Lessons learned on the road to becoming a CEO

In this episode of the Tech Story podcast, Charles interviews Brian Mullins, CEO of Mind Foundry, a company that develops AI for high stakes applications. Brian talks about the company's background, his career in technology, and how he became involved with Mind Foundry. They also discuss the impact of AI on the machine learning market, specifically mentioning the recent popularity of chat GPT.

‍

‍

Watch the episode

‍

Read the transcript

00:00 Charles Brecque Welcome to the Tech Story podcast, a place where we interview founders, CEOs and entrepreneurs in tech who are building really exciting businesses. Today I'm thrilled to welcome Brian Mullins on the show. Brian is the CEO of Mind Foundry, a spin out from the University of Oxford developing AI for high stakes applications. Brian, thank you for taking the time for being with us today. Please can you share a bit of background about yourself and Mind Foundry? Yeah, thanks Charles.

00:30 Brian Mullins It's great to be here. I am the CEO of Mind Foundry. Mind Foundry, spun out of the University of Oxford, is just over six years old. I've only been with the company for four years, originally from the United States. I'm actually a graduate of the US Merchant Marine Academy and I've been in technology and venture-backed technology for most of my career. After having been involved with hardware and software companies and some successes and some failures, I heard about an opportunity here in Oxford, a company founded by Professor Steve Roberts and Professor Mike Osborne, both who have phenomenal reputations in AI. We're doing some really exciting work on bringing artificial intelligence to high stakes applications where decisions affect people's lives or are made at the scale of populations. I knew both of them from their work, but when I had a chance to meet them and found out

01:39 Charles Brecque they were looking for a CEO, it was very easy to make that decision. Yeah, I can imagine. Yes, Steve and Mike are world leaders in AI. I guess AI has become a lot more popular recently with chat GPT. Is chat GPT responsible? And from your perspective, how has this helped the machine learning market? How does this affect mind foundry?

02:05 Brian Mullins Great question. I'm not sure if I would say any particular model is responsible or not responsible. Maybe there are some that are trained intentionally on a pretty badly curated data set, looking for a bad outcome. But I think, generally speaking, models are tools. It's how you use them and how you understand what the data represents, what the method that is used to create the model is capable of, and how those things affect what the outcome is and then based on what that recommendation is or what that decision is, what takes place in the real world. What's the intervention? What's the direction of the intervention? And if you can assess all of those things, I think then you can start to understand if you can use that model responsibly for a specific application. And in a lot of cases, you probably don't need to do that. If I'm trying to figure out the best time of day to send a marketing email, if it gets it wrong, the implications are probably not that profound. If you're making a decision that directly affects someone's life, their access to some critical services or some opportunity, then I think based upon what those stakes are, what the impact is, you have to put more consideration into how you use these technologies. That being said, chat GPT is exciting. These large models that offer the opportunity to take learning from one area and do multiple things with it. It's not general artificial intelligence, but generalizable intelligences. It's a new area of opportunity. There'll be many, many implications of it, many, many things that can be done with it. I think that's phenomenal. I think we're all still learning what those things are going to be, what the impact will be. We're very optimistic and I think equally concerned that they're used responsibly. And I guess, does that mean that a company like Mind Foundry can help chat GPT become responsible or be used responsibly? We tend to think of, it would be the second part. We work directly with organisations. Our largest area is in insurance, government infrastructure, so helping governments to make decisions about things like roads or railroads or bridges. And then in the area of security and defense, where certainly the stakes are the highest and using AI responsibly is critical for the safety of people's lives directly. We work with the organisations that are interested in using AI and help them with our platform to use them responsibly and to monitor their use in a way so that they can ensure that

05:21 Charles Brecque is happening on an ongoing basis. That's great. I'm sure the impact is huge, especially given that it affects lives and high stakes. Since being a founder and then CEO now at Mind Foundry, what's been your favourite moment

