You’re Not Behind: My System For Leveraging AI In 2026

We built AI models that beat a lot of American competitors that had raised 4 or 5 or 6 times the amount of money that we had raised. We were able to make them work. We were able to get them into production, shipped, work, you know, in market and being successful, which is ultimately what we were required. I don't believe we would have been able to do that had we had the same cost model or the same hiring restrictions that some of the other startups were like following. Another big advantage is, I think in today's world more than in the past, when you work in a distributed way, you have to be better at asynchronous work. You do your work individually from me, you pass to me, I do my bit and pass it back to you. In the world of AI, whatever, I don't think you can give up relationships with people. It's so important to spend time nurturing those. And I think it's so valuable.

What are some of the things that people should be thinking about in terms of async that maybe because they're mostly office-based, they just haven't had to confront?

Yeah, well, I mean, I'm Nick, founder of Gitlaw, also an investor. I've sold a few companies, bought a few companies. Most recently, I founded Gitlaw, which is an AI agent That helps you create, draft, and review legal documents for free. Before that, Whisk, which is a company I took from zero, just myself, to 120 people, sold it to Samsung in 2019, stayed with the company for 7 years, which was a lot longer than I thought. Also did some angel investing. So invested in about 20 startups, mostly kind of deep tech, mostly in the UK, some in the US, LP in a couple of funds. Before that, founded a few other companies. Some of which I sold, some of the ones I bought, merged, different things like that. But basically tech, building tech companies. Fab.

Excellent. And what, Nick, what are some of the things that you believe to be true that maybe other people disagree with you on or vice versa? What is it you think other people are just mental to still believe?

Well, I believe that the value of AI is undervalued. I think People talk about that being overvalued, that it's all crazy. And I agree, the stock market is a bit hyped right now, but long-term, I think the majority of people undervalue AI and the impact it's going to have. I have set up all my companies recently as fully remote distributed teams. I believe that's the most effective way to scale teams fast. So not in an office. I did that in an extreme way at Whisk, you know, -8 to +8 GMT, 120 people, different cultures all around the world, all in one team. I thought that was actually really effective.

Okay.

Most people would say that's a bit of a crazy idea. I think async isn't like just a productivity hack that some people talk about. It's actually like, should be the default operating system for how people, how all high-performing companies work. So not working, you know, on the basis of having to do live calls, but working kind of in a way that you can asynchronously work on things in your own time.

Yeah.

And then one that's quite self-serving, I guess, for me, which I strongly believe, because I've changed my career recently about all around it, is that I believe Creating legal documents will be about 100 times cheaper and 10 times faster in the next 1 or 2 years because AI will automate it all.

Fantastic. There's loads to dig into here. And I'm fascinated by the— because my philosophy is the opposite end of the spectrum. 'Cause I don't think hybrid is a philosophy. I think you're either, either you get out of bed in the morning and you are a fully remote or you're a fully office person. And hybrid might be where you end up being for whatever reasons or compromises. But God, I can't, what's the name of the guy? I just have to pull it.

Steve Blank, that's it.

So Steve Blank, who is the—

Yeah, I know him well.

Yes.

And so he put out a blog post a couple of months ago, which was looking at performance of startups in the US saying that those that were office-based were growing 3 times faster than those that were fully remote. And so I went, yes, excellent. Here's data that validates my biased position. But let's dig into that. So why is fully remote better than— and have any of your businesses been office-based or have you always had this bias to remote and therefore you've built a remote business because of that bias? Or did you used to have something else and you had a Damascene moment?

So my first ever business was a coffee company, which sold physical coffee.

Tough to do fully remote.

Impossible. So entirely on location. Since then, I've been building tech companies. The first 2 were fully in office. Um, Whisk started fully in an office and we switched to remote. Uh, the reason I think the thing that drove us to it was hiring because we couldn't hire the right kind of specialized talent locally where we were.

Where had you put it?

