Speaker 1 (00:00):
So Nicola, thanks so much for coming on The Unlock. Before we go into your story, just add some context around what PolyAI actually is.

Speaker 2 (00:08):
We build AI agents for customer service. We help some of the largest companies in the world be the best version of themselves in every customer interaction. So pick up every phone call, reach millions of customers with an urgent piece of information, and just show up in the best way possible. So we work with, I think everyone from a FedEx to large utility companies, banks, insurers, a lot of hospitality. I think we're in one of two casinos in Vegas, so all over really.

Speaker 1 (00:36):
So talk to me about what happens. I call up a hotel, I book a room, I'm speaking to one of your agents.

Speaker 2 (00:41):
Yeah. I mean, you've probably spoken to one of our agents or many of our agents without knowing, especially here in the UK, from just outside Hawksmoor is a client to really half the restaurant groups, but when you call up, we'll have our agents picking up the phone to, if you call a Metro Bank or NatWest, or you might get a call from one of them. It could be that you call anything from an NSNI all the way to through hospitality, through big utilities where maybe you have a billing issue, maybe you're calling to check where your order is with FedEx or to reschedule a missed delivery. And you call, an agent picks up, it may or may not tell you that it's human, that's up to the customer. You have that conversation and that's it, really simple.

Speaker 1 (01:29):
And is there a lag between the conversation when it picks up the phone and goes, "Hello, FedEx, how can I help?" Is there then a lag? Does it have fake typing in the background? How quick is the response rate? Because that's sometimes infuriating for people that are calling up agents.

Speaker 2 (01:43):
Oh, I mean, I think what's most infuriating to people is not being understood, especially in the UK, you've got 63 dialects and if you're Glaswegian and trying to speak to one of these systems, you're not having a good time.

Speaker 1 (01:56):
Can anyone understand a Glaswegian, humans included?

Speaker 2 (01:59):
I once left a Polish dude in a contact centre in Glasgow. I don't know where he is, but no, I find it to be one of the most charming accents. I'm calling

Speaker 1 (02:10):
Glasgow.

Speaker 2 (02:10):
Really? I

Speaker 1 (02:11):
Just don't sound it.

Speaker 2 (02:12):
True story. It's a great accent. But no, no, no. Look, I think that latency design, the look and feel, really the feel, not the look of a voice agent is really important. So latency is a performance word where our models right now are. We have smaller reasoning models. If you think of Opus versus Sonet, Opus is way better, but it's a lot slower. So our dialogue reason one model does just enough reasoning to be clever enough to do what it needs to do in a phone call like that, but respond in under 300 milliseconds. So then you add a bit of phone lag and all of that. It still ends up being well under a second, which is kind of the mental barrier to. You and I are not really responding to each other in less than a second. If we did, we'd be overly keen.

(03:03):
You'd be uncomfortable with how I'm communicating with you if I was that poised to respond. And it's funny, we used to think it mattered more than we do today. It definitely matters if it's really bad. You get people bolting OpenAI models on and then using different stacks and they get five seconds of latency.That's unacceptable. People will opt out of that conversation really quickly. But if you make it too fast and too snappy, that also has a negative impact because people are like, "This is not a human."

(03:33):
Like that uncanny valley feel where you're like, "Something's wrong. It's way too primed. This is not a human." And when you explain why you think it's not a human, that's where performance is the case the most because you're just giving me a human. And there's a lot of people who are not that into AI these days. I think it's easy to be in this bubble and be like, "AI is great and whatnot," but there's a lot of anti-AI sentiment as well.

Speaker 1 (03:58):
Yeah, for sure. And we'll go into maybe some of the structure later and how you maybe started super early comparatively to the masses. But you said on a podcast recently that you had something to prove you from a generation that had something to prove. Have you actually managed to prove it yet and what are you trying to actually prove?

Speaker 2 (04:14):
I don't know. I think that was in the context someone asked me about Serbia and how that generation of Serbs and all former Yugoslavia was impacted by it because I think that the 90s are, for most Europeans or Americans, a period of just unfettered growth, things going well. Whereas I think we were the part of Europe that collapsed in a civil war at that point. So I think made for quite a different upbringing where institutions collapse, countries collapse, probably about 10% of the population is unclear on what citizenship they have. So I think at that point, just that fall from. And Yugoslavia was really well-functioning. It was more like Scandinavian social democracy with a lie dictatorship on top during communist times. So that flip from being the absolutely most developed part of Eastern Europe with the most liberal values, Coca-Cola ads everywhere, to just turning it into the part that fell economically and not just quality of life, safety, that quickly I think it created for a very different outlook in life.

(05:31):
I think that my generation of Serbs, they kind of want everything and they want it now. It's not really a mentality of even proving. It's more a mentality of just breaking out one way or another. So I think it makes for a lot of ambitious people. I think you're going to see a lot come out of that part of Europe just because I think that good traumas make for the best motivators, long-term motivators in life. And

Speaker 1 (05:56):
There was a lot of trauma in Serbia back then.

Speaker 2 (05:58):
People

Speaker 1 (05:59):
Were losing their jobs hand over fist, companies were going under. You started a company, which is essentially pushing people out of jobs.

Speaker 2 (06:06):
You think?

Speaker 1 (06:07):
There's nobody that's going to be needed to answer phones. You're replacing these individuals.

Speaker 2 (06:12):
Have you been to a contact centre?