05:40 Brian Mullins so far? My favourite moment ever as a CEO or since the first time I was a CEO or founder. You can share both moments. Yeah, so, I mean, there's a lot really. I really like working with people. I'd say that the thing that I'm most proud of is when people that I've had the chance to work with and mentor go on to start their own companies. I think the idea that we could help people to reach their potential is the job of the CEO, whether it's in their current role, their future role in an organisation, or whether they go on in the future in their career to start something new. That's the thing that really gets me excited about is helping people to achieve their own goals and reach their potential. Then there's a million things in between. I love technology. It's easy with technology to be focused on tech for tech's sake. I've made something that's cool and the idea of it is really cool. But it's even more powerful when it solves a problem in the real world, when you see somebody who has a problem or a group of people. Then that technology can actually make a difference in what they do every day and then have a big impact on the real world. I think that in a lot of cases, we miss that most important step, which is let's build something cool with them. Let's make sure that it gets used in the real world to solve a real problem. I think most of the time, and I've been fortunate to see lots of these moments, I get really

07:23 Charles Brecque excited when that happens. I can definitely relate to when you solve a problem with technology and it's technology that you've developed. It's a very unique feeling. I can also vouch to when we worked together and I wanted to start legislate, you were very encouraging. That was also one of my favourite moments at Mind Foundry because now that I've built a team and obviously seen people go on and do their own thing, it's obviously very conflicting. You want people to grow with you, but I thank you a lot for being so supportive.

07:59 Brian Mullins You did everything on your own. It was great, the chance that we got to work together and super excited to see you grow and legislate and building a company. It's pretty amazing.

08:09 Charles Brecque Yeah. Well, thank you. What do you wish you had known before being a founder or CEO?

08:17 Brian Mullins What I wish I'd known before being a CEO, I think early on, maybe some of my previous answers reflect what I've seen as my own shortcomings in my career. Early on, I was excited about technology so much that I thought that was the goal. Let's make a thing. If you build it, they will come. Later on, I realized that organizations that are more focused around product actually get to solving the problem. I had to make a whole lot of mistakes before I figured that out. I think the reason why is that the product philosophy and process gets you closer to interacting with the customers, the users, the people that have the problem. It really informs your thinking about what you're building and how you create it and what the objective is. But then after years of thinking that, oh, this is so much more successful, this is fantastic, I also learned, and later really learned, that the things that would still hold an organisation back are not empowering people. The truth is, I think that a lot of the strength of the product process and the product focus in an organisation is because it's focusing on people in very specific places. If you take a step back, you realize that it's actually when you're focusing on people and empowering people to solve the problems that that's when you have the most success. It took me a lot of mistakes, I think, along the way to figure that out and a lot of good advice from people along the way. I wish I'd known that earlier on, but maybe you have to go through those mistakes, maybe

10:13 Charles Brecque you have to learn by doing to really appreciate it. Product is really hard and something that we always take for granted. I think especially when you're at the top with the vision, you just expect things to happen and build themselves. It always takes longer, but it's also an area where maybe it might seem expensive to invest

10:35 Brian Mullins resources, but ultimately it does pay off. Absolutely. You make a good point about how hard it is being at the top, trying to create a vision. Your job as a CEO is to create a vision. It's to build the team, both with new people with new abilities, but also the team that you have. Then it's to drive revenue. Revenue is the fuel that you need to accomplish your mission. You've got to do all of those things, but you have to do it knowing that when you're at the top, it's like being on top of an iceberg. You just can't see what's below the surface. As that iceberg gets bigger and bigger and bigger, your blindness to what is below is less and less and less. If you're not willing to stick your head in the water and really see the shape of things,

11:32 Charles Brecque you're likely to have a lot of problems really helping to do your critical role as a leader. I think the analogy of the iceberg is really true and something which you have to go down, but also come up. Great. Great. Now, Mind Foundry is a six-year-old company.