Well, we were based in Birmingham, but included that net to London, right? And I think it comes a little bit down how much money you've got too. I think if you're like, you know, Google or Facebook or someone like that, and you're paying people top rates, that's, you probably can get the top 5% of talent. But if you are not, then you can't. And it's actually not even just about money. It's also about brand, right? The top 5% of people, in London will also want to go for a secure role often. So just the fact that they're at Google or Facebook or whatever makes them feel secure for their career. It's good for their CV. They're not going to get, unlikely to get fired, you know, et cetera. So if you're not a brand like that, or you don't have that capital, it's really hard to find the top 5%. And actually, I think in the past, having that top talent was super important in building technology. Today, I think it's actually even more important. in every field. Because essentially, you've got individuals running teams of AI agents that are helping them, that are doing the work of juniors. So you need people that are really smart, really experienced, and finding those in a small area, or asking people, or putting a requirement that people have to move into your small area to be able to do the job for you, feels like totally non-logical to me. So hiring globally, you, you know, literally this morning I put a new job advert out. I posted it 5 times in 5 different countries as a fully distributed role because I want to be able to make sure that the job description gets maximum view, you know, views across the different parts of the world. And I don't care where people live. They can live anywhere in the world. I do care a little bit about time zone. So they have to be within 3 to 4 hours of time zone. That is one thing that we were a little bit overly extreme on in Wisk. We allow people to join from anywhere. And the acquisition by Samsung with Korea and San Francisco meant that we had minus 8, plus say GMT, which, and also very, very different cultures. That was a little bit, maybe a little bit too far. But for like, it worked for Samsung because Samsung's a global company. But I think the optimal is like, you know, 3 to 4 hours time difference. Find the best people in Europe or in kind of roughly that time zone, which is a lot, a lot of people you can choose from.

And what's your, that top 5%, how much better do you think that top 5% are than average?

Well, I think it's more that it depends what you're talking about, but like in technology, if you choose the wrong architecture or you make the wrong decisions around how data is stored or which models you use, or it can have, it can be the difference between being successful and not being successful, like literally. And if you look at our team, what we did with the distributed team at Wisk, we built software that was, you know, build AI models before they were all so accessible in the cloud. This is back in 2012. You had to buy graphics cards and put them into racks and stuff. We built AI models that beat a lot of American competitors that had raised 4 or 5 or 6 times the amount of money that we had raised. We were able to make them work. We were able to get them into production, shipped, work, you know, in market and being successful, which is ultimately why we were acquired. I don't believe we would have been able to do that had we had the same cost model or the same hiring restrictions that some of the other startups were like following. So for us, it was like, you know, crucial to have that quality of talent. And I've experienced it so many times where not, you know, it's not Nick's inability to hire locally. Like I've had recruitment teams, like, you know, Samsung's recruitment teams looking for the best candidates in San Francisco and then comparing them to a global hiring pool and just the It just makes, there are downsides of distributed work. Absolutely. But there are really big advantages. Another big advantage is, I think in today's world, more than in the past, when you work in a distributed way, you have to be better at asynchronous work. You have to be able to have like, you do your work individually from me, you pass it to me, I do my bit and pass it back to you. Or like, we work obviously ideally through some kind of collaboration platform, like Google Docs, SharePoint, whatever. AI needs access for context to know what you're working on. What are you doing? What's the challenge? What's the progress? What's the history? So the more you write down, the more you kind of work in this async way, it creates amazing context for AI. Because essentially by working asynchronously, you're working, you're producing work and sharing your stuff in a way that helps other remote workers do, you know, consume and collaborate with you. And actually, that's exactly what an AI is. It's a remote worker that's working with you. And everybody who's talking about AI and how productive you can be with AI talks about how important it is to give AI context. If you don't give it context, just like a junior employee, if you don't give a junior employee any information about how your company works, it's not gonna be great, right? Well, they are not gonna be great.

So—

So you, by default, you become a, you default to have a writing culture in the business as a way of disseminating information. And then coincidentally, that has this massive additional byproduct and benefit in the age of AI.

