Hassan Luwalira Podcast Transcript
Hassan Luwalira joins host Brian Thomas on The Digital Executive Podcast.
Brian Thomas: Welcome to The Digital Executive. Today’s guest is Hassan Luwalira. Hassan Luwalira is the founder and CEO of Unwilderd, the company behind Caira, one of the first legal AI products designed for consumers. He began developing Caira in London in 2023 after his own experience of representing himself in the legal system.
Without a background in law or software engineering, Hassan approached the challenge as someone who understood how overwhelming legal problems can feel, and how much difference clear, timely support can make. That experience shaped Unwilderd’s mission to build the best loved legal products powered by AI that make justice affordable, accessible, and equitable for everybody.
Well, good afternoon, Hassan. Welcome to the show.
Hassan Luwalira: Good afternoon to you, too. I’m really happy to be on your show.
Brian Thomas: Absolutely, my friend. I appreciate you making the time. Hailing from across the pond, as we call it, the, You’re in London, the UK. You’re there in England, and I’m in Kansas City. So we’re just traversing a few time zones, but I always appreciate that ’cause I know it’s hard to get on calendars.
So Hassan, let’s get into your first question here. You describe yourself as an accidental entrepreneur with no legal, software engineering, or STEM background, yet you founded an AI legal company. Your own path into this came through a lived experience in the legal system where you self-represented. Take us back to that moment.
What happened, and how did representing yourself turn into building Caira and Unwilderd?
Hassan Luwalira: Yes. So I have two children, and I split with their mom in 2021. And like many parents out there, we needed the help of the court to decide how to move forward. And I decided to self-represent because I did not wanna spend 20 or $40,000 on legal advice especially on legal advice that leads to an outcome that could be achieved without spending that amount of money.
And what was really unnerving for me is that legal advice doesn’t even have a ceiling. So from your first consultation until the whole ordeal is over, it could be 20, it could be 60, it could be $80,000. And I didn’t really like the idea of spending $300 on an email reply . I simply didn’t want to. So I self-represented.
I was lucky enough to be self-employed at that time for the first time, and so I had a lot of time to spend on learning how the legal system works and reading legal acts and previous judgments, et cetera. And but after the legal proceedings concluded in 2023, I thought, “I’d like to help people.” I don’t want other people to spend life savings.
Like, I read about someone that spent 18 years’ worth of life savings on legal advice, and I didn’t believe that that was right. So I came across ChatGPT, and around then it was GPT-3.5. That was the frontier. And I thought, “Lawyers manipulate words, AI manipulate words.” So I created a simple prototype using a no-code platform.
I was probably spending, like, $20 per month or something. And I thought, “Let me put this live. Let me write a custom prompt that’s basically designed for law in UK, and see what people think.” And surprisingly, people used it. And I thought, “Wow, that’s a little bit of a surprise.” Because when there is an incumbent and you create a product, there’s only two things that really matter is, will customers notice a difference?
And also, how quickly will they notice the difference? And in this case, they were able to notice the difference really quickly but AI was a slur back then. If you spoke to 50 people on the street, hardly anyone would know what ChatGPT was, or if they had any experience with it, they would think it’s terrible, or they would just think that the idea of consulting AI instead of a human would be s- like, really stupid.
So I didn’t really speak to anyone about it. But then GPT-4 Turbo launched in October, and I switched to that mom- to that model, and it was like a light bulb moment. I was like, “Wow, we’ve gone from AI that can barely write a coherent paragraph to AI that can draft a letter or a statement that is serviceable, that you can take and send to someone.”
And I was like, “If models make this improvement every single year, then this could work out.”
Brian Thomas: That’s amazing. I love the backstory, always do here on the podcast. Your story’s amazing. Your, you self-represented in a legal situation. You, you thought, gosh, going through legal representation is gonna be too costly, and you shared some examples how costly it can be.
So you did all the research on your own, and then over time, as, as AI rolled out, you talked about ChatGPT but again, ChatGPT was very novice in its time, and it could make mistakes, and it was biased and that sort of thing. So you decided with your AI prototype, and as AI has gotten better through recursive learning, you’ve seen just an amazing platform now, and I just love that, so thank you.
Hassan, you’re clear that Caira is different from generic tools like ChatGPT. It’s backed by thousands of official legal documents, uses retrieval-augmented generation, RAG, right? You can read a user’s files and draft them from across nine jurisdictions. For someone who assumes they could just ask a free chatbot their legal question, what’s the real difference, and why does it matter so much in law specifically?