11:59 Brian Mullins What's the vision for the next five, six years? We're focused on the high-stakes application. That market represents about 20% of AI overall, but it's the things where off the shelf won't do. Most of the way that AI tools are used and deployed today are through existing teams in enterprise or solution providers or in features and products that people use. Those tools solve a lot of problems, maybe 80% of problems, but when the stakes get high, things need to be different, not just the consideration for how to use it, but the technology, the models, the interpretability, the ability to operate within regulatory frameworks and compliance and to be able to validate that you are maintaining those requirements in operation over time as the world changes, it becomes critical. If you make a mistake there, the results are disastrous. When we really see the work that we're doing, we're hopeful that we can have a positive effect on some of these key industries. The more that people approach AI in this way, whether they're working at the algorithmic level like we are or whether they're implementing things that are off the shelf but in a more responsible and curated way, we're hopeful that then the trend will be more responsible across these high stakes applications. We're learning so much from our customers, the importance of the areas where we're already developing traction to society. They're there in this unintuitive way that we don't think about as individuals, but if you take a step back and look at the market, they're critical. Think about insurance. Most of us, maybe we get car insurance because we have to. Where I grew up, you need health insurance to get access to basic health care. That seems very important as an individual, but we don't really understand the scope of it. Insurance represents almost 9% of GDP worldwide. Just the significance of that alone makes it pretty critical. The services that it provides are so critical. Accessing insurance products is paramount for all kinds of communities that we know can be disenfranchised by AI or algorithmic decision making. Making sure that the products are able to understand risk without being exclusionary is really key. The opportunity is that insurance can be a profitable business and provide these services to customers even though there is risk associated with it. You don't have to remove risk. You just need to better understand it. That's something that AI does really well. The opportunity is to make these products available for more people to provide this critical layer of infrastructure that really works across economy. It's not just the products that we see, life and health and car insurance, specialty insurances, insurances that businesses have, essentially keep the economy running smoothly. Take the idea of delivery vehicles like trucks. Most large cities rely on truck-based delivery. They don't manufacture foods locally. They don't have the goods that they need to operate a city. If a truck had an accident, and most trucks are owned by small and medium businesses, sometimes owner-operators, they don't have the resources to repair a truck that has damage in an unforeseen accident. If they had to wait until their normal business operations allowed them to do that, then you take some of the capacity away. Market forces would then make a higher demand. Some people would go get more trucks, and they would make up for it. Those take a long time. If in that time, another truck had an accident, then you can have this cascading problem where which goods don't get delivered to that city. Then how do the prices change, or how do the fluctuations in availability affect the way that people get their food or their medicine or any number of things that they rely on? Insurance as a layer allows that to run smoothly even when there's unforeseen problems. If we can build AI that helps with that robustness, people may not notice, but I think it's pretty important work. You're the hero that no one knows of, but it is ultimately keeping the insurance companies running. I love that idea. I think my team would say that the companies, the insurance companies, the governments that are looking to use AI responsibly, those are the heroes. A lot of people want to adopt the technology quickly because they see the opportunity, you save money or improve outcomes, but they're not thinking about doing it responsibly. No, it's not maliciously. They're just not thinking about it. The ones who do, the ones that slow down and say, hey, let's do this the right way,

17:52 Charles Brecque I think those are the unsung heroes today in high stakes applications. Absolutely. I think we take for granted that AI models need data. If there's an under-representation of data, especially in populations, then that means that if you don't apply the AI responsibly, then you can ultimately risk excluding those populations. That's right. That's really fascinating. Best of luck doing that. As a CEO, I imagine you must have come across quite a few contracts.