Exactly. And I think we haven't seen that talked about much, but I do think like the default operating system of the future will be this. And if you don't work in this way, even if you're in an office, you're gonna have to start working in the same way that a distributed team works so that your AI has the right information. We're seeing a lot of it become a lot easier too with things like meeting note takers and stuff like that, right? So like even if you are having a physical in-person meeting, you can turn on, click one button and an AI will transcribe it all and kind of make it available for people who are not in the room or record it, right? So like there's, it's becoming easier, but that way of working, I think, will be the default operating system in the world.

I had Atakan on, who's the CEO of Jotform, and they deployed their own agent, they built an agent, deployed it themselves to start doing their level 1 tech customer support queries. And it could only answer 25% of the questions because the documents that it was attached. And so You just go, well, that means when you're training— so his thing was, which was really interesting, is that means when we're training humans, we have been crap at it because the training information only allowed the people that we trained to answer 25%. And we only knew that when we gave it to an AI.

Yeah.

And then we freed up our people to make it better. So now it's at 75%, 85% first time, first. But it means that they now have a much richer dataset than they ever had before. And they didn't even know how shit it was until They did it the other way around. Fascinating.

Yeah. And it works all the way through, right? It works the same way as like how you measure productivity. In the olden days, it was a lot of like, well, they turn up to work and they're friendly and they bring donuts and they make coffee, right? It's like, yeah, they're a good colleague, right? But like, were they doing good work? And actually in a distributed world, they can't bring donuts or make coffee or have a nice chat over the water cooler. Like they are, you judge them much more naturally on what they produce, the work they do. So, and that's obviously how you judge AI too, is the work. The AI is not bringing donuts, at least not yet. But so I think it's like the practices of async work or like distributed work actually help in-office cultures too, but you are able to circumvent them and be more lazy in how you run your physical office because you don't have, there are hacks, you know, like you can kind of, there's kind of proxies of like, I see a mat is desk the whole time, he must be working. And that's probably reasonable. Like, he probably is working. Like, you know, I don't see Facebook on his screen and therefore he's probably all right.

Yeah.

And it's probably reasonable. But in a distributed world, that doesn't work. You have to be like very focused on like, what are the outcomes? How—

yeah. What are some of the things that you do from an async perspective? Because we sort of talk— we're talking here about remote work and async and the 2 things mesh together. You can't— it sounds to me you can't have one without the other. So what are some of the things that people should be thinking about in terms of async that maybe because they're mostly office-based, they just haven't had to confront? And what is it? Are there tools? I mean, you've done this multiple times, so there must be a, like, you've got an ideal stack of tools that you recommend.

Yeah. Well, I mean, so, you know, some of them are not novel at all, right? So I mean, the main 2 tools I'm in most of the time is Slack and Notion, and it's just a project management tool essentially, or a tool where you can kind of like create lots of different flows within it. The main things are, again, not that unusual, but just lots of people don't do them very well. So it's like, make sure that projects are really well like structured and really well organized in a system. Make sure that all the tasks that link to the projects are well structured and documented well. Make sure you cover them regularly. I have a daily standup with the team, right? So every single day, the whole team is there for like 15 minutes to go through like, this is what I did yesterday, this is what I'm working on today, these are the things that are blocking me. And you can, and every single day, the whole team looks at a whole project plan for the whole company of like how everything is moving through the funnel. We automate a whole bunch of like things like when a user feedback comes in, rather than a human having to like, you know, classify it into the right thing in the system, AI does that. So you can start to automate parts of that flow, but ultimately just being really good at managing and sharing and communicating what everyone else is working on. Then there's stuff like when you create a project or when you want someone to work on something, rather than just go, let's have a meeting and talk about how we're going to make the website better, you write a spec for like, what's the objective? What are the key metrics we're going to move? Who's involved in this task? Who's the clear owner? Is there a single clear owner assigned? Obviously, you want to make sure that you're very clear with what the mission and vision of the company is. What are the OKRs or the objectives and the metrics we're after for this coming year and then this quarter and then this month and this individual? So just making sure that everything ladders together. These are not like genius concepts at all. They've been talked about in different formats for, I think anybody talking about how to run teams has talked about them, but it's possible to run companies very easily without them, without any of that, if you're physical in an office. As soon as you start to become distributed, it doesn't become optional. It becomes mandatory if you don't want the whole team to fall apart and be ineffective. So that is kind of the things you're doing are not, people listening won't be like, oh yeah, that's really unusual. It's like, yeah, obviously like that's important, but actually how many people do it, right? You have to do it when you are distributed.