Speaker 1 (06:13):
No, not

Speaker 2 (06:14):
Specifically. Okay. Glasgow. There was year where I went to Glasgow eight times in contact centres. The attrition in half of them was over 200% every year. So the average tenure of an agent, especially outside of the top 10% of management and senior agents were paid more and like their jobs. Their jobs are very stable if they want to stay there. It filters for people who have incredibly high EQ and resilience. So it's not sufficient just DEQ. It's like you have to be able to take abuse and just shrug it off. Very stiff upper lip mentality helps, but it's not for everyone. So those jobs are safe, very safe, before and after AI. Because what you have to understand is most companies for about a decade, if not longer, have just deflected contact. Deflection is the word used in the contact centre industry to talk about I moved them online.

(07:11):
And really it's just a euphemism for I've told you to disappear and all but those terms. I put you in a self-service labyrinth loop or I've asked you 10 questions only to tell you that you can do that online and you maybe can't even do that online, but maybe I'm okay losing you as a customer or maybe I don't even know that I'm losing you as a customer. Shouldn't assume bad intent. But the general state of disrepair in customer service and the entirety of the Western world is humongous. So we've artificially depressed the ability to pick up phone calls and service customers in the channel of their choice for way too long. Now you've got about 2% of the British population working in contact centres and probably about as many people serving the British market outside of the UK as well in offshore contact centres.

(08:05):
And you're still waiting for every other thing you call for. So the demand is absolutely humongous. And when you open up channel of choice and just do voice for everything, you probably need double the people. So the people don't want to do those jobs because however the market settled, they're being paid minimum wage to do a job where someone's just throwing verbal abuse at you. Sure, it's directed at the company that is not providing internet when they need to work from home and be on a Zoom call where they have to present or whatever their package didn't arrive and they're going on a trip and they need that whatever something. But they're yelling at you and that's why people don't really want to stay in those jobs.

(08:47):
I think that when you look at it as a zero sum game globally, I think there will be maybe some jobs that are lost, especially lower end of the BPO market is getting hit already, but it's very unclear what is driving that because in customer experience, both in America and in Britain, you've had these waves of onshoring and then offshore. And people would always prefer to hire Brits to service that demand, but can you pay Brits enough for them to stay in the job long enough to become good at it? Yeah, you can. You just need to pay more and provide a good working environment. I think as companies are able to do that and then you get AI doing much of the mundane work, you're really getting this air traffic control, like AI copilots of these agents, fleets of agents really, that are servicing that demand.

(09:42):
And then that tier two agent that used to have their time just interviewing people and worrying about whether they're going to have enough butts and seats come next Monday, they're now like, okay, I have agents. They're failing 7% of the time on this and 5% on that. I need to call someone and figure out what we're going to say in this situation and that situation and what a refund policy is for that. Well, that's a PM's job if you think about it. At the moment. Whatever a PM means. No, but the business of this is heavy. It's very deep. That I promise you because that's all I've done for years now. Getting into it and understanding what a company needs to do to actually automate a very large percentage of calls.

(10:23):
My joke is I was a researcher once upon a time, then I kind of became an aggressive software salesman. And then really I've graduated into that code that I took over there that I took off. Now I'm an Accenture consultant telling you how to rewire incentive plans for your middle management so that AI has a shot at doing better for your business because really it's about that. If you're not serious about rejigging your objectives so that AI can help you, then status quo is a wonderful thing. You just declare that waiting times of 10 minutes are okay instead of five minutes and you've solved the problem. And by the way, that's what many people do today as well. So it's not really putting people out of work where they need those jobs. And maybe I'm just coping, maybe there is some of this, but the people I know, and we, by virtue of starting earlier, have had some of these customers for five years now.

(11:18):
I know them. I know their first tier of employees. I call them to try new features and I trust them because honestly at this point they're pretty wedded to us in terms of they've invested a lot of their iteration loops with us to make our product better. There are people who used to be contact centre managers that have now presented to boards of multinational financial institutions because they're now AI transformation directors and talking about institutions collapsing. I remember

(11:50):
Post Milosevic fell from power in Serbia and my mom worked at the National Bank of Serbia. She was in this department that had 40 people. She was one of two that made it over the next decade and stayed in that job and it was very stressful for her. So I think about that a lot as I go through what we go through. I think it's impacted also the verticals and industries that we've gone after. We've gone after a lot of hospitality, a lot of places where missed calls are missed revenue and then people just won't pick up those phone calls and that's kind of institutional. Think of restaurants, hotels. They miss 30% of the calls because receptionists are asked to pick up the phone call, but also to clean up and to check people in and then other things and it's all hustle and great, but they can't pick up five phone calls at a time.

(12:40):
We've done a lot in those places where it's just not replacing jobs. But coming back to jobs changing, my mom stayed in that job because she was willing to go through learning how to use software and everything that a lot of other people just opted out of. She was resilient. Yeah. Well, she had to be, and I think she was, but it's really like with AI now, just at a much greater scale today than with the previous waves of use computers, use this software as opposed to the previous version of it. That's happening everywhere. And honestly, I'm very happy that it's happening to engineers first and then to lawyers seem to be on the chopping block, Harvey and Ligori as well. Accountants. But they're not really on the chopping block, if you see what I mean. I think that what we might get is just way more patent trolls in America suing people with AI tools and then you need AI tools to defend that it'll be a blob like cybersecurity.

(13:40):
So it's really hard to predict where the economy will go and just where we'll channel our efforts and is it productive or not? I don't really know in the case of - But

Speaker 1 (13:48):
You can predict that it will definitely start to churn people's roles within reason for low value tasks like a call centre or for low value tasks in a legal firm, for instance. You can predict that's going to happen. Do you think there's a thing with ethical AI? Do you think you're running a ethically AI orientated business? Is that a thing?