18:27 Brian Mullins What are the key legal documents that you look at and what can you share about them? From our perspective, the most important contracts, in tech everyone understands you're doing NDAs regularly. There's a certain number of simple documents like that that allow you to have the conversations and interact with customers, partners, and other organisations. But then I think there's two other critical areas of contracts. One, employment contracts. Because of the complexity, especially if you work across jurisdictions, the requirements there, this is a high-stake application if you think about it from a contracting perspective. On the other side of that contract, on one side it's the company and people operations. They're trying to do their best. They're working across jurisdictions. Maybe they don't know. They're trying to navigate on the fly. But on the other side is a person. That is that person's livelihood. Their family relies on what they do with their livelihood. The stakes get really high and the experience has to be good and simple and able to be understood. I'd say employment contracts are probably one of the most critical ones. Especially in competitive markets for talent, you have to move quickly with those. Simple and speed and timing are everything. The last one that I would say on the most important list would be the commercial contracts because it's the same way. When there's an opportunity to solve a problem there and an organisation has the bandwidth to focus on that problem, if I can't get a contract that's simple and easy to execute and get it signed, we can't move forward. Another priority could come up that maybe this is still absolutely critical, but they don't have the time to get through it. That process has to be simple. It has to be adaptable to the fact that when you deal with a large customer, a large insurance company, a large government, there's the big organisation and the small organisation. You have to be able to have the flexibility to work there and they have to be clear and

20:50 Charles Brecque concise and also accurate. Great. I absolutely agree on employment contracts. It's really, really high stakes and someone's life. Obviously, half our team is based in Spain when we set up a branch and they're now all on employment contracts. The difference in obviously holidays, but also rules around notice periods in the UK, there's a rule of thumb around what's reasonable and what isn't. Whereas in Spain, there are actual trade unions that have set for this specific industry, this specific role, the notice period has to be X. That's something which we've had to learn. Of course, there are ways of ensuring that the company is protected or that there's alignment between jurisdictions, but it's just complexity that you can't just sign a document and expect it to work.

21:46 Brian Mullins It needs to be done properly. Regulations change. Those change and you need to know immediately which of your contracts are affected by the

21:53 Charles Brecque change and where do you need to make an update. That's exactly what we're trying to do at Legislate. Great. I guess with these legal documents, are there any common areas of friction that you've encountered? If you were giving a tip or piece of advice on how to overcome legal to a current or prospective

22:17 Brian Mullins founder, what would it be? I think two different questions. One, the friction in legal generally and then contract-based frictions. I think part of the problem is the friction in the legal itself because there's slightly different motivations. The legal practice, smart people, good people, their goal is to help you understand all the detail, protect all the possibilities, even if they're very remote possibilities. Then on the other side, you need to get your business done. Those don't always line up in a way that favours speed and readability. If I look at the other side, the frictions around contracts themselves, it's speed and readability are what you need the most out of your contracts, especially if they're with an individual who maybe they have a lawyer look at it or someone in the family who's a lawyer look at it. Generally speaking, a lot of what they have to understand from the commercial impact, their own employment impact, would be something that they're reading themselves. I'm always striving to simplify contract language as much as possible, but then still it then account for the fact that when you do business with enterprise customers, you have to use the right language the right way. Ultimately, if you can transact quickly and cleanly, but still also protect everyone in the eventual, hopefully not eventuality that there's some kind of litigation or dispute involved, that there's enough clarity in the contract from a letter and a spirit to help people understand what's meant and what's binding.

24:17 Charles Brecque Yeah. Well, I guess a key summary is make sure you use the right contracts or the right language, which means seeking legal advice. Because ultimately contracts only revisit it when something goes wrong and you want

24:30 Brian Mullins to make sure that you've got the right framework to protect yourself. I know everybody wants to get that deal closed, but if you've got a term in there that's wrong, it will come back to haunt you and it won't be worth trying to gloss over that detail

24:47 Charles Brecque on the front end. Yeah. I've heard lots of horror stories of founders who signed a contract too quickly and then couldn't terminate the agreement and ended up not being able to fulfil the contract or having a supplier who's sort of letting them down and then they just end up stuck and sometimes it does hurt the business.