Well, do you know what? One of the things, it makes me smile because one of the things I've said to clients, do a daily huddle, daily standup in every team, make it mandatory. And they're like, for some people, they go, I don't want to do it, but I did it and it's been transformative. I can't believe we didn't do it before and I can't believe how impactful it's been. Other people just go, sounds like a pain in the ass. But the people that do it go, I've got 4 hours a day back. There's so much email I don't need to read that just because it never gets— there's so much email I don't get sent. But that, I guess, is why you need that time zone overlap because otherwise, Korea, San Francisco, somebody's getting up at 2 o'clock in the morning to join that huddle.

Yeah. And there are ways of doing it. You can do async standups too, but they're not as effective. And like, yeah. So it sounds like you're actually advocating for a lot of async practices. You really just try it.

I think it's, you know what, it's funny because as I was listening to you talk, I was thinking of Kieran Cosgrove at Neoform. And when he took over there as CEO, he said, because he'd only run companies that had been in office and he took over a business that had been fully remote for a while. And he said, you know what, the biggest thing is that nobody comes here for the culture. That they might be working for us and then living in the Alps and skiing. And so the going out for a pint after work thing is just not a thing. It's like you just have to say it's a narrower focus. I'm doing interesting, meaningful work with amazing people, some of the best people I've ever worked with. And talking shit, I do outside of work.

Yeah.

And he just said, once you get your head around that, Then it's okay.

But there are lonely aspects of it too, right? Like you don't get to talk shit about people because you don't really go, hey, do you want to jump on a quick Zoom call to talk about nothing? To talk about nothing, right?

You don't.

So even though sometimes you try and manufacture, you know, I have in the past tried to manufacture those kinds of calls in a team to create that space, but it doesn't work brilliantly.

Do you do?

What I do do, Yeah.

No, I was going to, yeah, good. Tell me what you do do. Yeah.

Yeah.

I bring everyone into one place physically once a year. And I used to do it once every quarter, but now once a year is kind of, I've been doing for a long time. And even at Samsung, we used to do that. We maintained that. And when I first told them that, like, we're going to fly everyone to a location that from an Asian Korean perspective sounded like a holiday destination because they're like capitals of European cities usually, which, you know, from a Korean perspective is like, oh, that's, you're all going on holiday for a week. And I was like, no, no, no, it's not holiday. It's like, Yes, we do go for dinners and lunches and stuff together for a week. And it's an intense week of spending time together. But actually, we talk, we do a lot of workshops, we do a lot of like things that are productive. But the main objectives I always, the first thing I always do in those weeks is I stand up in the beginning and say, hey, look, the objective for this week is not, you don't have to produce anything useful. Like what you need to do is just spend time with each other. Like I want you to build relationships because those relationships that you do build when you're physically together, they will last and help you for the rest of the year, work better with people, have more empathy for what they're asking for, you know, give them that extra yard when they've messed up or whatever. So like, it's, I do think there is still, it's super important still to have like physical connection and like relationships with people are so important. And that's one thing that I don't think, you know, yeah, in the world of AI, whatever, I don't think you can give up relationships with people. It's so important to spend time nurturing those. I think it's so valuable. But yeah, I'm lucky. I've I've got a wife, I've got 2 kids, I've got friends. Like that, I do. Yeah. I get my interpersonal energy from them, not from work colleagues as much because like, yeah, no office.

Well, at Pier 1, we were an in-office culture, but because the executive team was spread across 5 different cities in 3 countries and 2 continents, we would spend a week living together for a quarter. We'd travel on a Monday, travel on a Friday. We'd have Tuesday, Wednesday, Thursday. We'd rent a house. We'd cook for each other. And it was the same thing. We don't necessarily need a business outcome at the end of this 3 days, but if we don't do this, the business outcomes will be so much harder to get done.

Yep.

Because it's dead easy to just misunderstand what people say when you're doing hard stuff. You know, it's easy to misunderstand an email.