Speaker 2 (14:06):
No. No, I don't think it is. I think there's a lot of copium out there. I think if we completely just go like 30,000 feet, you could make the case and I think you wouldn't be wrong if you could enforce it to pause AI for a while. I think that would probably be a sound approach for humanity.

Speaker 1 (14:28):
Yeah, because it wasn't Fable one recently that got really caught out of hand with Claude and that was quite dangerous actually.

Speaker 2 (14:34):
And now it's like I use my Fable budget every week on my max account and then I spend another thousand bucks every week on it. And it's phenomenal. It is dangerous. Exactly. It could be, yeah. But if everyone has it, it's kind of like guns in America, right? Yeah. If everyone has them, they're not as dangerous because if someone starts shooting, everyone has a gun.

Speaker 1 (14:54):
Well, one could argue that it's still pretty dangerous.

Speaker 2 (14:56):
It's very dangerous. It's very dangerous. Britain is a lovely island which doesn't have that problem and it shouldn't go that way. Albeit

Speaker 1 (15:04):
You're moving to California soon.

Speaker 2 (15:06):
California is pretty anti. But no, I mean, look, I think that ethically, I think there is a capitalist zero rush, which we're part of, that doesn't really care and is not really thinking about any of it. And then it spends 5% of its effort on some copium around safety and whatever. I mean, Anthropic is, I think, the butt of that joke at this point where I genuinely truly believe in Dario's honest intentions starting this. It's gone out of hell. I saw a meme this morning and it was like Dario's shocked that people are joining Anthropic to make money. And the top comment below was, wait till he hears what motivates his investors.

(15:46):
I think it's just. Is it just a good marketing strategy? The whole doomerism. I'm very optimistic about what it'll do for us, but I do think we are going through already the very early days of a period that will probably be marked with serious revolutions, upheaval. There will be winners and there will be losers like in any industrial revolution. And that's nothing to scoff at. I think if politically we can find a way to soften the blow for people impacted, we absolutely should. But yeah, I think companies, we are a very small part of the space. So me talking about it or pretending that I am able to redirect the forces that move this wave of automation would be illusions of grandeur that aren't even I'm capable of.

Speaker 1 (16:37):
Because I mean, you have competitors, right? Sierra doing 100 million ARR, you've got US rivals that are doing four times your actual revenue. You guys are valued at 750 million quick, give or take, right? So businesses in 2030, 2031 with these valuations will be disappearing, but people will pump capital into it. What makes an AI business stable, sustainable, scalable, and profitable? Because so many of them aren't.

Speaker 2 (17:00):
Yeah. It's hard to go through the noise of the different companies in this space. I think what's not really questionable is that we all have product market fit and the world growing really fast and that there's a lot of appetite for this. I think a lot of companies are bucketed into the same approach. So for example, Sierra, right? That's a competitor. Then you've got many others. I think that the way I think about the whole thing is we're more of a technology company than they are. They're like a massive consulting business pulling people in from the McKinsey Salesforce. They're pioneering this outcome based pricing. We've done that in large deals as well, and it can make sense if you have the right audience, if you have the CEOs and CFOs, COOs listening and they can align with you at the very top level around what a good outcome is.

(17:53):
I think for us, we're more of a tech company where the goal is to just give people a platform that allows them to gain massive leverage for their business using our technology. To us, it's like we have our own models. We are able to price them competitively so that we make a very healthy margin on the platform and the tooling we provide to them. We are able to get them going and live really fast. Then it's really on them to work on making their business talk to their customers with AI and become capable of automating a very large number of those calls, chat interactions, et cetera. But the top end of the market is these forward deployed motions where there are entire teams working on behalf of the customer. And that's what we did first and we like that. But I think that over time we really have become much more passionate about giving those tools and creating that category of.

(18:52):
You know how Salesforce has sales ops people that live inside those companies, have access to everything in that company to create those dashboards and things. Similarly, to us, it's really about creating this vanguard of AI transformers in the contact centre that are going to take the ownership of our platform and create this AI effication of their contact centre and broader than the contact centre because we do a lot of sales, we do a lot of marketing, lead follow up. It becomes a whole customer funnel for both support and sales and ongoing relationship with the customer.

Speaker 1 (19:29):
I mean, in terms of the model that you own, you build upon, et cetera, it's changing at a rate of knots. Every 18 months you're stripping something out, you're having to rebuild the whole business. That costly, right? But arguably, how do you know what the right thing to do at that point in time is? And is that hedge the correct hedge to do? I mean, you've done it recently. In July, you uprooted everything and had to rebuild the whole pipeline.

Speaker 2 (19:51):
Well, I mean, what do you mean?

Speaker 1 (19:53):
In terms of your tech stack, you had to rebuild.

Speaker 2 (19:55):
Oh, that's not just July. I mean, that's been happening consistently, right? So when you say July, what do you mean?

Speaker 1 (20:00):
Well, I've just got it in my notes.

Speaker 2 (20:02):
Oh, the release of model. Yeah. I mean, that's just the public release of the model where others can use it. So we now have competitors, frenemies using our model. Well, yeah. I mean, you live in the age of hyperscalers and big platforms, contact centre platforms that want to sell their own package products. So you know how good car companies are good watch companies sell their own movements or good car companies sell their own engines. I think the Rolls Royce engine is famous on its own outside of Rolls Royce, the premium car, right?

Speaker 1 (20:33):
Well, Trump specifically specked his aircraft with Rolls-Royce engines for that reason.