25:08 Brian Mullins But great. I got some great advice once earlier on in my career to mentor, tell me even if you're small and nobody will use your commercial paper, like nobody wants to use your contracts, you've got to end up kind of responding to their contracts, make your own paper, make your own contracts. Eventually you'll be able to use yours in a standardised fashion and by making them in advance you start to shape how you do your deals and you have them and you have the time to make them right. And so don't think that you'll get the right contract for you in the eight hours before the end of the month trying to close the deal with that customer that you really, really want because you're probably not going to be thinking clearly about what you need to

25:54 Charles Brecque do to protect yourself. And if you have your own contract then you know at least what to look out for. That's right. In the client's contract. Great piece of advice.

26:05 Brian Mullins And if you were to receive a contract assigned today, what would impress you? Simplicity. If you can say what you mean in simple terms, you know they still have to be legally robust but in simple terms I'm always impressed. I think a lot of people will use really legacy templates and maybe it's not because of them but it's because the company has always used them or their legal counsel that they've worked with has been using them for the last hundred years. What's wrong with that? I think that it says a lot about an organisation's desire to get things done and to really have mutually beneficial exchange when they have simple to understand contracts and they don't

26:52 Charles Brecque have 200 pages of things hidden in the fine print. That reminds me of the saying that you'd always say to me is the burden of communication is on the communicator and ultimately you need to communicate simply and if it's not understood then it's your fault and I think the same applies to contracts. It needs to be simply drafted so that it can be understood.

27:15 Brian Mullins 100%. I mean if you're trying to get away with something in the fine print, your motivations are probably not here.

27:22 Charles Brecque Absolutely. Since we're on the Tech Story podcast, what's your favourite tech product? My favourite tech product of all time? It doesn't have to be Legislate.

27:36 Brian Mullins It can be anything. I work with AI and there's a lot of things that I'm really excited about and things that are probably my current favorite products. I'm not going to say them though. I'll actually talk about earlier in my career, I worked in augmented reality and I still am really excited by technologies in VR and AR that help you to both interact in digital worlds in new ways but also then to bring ideas in the form of digital content into the real world, creating an interface out of the world, telling stories in the environment that you're at. I still get really excited about that future. It's still years away but I can imagine wearing a pair of glasses and being in my living room and seeing a character walk through the door in my living room and the story plays out there and they sit down in one of my chairs and have a conversation with me. It's kind of the same magic of immersive theatre where you get to interact with characters and each time the show happens it's slightly different based on what the audience does. It's this kind of ultra personalised experience but can be scaled and can be in a very personal context of where you live, where you're at.

29:04 Charles Brecque I think AR is pretty amazing. Have you heard of a company called Humane? Rumours, but I don't think I know anything about what they're doing. So they're a Silicon Valley company. I think it is one of the teams that created the iPhone. There's been a lot of mystery around what they've been doing and building but yesterday they demoed a version of the product where it's a little device that clips onto you and projects your phone onto your hand so that you can take calls, speak and I guess they're sort of, I don't know if you'd probably be able to say if it counts as AR or not.

29:47 Brian Mullins Yeah no it definitely counts as AR. I think whether I'm wearing glasses or I have a projected interface, ultimately there's kind of seamless move between those two. If I'm wearing a pair of glasses on the street and I get into a car, the car is likely to have an AR interface. I'm going to want to take the glasses off. I think a lot of people who wear glasses spend a lot of money to get corrective surgery so they don't have to wear glasses. You don't want them on in all environments. I love the idea of wearable projectors or projectors in an environment and interfaces on devices.

30:23 Charles Brecque I think it all counts as AR. It's interesting to see what they've built and I guess, I don't know when it will be available to the public but I guess it's something interesting to look out for. I'll have to take a look at the talk. I know you love gadgets and devices. I do. Guilty. Great. Well thanks a lot Brian for taking the time to be on the Tech Story podcast and best of luck growing your mind, Andrew. Thanks Charles, this was great.

‍