Let me interrupt the conversation just for a minute because I don't want you to miss this. My new book, Mind Your Fucking Business, is out on Amazon. If you're thinking yet another business book, I think you're wrong. This is actually a blueprint to cutting through the noise and achieving unparalleled success for you. It's based on 10 years of business coaching, spotting similar problems that people have had and coming up with real-world solutions. If it's supercharging sales, if it's mastering talent acquisition, building a powerhouse brand, cultivating a company culture that breeds success, then this book is a ticket to your transformative journey. It's filled with practical tools and frameworks that I use every single day with my clients as part of their coaching program. If you want to take your business from good to great, look, you can go to Amazon and you can buy it, or you can click the link in the show notes and head over to monkhouseandcompany.com, and I'll give you a copy. I'll send it to you today. I'll send you a copy for free. I want to help you build a fantastic business. Now, back to the conversation.

Why do you think AI is undervalued?

Well, it starts with kind of the anecdotal stuff, which is, you know, as I listen to what businesses are doing with AI and I listen to kind of how they've implemented them and the challenges they've got on them, it's just like, this surface-level implementation, and they're really not digging at all into the depth of like what AI can actually do. Secondly, the quality of the models is improving so drastically. So like, you know, we're now seeing real competition between Gemini and OpenAI. We had before, you know, competition between OpenAI and Anthropic, but now we've got the third party, Google's really catching up too. So basically you've got these massive companies investing billions of dollars building models that are literally new releases coming out every 2, 3 months. Like it's every 6 months, there seems to be like major updates, but every, it's really, really, really fast. So today I'm seeing surface level use of AI. Like people aren't using it for nearly as far as much as they could do. I'm doing basic stuff with it sometimes in the business and people are like, wow, what? You're able to do that? And you're just like, yeah, like it's, this is like, This is like the surface level of what you can do with it. Um, you know, I'm building agents, so like I'm seeing the depths of like what they're capable of. I can see the gaps too. They're not perfect at all, but the models are getting so good. And then we can do this very rapidly.

Have you got agents replacing employees now? Have you got, you know, historically?

The business I've built is not, is quite new. So we haven't had to, we haven't fired people in return for an agent, like one-to-one, but I'm certainly hiring a lot less. If I, I'm moving at the speed of, you know, it's hard to say at the speed of because like there's a, well, so I think I'm probably, we're doing the work of maybe 50 people and we're like 15. But the problem is once you get to 50 people, the business slows down, right? Because you have people communication, which is like, you can't, with 50 people, you can't do the work of 50 people. You kind of almost need like 100 people to work at the, do the work of 50 people. So it's like, this is, You don't have any of those issues, right? You're in this like beautiful moment where your team can still know the names of everyone else and know what, yeah. So, yeah, like code, a lot of code is written by AI first. Code is reviewed by AI. Designs are built with AI. The first versions of designs are done and then a human just improves them. Content is written by AI and a human just improves them. So like it's, AI is doing first pass on so many things. AI is organizing user feedback, classifying into how important is it and, you know, like classic, even suggesting solutions for it. So like AI is doing so much, which is why you need really, really smart people to be in charge of AI. When all this stuff is coming at you, you have to be able to make decisions of like, yeah, the AI is right here. Nope, the AI is wrong here. And that often requires experience or really smart people to be able to make those judgments quickly, right? That's kind of what managers, if you like, I hate the word managers because managers You know, has a bad rep, but like managers of senior people in teams, that's what they do, right? They make judgments on like using their taste on like what's right, what's wrong. So yeah, that's why you need great people.

And legal docs, 10, 100 times faster and cheaper than today. So that's the business you're in now.

Yeah.

I don't know. Did you say what it's called? What's it called?

Git Law.

Git Law.

Yeah.

Sorry, you did say at the beginning. Yes. And for free, you say?

Yeah.

So our basic plan is free if you want to. Okay.