Speaker 2 (20:36):
Hey, that man knows quality. So I think in the great lineage of British businesses that went on technology and that flavour of design, so do we. We worked on this stuff, but no one worked on this stuff. I mean, I was at the open AI launch party and I think probably the only person there with my co-founder who cared about this stuff back then. So we've worked on this in our team, our senior researchers. We've got more than a decade of work on deep learning

(21:04):
For conversational AI. And that just means that we're part of the Vanguard that is at the frontier innovating for that front. Now it's also like what data you have. And while Google of course has more data than PolyAI, they don't have more data of the kind that you need to advance this. No. Because human computer interactions that are really good are few and far between, and we've done more of them than anyone else. So when you take that data and the model and the infrastructure changes, then you take that innovation, you go back to your engine and you go like, all right, we're now doing V12 engines. Now we're doing turbocharged V6s. Okay. How do we do that? Let's think. BMW is not part of F1. I don't think anyone really doubts that BMW could be an F1 and do really well. So there really the question is just what matters at what point?

(21:57):
But we are very much like a tech company, an R&D company first, not a AI consulting business.

Speaker 1 (22:05):
Yeah. But you're very much B2B fundamentally. Oh yeah, we are. Of course. Our clients are

Speaker 2 (22:09):
Enterprise.

Speaker 1 (22:09):
100%. And I find it very interesting how you guys spent millions of pounds on a Super Bowl advert, which fine, there's millions and millions of people watching it, but a B2B business wouldn't necessarily advertise potentially on the Super Bowl because there's so many consumers that are watching it. Beer brands, food brands, that kind of thing you'd think would be more akin to the. Why did you guys

Speaker 2 (22:29):
Decide to - Oh, you missed the whole memo. I mean, the coolest thing that you can do is like - The coolest thing.

Speaker 1 (22:34):
Yeah. Well, I mean - Was it a vanity project?

Speaker 2 (22:37):
Everything's a vanity project, but Trump's a vanity project and he's doing his second term. I do a third, right? Questionable. Yeah. Look, the way you project power in AI is by having mind share. Yeah. And you're expected to do that. And you're expected to show the world that you have the best thing through technology, but also through though leadership, creativity. When I started the company, I started it because I got fed up with people telling me the serious shit. I was at Apple for two years after they bought a previous company. And where do you work, Apple? What do you work on Siri? Sirius shit. Hello? Nice to meet you too. Why is it bad? What do you hate about Siri? And people would be like, "Well, it doesn't work. It doesn't do this." And they're valid concerns like you set an alarm for 7:30 and it didn't set an alarm for 7:30.

(23:32):
That's okay. That's an objective failure, right? But if you go like, "Well, I wanted to plan a holiday for my family." And I'm like, "You don't have authorization from your wife to plan the family holidays. So that's not a problem with Siri. That's just you being negative because you watch too many sci-fi films." So I just got fed up with people hating on it. I was like, "Where can I build a smaller thing that's going to be really good?"

Speaker 1 (24:00):
So that's the reason you fell out of

Speaker 2 (24:01):
Love with

Speaker 1 (24:02):
Apple?

Speaker 2 (24:02):
No, I love Apple. I think Apple is a phenomenal company. By the way, I think Apple called it perfectly because we were acquired in Cambridge. It's not the largest machine learning office that Apple has outside of America. And Apple is a phenomenal company with a very specific culture. And this was vocal IQ, just to be clear. Yeah. I was the first employee. I wasn't a co-founder, but I was there from the very start. It was a really fun journey. I think it gave me the startup bug, and I suspect I wouldn't do all that well in a big company if I had to do it now. I love the ownership, but the interesting bit was Apple understood that people don't buy iPhones over Android phones because of Siri. That hasn't changed. I remember Alexa hired 14,000 people, including I think more than a thousand in Cambridge down the road.

(24:57):
We were like, "What is Apple doing? They're squandering a first mover advantage."

(25:03):
I think they called it perfectly. They were like, "This won't make a difference." And it didn't. No. So Bezos blew probably 10 billion plus on Alexa to help you buy toilet paper with your voice, and you never even bought the toilet paper. So Apple called it well. So I'm not out of love with Apple. I am very much in love with Apple, and I think it's a phenomenal company, one of the best in the world. It has its culture and it's very specific, but I think if you want to do incredible work and you don't want to be loud about it, Apple's the best company in the world to work for.

Speaker 1 (25:34):
So is that one of the reasons then that you started PolyAi, because you could see all of that going on at Apple at the time?

Speaker 2 (25:40):
It just wasn't key, right? Whatever you want to work on, if you really want to work on that, if it's not one of the two, three most important things for that company, you're not really going to have a great time because you're working on a side project. So you got to take that thing and make it your main project. And what is PolyAi known for? Branded voice experiences that are the best in the world, that are used by Gordon Ramsey, that are used by Caesars, that are used by the Rosewood, by a number of five star properties and Michelin star restaurants around the world. And that to me was always the key thing. If they're willing to use it, they're not using it to save a bit of money. They make enough money and the experience is holy. It's sacred. I think Gordon, to me, he's always had a very therapeutic effect on me.

(26:29):
Why is that? I don't know. It tells you something about my personality, but he is known for not suffering any deviation from perfection, and that's it. That's Polly AI. Th's our brand in a nutshell. He's also a British thing that became bigger in America than arguably, and that's very Polly AI. 80% American revenue. I think most of our American clients have to think, oh yeah, yeah, I guess British kind of. I mean, half our employees are in America, so we are very much like an Anglo-American company where you kind of start work late morning London time, and then you work till close of business, California, which isn't really an advert for quality of life. But hey, if you want to work on something important and exciting, it's the place to be.

Speaker 1 (27:18):
I guess in terms of working on something important and exciting then, and if you are based in the UK and you work in the US Times, from a cultural perspective, how do you instil that in your team to work hard, work late, make a difference? What do you have?