Hassan Luwalira: Yes. One of my popular sayings, which is not a criticism to anyone, is a fool with a tool is still a fool. When you use AI products, it’s really important to understand how they work, and especially for legal matters because legal is very nuanced. Like changing one word could swing the outcome of a case.
So it’s very important to understand how the AI products you use actually work. When you use general purpose tools, they perform a web search essentially, and then they use those sources to give you an output. So for legal services, this could be using sources such as blogs, forums, social media but those sources can have information that’s incorrect, out of date, biased Made up or not relevant to your jurisdiction.
And I think this is extremely important for people outside of the US because the majority of information on the web about law is US-specific. So, so, like, therefore, when you ask ChatGPT or Gemini or Claude a question, it might not actually have access to the latest or best information for your jurisdiction.
So where Caira differs is that she’s designed for law across nine jurisdiction. And she’s not only designed to give you a, a, a an answer that’s as accurate as possible, but she’s also designed to show you in the best light, because it’s not only about being right, but it’s about showing sometimes that you’re amicable or that you’re diplomatic, et cetera.
So when you ask Caira a question, she uses sources that’s been provided by ourselves. So for the UK, we have a knowledge base of legal documents just for the UK, so that is judgments and acts. And if you change the jurisdiction to the US or Singapore, then it uses sources for buckets that are specific to those countries.
So Kaera doesn’t have access to the web at all because that would literally put the, put the users at risk because we wouldn’t really be in control of the sources that she would use for the answers. And, and also our product is very much designed to be as consistent as possible because when you use the general purpose tools, the underlying model changes all the time.
Sometimes models are quantized or the retrieval augmented generation is changed. So you can ask Gemini, Claude or ChatGPT a question today and ask them the same questions in three weeks and in six months, and the answer could be drastically different. But with our product, we keep it very consistent so that you feel like no matter what time of year you use it, the answers are going to be as consistent as possible.
Brian Thomas: Amazing. Thank you. I, I liked how you highlighted a fool with a tool is still a fool. Pretty cool. But yeah, you must, you must understand the tool obviously because it’s, it’s gotta know local laws, the nuances around that, and you talked about how AI tools can be outdated, biased, not access to a particular jurisdiction, and you’ve c- you’ve solved all that with Caira.
You’ve designed it specifically for that local jurisdiction, and it doesn’t have a- access to the internet. Again, we talked about this. It, it, it’s a retrieval augmented generation, RAG which obviously is gonna keep, keep you in check. So I appreciate that. Hassan, law is high stakes, and you’re upfront that AI responses can contain mistakes, that Unwieldy isn’t a law firm.
How do you build genuine trust with people making serious legal decisions, and where do you draw the line on what Caira should and shouldn’t do on its own?
Hassan Luwalira: Law is very new to most people. It’s not something that’s taught in school. And one thing that surprises people is the English language. The English language that’s used in law is very different to the English language that we use every day.
And what our product tries to do is not just give a user outputs, but also help educate them on how the law works and how n- how nuanced it is. And and ’cause there’s, like, two challenges when you go through legal matters. The first one is intellectual, and the other one is emotional as well. So we try to help users with both, and helping them keep calm, but also teaching them that with legal matters, the answer is never just really simple and straightforward.
‘Cause we do find that general purpose tools are very quite prone to tunnel visioning. So that is when you ask them a question, and then they give a probable path, and then you just discuss that path, and they literally lead you down that path without perhaps asking about additional documents or evidence that could potentially create branches of three or five different paths.
So Caira is very much designed in a way that she helps you explore and understand all the paths that are available, but also caveat where necessary. Because one thing we try to avoid which can be prevalent in the general purpose tools, is not necessarily hallucination, but confirmation bias bias.
And this happens when a user has a question, but they almost have a preconceived answer in their mind, and they’re sometimes using AI to basically try to confirm that. But our product doesn’t really do that. Caira doesn’t do it. And even myself, I’ve used Caira a few times where I’ve accidentally almost been seeking confirmation bias, and she’s really pushed back and not given me the answer that I really wanted.
And we’re very proud of our users. One thing that’s really surprised me is how savvy they are. They don’t take AI’s output as face value. Some of them use multiple AI tools, where I’ve spoken to them over emails, and they’ve disclosed that they sometimes copy Caira ‘s answers into ChatGPT or Claude answers into Caira and vice versa.
So they are a lot savvier than I expected, and they are incredibly involved. Like, we have users who in three months have sent Caira 4,000 messages. They’ve uploaded 800 PDFs and 20 JPEG pictures. Like, they are really involved, and they’re getting to the nitty-ni- like, nitty-gritty in a way that’s really forensic, in a way that’s probably much more detail than they might have if they didn’t have AI, ’cause then they would have perhaps, been in the dark and essentially relied on legal professionals to do all the research for them.