Yeah. So like, if you look at the benchmarks where people are testing these models against humans, the models are already at par or better than humans when you give them really specified tasks. The challenge is that obviously when a human's asking, it's obviously not very specified. So what we're doing essentially is orchestrating AI in a way that tries to understand what a human wants, uses all the context that it knows about the human, and then gives them, either writes the document or reviews the document for them. We will also, at some point, loop in lawyers to actually check the work for them and make sure that, yeah, we're happy with what the AI did here. So the human has the trust level because people do have a trust problem with AI, right?

Yeah.

And that's with the current level of models. Models are going to get better. And it'll get, um, uh, right now it's depending how you look at it, it's not, it's not as good as you went and hired a top-tier law firm. But the reality is, um, 90% of small businesses don't use a lawyer when they need one because they're too expensive or they're too slow. Uh, and I've experienced both of those situations, right? So they're, uh, legal services are not accessible to small businesses. And it's essentially an unfair, biased market. Only the richest companies can use lawyers to do one over on the smaller companies, right? I think it's actually a really exciting mission to equalize the playing field a little bit.

And so because you've got that vertical integration, you're putting some really tight guardrails on the model and you can then swap whatever model out at the back end?

Yeah.

Oh yeah.

We also have multiple models. Yeah.

And so would, do you have a, do you have a favorite? Are you, I was reading something the other day going, saying that some of the Chinese models are actually a third of the price of the US models. And so a number of startups, certainly in the US, are now using some of the Chinese models because it's cheaper. Are you still in, are you in that position or?

No. So we're not currently focused on price. Um, price isn't the problem for us, to be honest. Like, We, you know, it, we're not that cost sensitive. Um, so like the, because obviously, you know, on our platform, you get like, you know, we give everyone $5 free a month, which is enough to do a lot of work on the platform. But after $5, you have to put your own credit on, like put $20 on or whatever. It's still super cheap, but like, um, you have to put some money on. And we're not seeing people run into cost problems. So that's why we're not kind of solving the cost problem. Yeah, you can run DeepSeek or one of those models yourself. They're open source. You can tweak them. You can adjust them to your own needs. And yeah, I've definitely seen some people talk about the fact that they're getting even better results from these open source models that they've kind of like tuned to their own needs. But to be honest, the counterargument to that is these frontier models, frontier models being like OpenAI, Anthropic, and so OpenAI being ChatGPT and Anthropic being Claude and Gemini owned by Google. Those 3 models, they're getting, they're so good and they're getting so much better. You can basically just send different things to the different models that you know are good. And they're all different. They're good for different things. I personally prefer how ChatGPT talks. Claude, I think, is better for a lot of the actual, like, thinking work. Gemini, very recently, last 2 weeks or so, the new model they released is very, very good. So, you know, I think Sam Altman sent a couple of weeks, days ago, sent an email out to all of his employees going, you know, read Code red. Gemini is catching up, right? Or has caught up. And that's obviously scary from a ChatGPT perspective because Google are so good at scaling stuff, right? They weren't very fast to the races with the AI model, you know, but they're really good when it gets to like, let's keep, let's scale this. They've got the engine, they've got the customers, they've got the trust, et cetera. So like, I think, yeah, it's all, it's those 3 that we use. Um, and depending on what task you have, we use different tiers of the models. So either really fast because you want to, you know, edit the document title. You don't need a smart model for that. Look at the whole document and tell me what I need to edit. You need a much smarter model for that.

And then you've got a, on top of the paid tier, you've got the pay now to have a lawyer read it to make sure that you get a human, human check.

We will do soon.

Oh yeah. Okay. Right. That's coming.

Yeah. So basically you're just Because sometimes you don't trust AI. I try to understand why.

Yeah. And well, it's, but it's interesting, isn't it? Because, you know, 2,000 people die on the roads in the UK. We'll end up with, you know, Waymo all over the place. Waymo will kill somebody. Waymo will be involved in an accident when somebody dies and it'll be, the machines are killing people. It's like, but let's forget the 2,000 people who died before the machines arrived and maybe forget the 1,000 people that didn't die because the machines are here. It's just, it's a different level of trust. You know, it's like we expect the lawyers might make a mistake, but we want the machine to be 100% perfect.

