty huge. know how many people manage that. so, know, so I'm very much for it. think law as a profession has generally been

seen as not very entrepreneurial and I think that there is a little bit of truth in that but I think it's also a bit overblown because there's plenty of lawyers who set up their own law firms like I did as twice as it happens ⁓ but yeah I think go for it that's what I say.

Alex (43:26)

Great, thank you. And final question for you. What's the one thing that you wish someone had told you before you got started?

Philip (43:36)

⁓ I'm facetiously tempted to say the EuroMillions lottery numbers, one of the weeks, that would have very helpful. What one thing could somebody have usefully said to me? That's a very good question. ⁓

Alex (43:45)

you

Philip (43:56)

I think in retrospect, we probably could have started earlier in terms of thinking about a sales team and things. ⁓ So we've just, as I said, taken on our first salesperson, but we probably could have done that sort of October, November time last year. ⁓ So that would probably been quite helpful actually, because it would have taken some of the burden of doing demos and doing sales and speaking events and things off my shoulders. ⁓

That said, I don't mind doing that. It's quite fun. But yes, so maybe we could have moved quicker in that regard. I think as well, the other thing is, and this is something I thought about very carefully, is how you're going to fund whatever you're going to do. And you know, what sort of money do you want behind it? Do you your own money, angel investors, VC money, both sorts of questions, or maybe some sort of deal with the law firms? lot of law firms are looking about making investments into law tech.

And I think think very carefully about that because whatever type of money you get will have a big impact upon ⁓ the incentives and stuff and therefore the behaviors in the business.

Alex (45:03)

Yeah, amazing advice. Thank you for that. And, Philip, I have enjoyed this immensely. Thank you for being generous with everything that you've shared.