Speaker 2 (27:31):
You can't instil it, man. You can't instil any of this. You've

Speaker 1 (27:34):
Got 280 staff.

Speaker 2 (27:35):
Yeah, you can't instil anything. People either want to work that hard or they don't. A company is a mesh of people that grows from a core of founders, then this layer of early employees who are basically co-founders, who've been with us for seven, eight years, and they're all like family to me. If something's failing somewhere for any one of those people and they call any other person, it's like calling their sibling. It will be done. That person will do it for that person because they'll care about it and they're in it together. That's what a company is. I'm not saying this is right. I'm not saying this is good. It's just how it is and how PolyAI is. I think you'll have circular groups of people who then kind of join in waves and those that make it and stay are people who adapt to the culture and they also influence the culture.

(28:29):
To me, it's very much if I see some of our sales and channel guys taking someone out and then three engineers end up having beers with them from 2:00 to 4:0 PM and I'm in front of the office, probably on a phone call running back to the office from a meeting, I'm like, "Oh, lucky." But I love that. That to me is the same hooliganism that you get as a thank you for like, "Oh, there was a thing at 10:00 PM you had to be on the call for and you were there slaving away till 1:00 AM." You just have to build a culture that works for you at the end of the day. I think the notion that some people are out having beers at 2:00 PM is very not 996 American, but also the 996 is such trivial hours for someone working in London because if you're serving America, you don't get to clock off at 9: PM.

(29:19):
I always though the Californians were very cute with their hours and stuff when we all know that 6:00 PM in San Francisco, it's a complete wipe out and they've all had the time to commute home as well. I think that we all take the liberty of adopting myself into your nation here, but Britain doesn't brag. No. It's anticultural. That's why I think many of us fell in love with Britain, but you have to force yourself into it a bit because Americans make a big deal out of, "I had breakfast." I think here people are like, "All right. Yeah, I had to work a bit late. How late for work? 1:30 AM. But the American culture is so

Speaker 1 (30:01):
Different period from raising capital as well and failing companies. So you guys are running at a loss, you've raised a lot of money granted, but you don't really shout about that. But in America, that's okay. But in the UK, if you failed, people don't really like that. It's much harder to raise money over here. Why do you think the culture is so different in the UK from a failure perspective and a capital raising perspective to what it is in the US? When you rightly pointed out they are more brash and outlandish and conversational.

Speaker 2 (30:27):
They're great. I love America. Professionally, I think America's the best place in the world to work. They're entrepreneurial. They take big swings, they take risks. A lot of the American bravado is bravado, but a lot of it is true. And what should be on a shadow of a doubt is that they take those risks and they celebrate those winners. And that just people get bumped up and they do it again and again and again. And it's just like if someone took a massive run at something, they probably could have made that company profitable earlier and sold it to private equity. They could have. They learned something unique in trying to do that crazy scaling before. If they try it again, they are the ones most likely to succeed because they will have known where they missed up last time. They could fail again, but then that's where you have different categories of capital, investing in venture capital, then you have different forms like growth equity, private equity, et cetera.

(31:24):
So you choose your game and I think that what is true beyond shadow of that is they have far more venture capital, truly venture capital. I mean, Bezos talks about it a lot where it's like someone will just throw 50, 100, 200 million into something that's a complete moonshot. We don't have that kind of money here.

Speaker 1 (31:41):
At all.

Speaker 2 (31:42):
We don't, we're poor. I think we have to acknowledge that we're poor. We're very poor. I think that the poorest US state still, I think at this point has higher average earnings than the UK. And isn't that a change? And I think we have to just internalise that. I think it's easier for us Eastern Europeans to acknowledge it because we were always the poorest part of Europe. And I think we've just learned that. It actually rhymes with British culture, that whole self isolation and deprecation. So I think while the loudness is different between Cisrabs and Brits, the sense of humour is the same. So I think that self-deprecating bit, it serves you well if you channel it the right way because then you're the underdog and then you're the most dangerous. And by the way, Americans love that. They do love that. They love underdogs.

(32:27):
So I think that apropos the whole raising money, why is it like this or that here or there? You know what else American VCs love more than anything? The European summer and London. They love London. Honestly, if you have a British accent, you will just be deemed to have IQ 20 points higher than your actual IQ, no matter where it is. I don't think it's ever

Speaker 1 (32:48):
Worked for me.

Speaker 2 (32:49):
Sure?

Speaker 1 (32:50):
Don't know.

Speaker 2 (32:52):
Got to try more. But look, I think that they love funding British companies. They love spending time here. You can raise money here, no problem. Have you ever failed? I fail all the time. What do you fail at?

Speaker 1 (33:04):
What do you want to know about it? What's your biggest failure? One that you couldn't turn back from that you thought it was all over?

Speaker 2 (33:12):
I mean, we've had a lot of. I think the first seven figure deal that we were supposed to close, I think went into London, did a photo shoot with BT leadership and a contract that was in the works for about a year. It was supposed to be our first seven figure contract, just disappeared into thin air because commit to hyperscaler. And we've seen a lot of that over the years, but that one was a pretty tough bunch. And then you just realise, well, it's the name of the game. So you'll get -

Speaker 1 (33:47):
When was that in the journey?

Speaker 2 (33:48):
Very early on. And you'd raised capital prior

Speaker 1 (33:50):
To that?

Speaker 2 (33:51):
Yeah, we did. We did. Yeah.

Speaker 1 (33:53):
Did it hit you hard, that failure, or was it because you'd raised capital that you had enough money in the bank from a burn rate perspective to go, "That's cool, we'll go again."