Just, yeah, it really depends what you're using it for, right? And there's so many interesting, if you go into the mistakes of AI, like there's so much, so many interesting things have happened. Like, you know, Alexa, there's a kid, I think a 6-year-old, uh, said they wanted a doll's house, which Alexa sent them $120 or $600 or something doll's house. And they wanted some chocolate cookies and the Alexa sent it to them. And then a news anchor repeated the story and then thousands of of Alexas in the room with the TVs ordered the same thing. That's an example from a long time ago that was famous because it was funny. Less funny examples is when Alexa tells a kid to balance a pound coin or dollar coin in a socket as like a, give me a challenge to do, right? Which happened as well. Less funny is when a character AI psychologist tells someone to kill themselves or supports someone's suicide.

Yes.

Or in business, AI hiring, there's lots of people have talked about discrimination where, you know, it kind of downranks people of color or women or things like that. So there's like, it's so interesting. You have to be really careful with AI because like these are basically statistical models that take what's on the internet and, you know, kind of predict based on what's on the internet. And some of the stuff on the internet is not very nice. So you have to be, and some human behavior is stuff that we don't want to amplify. But AI doesn't know necessarily what to amplify or not. It hasn't got a character or personality. It's just a statistical model. So yeah, you have to be careful with AI for good. And also ChatGPT is way overconfident, right? So ChatGPT, you'll ask it something like, is this contract okay? And it will go, yeah, it's great. I've read it all. Dominic, don't worry, you're good, right? And then you kind of look at it and tell it a mistake that's like, you know, a third of the way down the document. And it goes, yeah, sorry, I thought I missed that. You're right, Dominic. Well done. And very positive to you.

You're telling you that you're fantastic.

And you're like, but why did you miss it? And then you dig in and it will say, oh, I only read, only skim read it, Dominic. I didn't actually read it. So this happens often, right? These AI models present wrong results very confidently. And because of the confidence they present wrong results with, it erodes the trust. But I think actually, in some cases, it's fair to erode the trust, but actually there are parts, there are ways of orchestrating AI that you can trust. But it's going to take some time for us to get there. And therefore having like some trust loop where it sends it to a real lawyer to actually do the validation is helpful.

And do you know what? I came to the, I was in India last week and I was trying to find my bike in Delhi airport. And I came to the conclusion that ChatGPT was actually an Indian because whenever I asked anybody, they confidently told me where my, where I could go and get my bike. And when I went back to them 20 minutes later, they admitted they had no fucking idea where my bike was. And they were very sorry and I had caught them out. And it's like talking to ChatGPT, but it's people. This is the most bizarre thing in the world.

Exactly.

So in my mind, I then had all of these Indian-origin programmers in San Francisco tapping away at keyboards and ChatGPT comes to fruition. Nick, What books have you read along the way that have made a difference for you that you think people should pick up?

I would say the main books are actually not ones that you might expect me to talk about around like, you might say to me, oh, well, AI or asynchronous work or the problem is those topics evolve so quickly that Actually, I think books are not even worth reading on that stuff. You should read blogs or podcasts or try it out. The stuff that I read, and I read quite a lot, probably half an hour to an hour a day at least. My recent read, which I really enjoyed, was Poor Charlie's Almanac, which is like Charlie Munger, Warren Buffett's business partner. And he basically talks about sort of timeless fundamentals that he's learned, sort of mental models of incentives, human psychology a lot, which I thought was really good. And it's like those kind of fundamental truths of how humans work rather than like, you know, what's happening in technology right now, because I think those fundamental truths will help us, help all of us understand each other and work and build good businesses. Second one will probably be 4,000 Weeks by Olivia Burkeman. Um, I enjoyed that because like it reminds me that we all have 4,000 weeks and every week is ticking and it's, you know, it basically gives a little bit of urgency and also a little bit of like time to reflect and go, whoa, how am I spending my 4,000 weeks? It's not all in work, you know, family, friends and stuff like that are really important too. Um, and then I think my fourth one is probably, yeah, How to Win Friends and Influence People by Dale Carnegie. Again, a really old one, right? But like, it's basic stuff, super basic, like show a genuine interest in people, avoid criticizing and blaming people, like just really basic things. But actually, you know, you could read it and go, well, that's really basic stuff. But actually, when you read it, it kind of does sink in and it kind of like, and the stories, the way he tells about it, it sticks with you. And it's so true. And actually a lot of it's very similar to what actually Charlie Munger talks about too. It's like, it's a basic human truths of how we should interact with each other. Actually, I think if you want to be successful in life, or if you want to enjoy your life, or if you want to like be a good husband or a good wife or a good parent, like these are all just, those are the things that actually I think are, I enjoy like reading in books. But then when it comes to like the latest technology, Lenny's podcast, which is a podcast around technology and AI, and he has all the leaders of like OpenAI and Anthropic and all the kind of latest, you know, frontier models and people working with those. I find that super helpful to help me learn about what's happening in technology.