But we’re also responsible and include disclaimers in our T&Cs and also in the chat below the input field. But also Caira, usually when she responds, she, she, she does discuss limits of her output if certain evidence or facts are missing, for example. And if external help is needed, she is often very good at signposting, saying, “Speak to X, Y, and Z to, inquire about this, and then bring that information back to me and we can continue.”
Brian Thomas: Amazing. Love that. Just to highlight a few things here, Hassan. Absolutely law is n- very nuanced. You talked about that. It’s not taught in school typically. And English law is very different and nuanced than our everyday English that we use on the street, right? So Caira is good at determining the, the nuances, the biases, and what I really thought was interesting is you have an amazing community, amazing engaged community which is really great to hear people that are actually involved and engaged and providing you feedback to improve your product.
That’s, that’s awesome. Hassan, last question of the day, you’ve engaged with UK’s AI Action Plan for Justice and argued that the legal aid system can’t keep up with rising demand just by adding budget. Looking maybe five, even 10 years out, how do you see AI reshaping access to justice? Will it replace lawyers, work alongside them, or something else?
And what has to happen for the future to be genuinely fair rather than leaving people behind?
Hassan Luwalira: I think if you look at technology, it often changes how we transform… If you look at technology, it changes how we interact, communicate, and transact. We, we, we have seen how technology has transformed financial services like banking and also media, and I, I expect the same transformation for legal services or knowledge work in general, especially since consumers are not just using my product to ask about legal matters, but they’re also asking about financial and health matters.
And traditionally, you would go to separate professionals for that, like a lawyer, financial advisor, or health professional. But now with AI uses like a single point of contr- of contact, so that means that regulators Need to rethink how regulation works. Like for me, for my business who’s going to have oversight of us?
Is it the Financial Conduct Authority? Is it the Solicitor regulatory, Regulatory Authority? I just don’t know. And but AI is going to solve access to justice. It’s going to make the best legal services available, accessible, and equitable to everybo- everybody globally. But courts and legal systems are going to have to change because we’re already starting to see the second order effect of AI in legal.
The first order effect is that consumers or businesses are able to pursue or inquire about more legal matters, but the second order effect is that courts, especially in the UK, are seeing a rise in legal complaints and applications being filed. And I don’t believe there’s ever going to be enough legal professionals and judges specifically to meet the demand for legal services or meet the demand of, like, the volume of work that needs to be reviewed.
Because any output that’s made by AI needs to be reviewed by a human, and there’s just not enough of them. So I think the legal systems will likely employ AI to help determine the courses of action for some Legal cases because there’s just simply not gonna be enough legal professionals to essentially review, like, the rising volume that’s out there.
Do I believe that people are exclusively going to use AI for legal services? No. I think there’s going to be a hybrid model where some exclusively use AI, some just use legal professionals, or some use a mix of both depending on where they are in the journey, whether it’s at discovery or at a final hearing, et cetera.
So I think, yeah, it’s really going to be a hybrid approach. Like a bit like home improvement. If you wanna renovate your house, you might do some bits yourself and other bits you might work with contractors, for example. But I believe that in 10 to 15 years, the legal system is going to look very different to what it does today, and I think it’s pretty admirable to see that the Ministry of Justice in the UK is at the forefront of recognizing the need for change and that AI is here to stay.
Like, the genie is out of the bottle, and going forward, businesses, consumers, they’re all going to use AI and that yeah, that the legal system needs to also change.
Brian Thomas: Thank you. Appreciate that. You’re absolutely right. Technology has transformed many industries, and you expect this transformation to, to happen further, obviously, in research and legal matters o- of course.
Now I, I thought it was interesting you mentioned AI’s gonna solve access to justice, but it’s also gonna change the entire law and courts the landscape there, how they’re gonna operate in the future. And of course today or in five years, there’s still gonna be an overwhelmingly amount of legal matters that need to be addressed.
And I liked your insights with the hybrid approach, keeping that human in the loop obviously, but we’re gonna definitely have to leverage more AI to keep up with all the AI output, so thank you. And Hassan, it was such a pleasure having you on today, and I look forward to speaking with you real soon.
Hassan Luwalira: Yes, thank you very much. Been a pleasure, and hopefully we’ll speak again.
Brian Thomas: Bye for now.
Hassan Luwalira Podcast Transcript. Listen to the audio on the guest’s Podcast Page.