Speaker 2 (34:00):
Well, I think it kind of got me really to understand the flows of money and the distribution of software, which is very different now. I don't think enough people talk about it. It's not a game where the best technology wins. It's the best distribution plus technology and people buy at different levels and it's very complicated. And with AI, it's going to get even more complicated. But yeah, I don't know if that answers your question around failures because that was just a setback. In B2B, you're going to get these lumpy deals and things that just some years you're a hero and you think you can do no wrong. Some years a few of these fall through because death by hyperscaler or something like that. At that point it's very humbling and you have to go and rethink your strategy. But Altman had, I think, the best advice in his YC days, which is don't die.

(34:50):
You just can't die as a company. I think that really comes back to the core of the company and the mission and are you having fun? It's not just the mission. The mission is part of it, but it's also the double flywheel of that is the people working on that mission together. Are they having fun? Is this exciting? Is it cool? Are you excited to work every day on the thing that you're working on? I think that those two things together multiply into are you going to survive or not? It's never really about the money. There's always money because if you're excited enough, you can tell the story. If you can tell the story, you can raise money. So you can't that, you're invulnerable. Elon has proven it over and over again. Sure.

Speaker 1 (35:31):
Elon is -

Speaker 2 (35:31):
No one really thinks that he is. I mean, he is incredibly smart, but he's not cracking the launch of SpaceX rockets himself, right?

Speaker 1 (35:42):
Well, one could argue that he's been heavily involved in the process.

Speaker 2 (35:45):
Oh, I think he's the energy -

Speaker 1 (35:47):
Above and beyond most CEOs and founders potentially at

Speaker 2 (35:49):
That. 100%. No, no. He's a legend, right? But it's like the energy, his belief is what takes that flywheel of motivation of that whole team to persevere and find a way. That is what he injects. That's what makes him special. So I don't think we were ever, ever in existential risk because we care and we're good at what we do. And there's a sense of almost like a promised land and destiny with PolyAI where this is our turf.

Speaker 1 (36:17):
I think that's the same with all AI businesses. No, no

Speaker 2 (36:19):
It's not.

Speaker 1 (36:20):
Is there not this future kind of we'll get there eventually, but we don't know when. When will you guys get to where you want to get to and where do you want to get to? Is it an IPO? Is it profitable? Is it an exit? What is it?

Speaker 2 (36:30):
Those are all just words. They mean nothing. So

Speaker 1 (36:33):
What do you want to do with PolyAI? Where do you want to get it

Speaker 2 (36:36):
To? I just want us to create the future of this human-machine interaction. It's about much more than even customer service. Customer service is just the largest surface where you get to do it, but I'm as excited about that as I am about an application of this to absolutely everyone everywhere. We might be living in a world where you no longer have keyboards and that tablet is like. And isn't that going to be a funny thing where we remember all of us typing away at laptops and things as the hack mouse? Well, mouse is kind of gone. Some people still use them, but it's rare. And I think that we'll get to different interfaces for communicating with these machines. I think we're evolving as a species, right? Yeah. And I think that that's just part of this journey where we're trying to be a big part of it.

Speaker 1 (37:27):
Because you were so early on in this world of voice, essentially, after you left Vocal IQ and you started this company. Where is Voice going now? For instance, I'm in my workflow using WhisperFlow, you might heard of Wispaflow is unbelievable, so much better than any other sort of voice detect platform I've used. I've also, and actually I think it is potentially a Serbian company, but Victor AI. I don't know if you've heard of Victor. I

Speaker 2 (37:47):
Think they're Polish. Are they Polish?

Speaker 1 (37:49):
That's unbelievable from an agent perspective, pumping that into Slack. And those are two key things that I'm using in my workflow consistently, which has sped it up and made it easier. Voice being a main part of that. Where do you see the industry from a voice perspective or indeed a workflow perspective going for somebody that is still in a job?

Speaker 2 (38:06):
Yeah. I think one thing that's crazy if you just enumerated it and look to the world in workflows, there's so much human power workflow that shouldn't be there. For example, we work with companies that distribute medical devices for sleep apnea in America. I think we probably have more than half that market powered by PolyAI at this point. And there are about five calls that happen between someone calling for a new device, then them calling the insurer, that process getting resolved, calling them back, delivering the device, on average five phone calls every month for just a new CPAP pillow and whatever else is needed for that device. So you look at those workflows, they can be better. I think that AI is creating a more convenient life for people doing that. Or for medical transport, we I think deal with about 20% of US non-urgent medical transport.

(39:01):
So people going for oncology treatments, for dialysis, for high grade medical care tasks. Their life is better if they can just deal with calling a phone number the way they always have, but instead of some clunky contact centre experience where people are finding their ride or whatever, you know who they are, you can suggest things, you can get them booked and reassured in 20 seconds so that they just go on with their day instead of them learning to use this app. Because apps themselves are gooey as well, the same way that a keyboard is. So I think in the future, technology will probably play a bigger part in our lives, but less visible. Why are we all transporting this hardware around with us? I think that we're still some bits away from. I mean, the phone seems to be like that carrier device, but it's still clunky.

(39:56):
You can't really use Whisper Flow in the shoot. No, no you can't. In theory you could, but it's really good. It's incredible. I think they've done a really good job. But even then, I find it funny, right? I'm a very senior tech leader who's calling me a Luddite for not agreeing that keyboards will be gone in two years. And I'm like, "I've given my life to voice. I don't deserve this treatment." But I saw what he meant, right? So

Speaker 1 (40:20):
What's going to replace the keyboard? Literally voice control?

Speaker 2 (40:23):
I don't think so.

Speaker 1 (40:23):
Eyes? Gaze?

Speaker 2 (40:26):
I think gaze is a huge part. I think once that's cracked, we're in business.