Yeah, fab. That's brilliant. I just, I've got a sort of perennial question. It's interesting, that whole get 50 to 100 and everything slows down. Absolutely. I spend time with people in that no man's land challenge every day, but I'm intrigued when you built Whisk, And you went from 0 to 100. How many people on your original leadership team were still there with you at the end?

Actually, so many. Like I've probably had—

How did you, how did you manage to pull that off? Because that's tough.

Well, so I didn't hire external managers in. I didn't go and like hire a new leader for like technology and sort of parachute a CTO in or anything. What I basically did is I've promoted from within. And remember that within is like people I hire from all over the world, right? So those individuals, some of the people in my team have been with me for like 15 years and have actually followed me along the different businesses that I've built, right?

Yeah.

So in fact, I would say every single person, so the head of design at Samsung Food has been there for, I think, 15 years. Head of operations is like finance, legal operations has been there for probably 8, 9 years. The head of technology, probably 6 or 7 years. And before that, 10 years, you know, head of product's been there probably 6 years. And before that, it was me who's been there for a long time as well. So like, it's really like people, it's all, I've kind of just naturally always gone to promoting people rather than bringing people in. And just, I don't know, I actually think it's more effective, but I haven't tried the parachuting thing. I have, I've had it done on me. Uh, where I've, you know, when we were acquired into Samsung and people parachuted people above me, and that, it didn't feel great. Um, and it actually wasn't effective. So I guess maybe I just, you know, I don't know. It's interesting.

So you're, well, you're, it's interesting because it's, it's linked back to something you said earlier, which is because you're fully remote and you're hiring the best people you can afford, you know, you're taking, you're taking a chunk of money. You know, if you'd use that money to hire somebody in Birmingham, right, you know, you might've got to the point where that person was okay at 30, was no good at 50 and terrible at 70. But because you've got, somebody over here who is, you know, accessible to you at a price. You're getting people who have faster personal growth rates and they're staying ahead of the growth of the company.

And also I think you get to shape them, right? Like how you as a founder, it sounds quite self-centered, I guess, but you shape people when they join a company early, how you want to work and, you know, or how the company works. And then, so then they become a leader and they're already like fully on board with the company culture, the history, the people, they know everything. It's not like, it's not something new. And the problem I think I've noticed with parachuting people in is that often they come with a different philosophy. And sometimes that's really helpful, right? I imagine. But sometimes it also can be quite jarring for people. And then like, yeah, I think you, I don't know. I think that the context that people have from the tenure length that they had and have, I think is super helpful. I find it necessary to be productive.

Nick, that's absolutely brilliant. Thank you for coming on the show. And thank you for sharing these great stories and your time with us today.

I really enjoyed it. Thanks for having me.

Next time.

AI won't replace most leadership jobs, but it will expose those who are not pulling their weight by adding real value. And if they don't know how to use AI, they are going to fall. There's too many people spending hours trying to put the data together just to put into a of report, valuable that it is, and it's right to do, but we can automate that, saving a leader probably 2 or 3 or 4 hours a week. Now all you need is somebody to oversee the agent that produces that for you and to tick off and validate. Yeah, but I don't need lots of humans to do that. That is saving so many man hours.

At a leadership team level, I'm intrigued. Are you working with some clients at the minute on some experiments where You are making the leaders more efficient.

The first thing is—