Speaker 1 (40:30):
Well, going back to Musk, I guess with Neuralink, that's

Speaker 2 (40:32):
A

Speaker 1 (40:32):
Huge part of that.

Speaker 2 (40:33):
Neuralink is the thing.

Speaker 1 (40:34):
Okay.

Speaker 2 (40:34):
Neuralink is the thing.

Speaker 1 (40:35):
But how will people feel about having a chip literally planted in their head?

Speaker 2 (40:39):
Wire me up.

Speaker 1 (40:40):
Wire you up?

Speaker 2 (40:41):
Please. Why would you not?

Speaker 1 (40:43):
Well, I mean I - How

Speaker 2 (40:44):
Would people feel about carrying potentially radiating device in their pockets, putting it next to their heads and their pillows while they're sleeping?

Speaker 1 (40:53):
I think there's a psychological difference between the two because you're holding a

Speaker 2 (40:57):
Device which

Speaker 1 (40:58):
You think you can get rid of. But if you've

Speaker 2 (41:00):
Got a

Speaker 1 (41:00):
Chip in your head.

Speaker 2 (41:02):
Okay, let's make it remotable.

Speaker 1 (41:03):
Still, you've got something in your body. I'm not sure. How do you think the market will react to that?

Speaker 2 (41:07):
Since my first kid was born and my wife kindly on maternity allowed me to sleep. I put in AirPods and I often wake up and there's at least an AirPod in my ear. I'm very happy it's there because if both are not there, then I wake up more and this just makes me sound like a terrible father. I was going to say. I do weekends. I do weekends. I do a lot. Two weekends. No, no. I do a lot, but this is our maternity, not when we're both working. By then kids sleep through the night, but different podcast topic. I don't know. I think it's just like.

Speaker 1 (41:43):
And you think

Speaker 2 (41:43):
In two

Speaker 1 (41:43):
Years keyboards will be gone yearly?

Speaker 2 (41:45):
No, I don't think keyboards are gone in two years. I think they're kind of gone. Not gone. They'll be used less. I don't think we're there yet. I think we got to crack the UX of these things a bit more. It's not there just yet, but with Neuralink, I think they're beyond gone. I think that computers could be gone because I think. Well, I mean, Zach bet on AR early and he wasn't right for now, but that piece is the obvious one. Why do we have all thesere? With the metaverse. Yeah. Why do we have all these screens and hardware and stuff? It's obviously not optimal, right?

Speaker 1 (42:21):
Yeah. I think some of the hardware is too big anyway. You look at Oculus, for instance. Yes, fine, you can have virtual screens in front of you, but you look like an idiot wearing that in a cafe. Do you know what I mean? Until you've got glasses that look like glasses, then it might work.

Speaker 2 (42:35):
But even then, you'll look at someone with them and be like, "Are they talking to me? Are they watching YouTube?"

Speaker 1 (42:40):
Of course. But I mean, I think it's become even more accepted at the moment to talk to your phone to write messages, as we were saying with

Speaker 2 (42:45):
Whisper Flow.

Speaker 1 (42:45):
Yeah. People are understanding that actually that's okay. Five years ago, you would have looked a bit weird talking to your phone to write a message unless the phone was sort of on your ear, right?

Speaker 2 (42:53):
Well, it's not really a hisperflow thing. That I think -

Speaker 1 (42:56):
Well, certainly not Siri.

Speaker 2 (42:58):
Well, no. I mean, I think that voice notes have been a huge thing. If you looked at LATAM, I hate them too. But many cultures like LATAM, Israel, China have used voice notes for years now, and that's been the default. Yeah, no, I feel like a violation. When someone sends me a voice note, I was like, when I'm supposed to sit here and listen to you talking.

Speaker 1 (43:17):
Exactly. Exactly.

Speaker 2 (43:19):
But that's that crippling social anxiety.

Speaker 1 (43:21):
But going back and talking to your glasses or having a pair of glasses on, I think in time will be more accepted. I think it would be okay. I don't think you'd look that strange if you just had that.

Speaker 2 (43:29):
I think it's more the privacy part, which is like with whisper flow. With a keyboard, I can be in a meeting and do whatever I want to do. I'm sitting there running a mega task and clawed and pretending to be listening to you. That's what I do in most meetings. Whereas if I had to use my voice, I'd have to mute, go off camera or just show you blatantly that I'm doing something else. So surely that's positive. I need my keyboards to keep that semblance of respect for you while I'm actually doing something else.

Speaker 1 (44:01):
Do you not think that in that case you'd be more attentive in a meeting because you have to, for instance, listen because you can't go and type because you're not able to use your voice to do something else?

Speaker 2 (44:11):
So

Speaker 1 (44:11):
You're not multitasking, so therefore you're more in the moment. Is that a positive outcome?

Speaker 2 (44:15):
I think you shouldn't have meetings where you're not doing that and you should leave them if you're just flying the wall that's doing nothing and doing something else instead. But that's nice in theory. In practise, it happens a lot more than we all want it to, right? So I don't think we should. That's not the reason to adopt these things, to deny yourself a keyboard. But yeah, I don't know. I think that adoption will fall in places where it's the most logical and convenient thing to use. There's a bit of a social stigma, but it's going away. But I still don't think it's the fully most natural way of doing these things because I think people like the privacy of an input that is text-based where people can't really hear what you're doing.

Speaker 1 (44:57):
What's your thoughts on the stigma around vibe coding and individuals just trying to build AI apps from nothing and raise millions off the back of it?

Speaker 2 (45:04):
What stigma?

Speaker 1 (45:06):
Well, some people are saying that vibe coding actually is just going to commoditize your industry specifically.

Speaker 2 (45:10):
Yeah, they need to try vibe coding.

Speaker 1 (45:11):
Right. Okay. Go on.

Speaker 2 (45:12):
What do you mean by

Speaker 1 (45:13):
That?

Speaker 2 (45:13):
Well, I mean, look, I think it's natural fear, especially of people who are deep specialists in a certain thing to kind of go like, "Oh, it's slop, slop, slop." And I've had that term thrown around poly a lot as well. And slop is just a phase. You have to go through that token maxing phase of just punching through and learning how to use it. Our fast cars just slop replacing horses because they can get out of control and get on a curb and kill 50 people. Yeah, they can do that. There's still a better way of transport than carriages. You just need to figure out how to use them safely and get some rules and governance around it and improve the technology. I think the pace of change with coding tools is staggering. Yeah, if you try to do something where you've architected a system that needs to work with Five9's reliability, I dare say that a strong team with strong DevOps doing that is still going to do it five times faster and probably better if they have these tools versus if they don't.

(46:23):
But they just need to know what the underlying assumptions are.

(46:28):
I think when I got into programming, we still use Pascal and Delphi and a few of those very low level programming languages and the stag just moved up and now writing Python feels like you're writing assembly code. But it's so much

Speaker 1 (46:43):
Easier to do.

Speaker 2 (46:44):
Well, it's easier, but there's a part where I remember the delight when we had the first few large memory leaks in Python, because for the most part, they don't happen nearly as much as they would if you use a programming language like C, where you had to just manual garbage collection. And that now I think to people joining the tech scene now, they probably don't know what I'm talking about. That's okay. It's just that this problem could happen and it's like they'll uncover something four layers down. Oh, and it's going to be an archaeological issue. They're going to be like, "Oh, I remember some dude on a podcast 10 years ago mentioning this once. I wonder what it is." And then they'll go and learn about it and it's delightful when you have to. I think that's what all of us technologists really, really like because you're building layers and layers of abstraction.

(47:33):
The internet is built on seven layers of abstraction. It's very complicated. It's like if you really follow through, it starts with a postal system and ends with a Zoom call between you and someone else that works perfectly even though packets are dropping left and right. So similarly, now we're just increasing the layers of abstraction. It's more powerful. We'll invent new things with it. Obviously, someone joining at that layer has much to learn about the layers below. And always, most of the people joining that layer and building in the layers that end up, because there will be more layers always, they won't know everything that those guys there know. It's all symbiotic. They shouldn't be against that, but I think they should be aware that most of the growth and value is going to come in those new layers. I think that you can be - Stacking.

(48:20):
Yeah. Yeah. So I think the stigma is wrong. I think you can definitely get yourself in trouble when I'm up coding. And I think the pace of improvement is such that it's not really. I'm at a point now where for the first time after years, I make changes in our platform. I submit pull requests and I can do that because of coding tools and because they've obviously improved relative to six months back where I would not have dared to do that. And do I know exactly how the full CI/CD pipeline works and stuff? No, I don't, but I know enough and people who know what I know enough about what I've done to kind of go like Alex, fine, let it in.

Speaker 1 (48:57):
What's interesting is Sam Altman said the next billion dollar company will be built with one person through AI, through Vibe Coding, for instance, right? You're nearly a billion dollar valuation, but you've got 300 staff. Why do you need so many?

Speaker 2 (49:10):
I think it depends what you do. I think the whole one person company is irrelevant. Why? It hasn' happened. Why is it relevant?

Speaker 1 (49:19):
It's more efficient. You can do it with more, as we say, no code tools,

Speaker 2 (49:23):
More

Speaker 1 (49:24):
Accessible.

Speaker 2 (49:24):
Okay. Is that a company or just a solo entrepreneur doing something useful?

Speaker 1 (49:28):
Well, what is a company? An amalgamation of people, is that what you're saying?

Speaker 2 (49:31):
Yeah, I think

Speaker 1 (49:31):
So. Or is it a service that you're selling to people?

Speaker 2 (49:34):
It's both. The old flywheel again applies. You don't have it without both things being true. It has to be efficient enough. And look, I think that one of the greater joys of the new wave of AI is just how much more you can do. And then you get to work with really good people, smaller teams. We're in Canary Wharf. I think that what I've always -

Speaker 1 (49:57):
Full of people.

Speaker 2 (49:58):
Well, no, no. I mean, what I've always discussed with my wife that is a difference between high finance and startups is that compared to, historically, at least compared to small elite finance teams of five to 10 people where they've handpicked people very carefully and they'll hire a few every year and they'll rise through the ranks slowly. Startups have always been about this. Hire a lot of people, grow really fast. And it's a bit like the Eastern Front, people dying left and right, sellers hitting quota, not hitting quota, people just rotating through different stages. It's always been messy. It's always been high casualty rates. I think with AI and these levels of efficiency, it doesn't have to be that way. It can be much smaller teams supporting a much larger revenue base. We see that too now. I think that our outlook is very different now compared to where it would've been with this scale even a year ago.

(50:53):
But I'm not doing this for efficiency. I'm doing this because I want to do it and I want to do it at a very large scale. So the whole one person company, irrelevant, fewer people so that. And it's really about holding context. The fewer people you have, but if those people are willing to really start themselves and hold that context, with AI tools, they become juggernauts. They're able to just do all these things because you cut the time of the debate, of the alignment, of the opinion, because it's just one person with all the information available to them. Obviously, if they get it wrong, they get it very wrong and they can make very bad mistakes as well. But I think I'm really excited about that. So look, I mean, I think we don't have to grow headcount from this point on to. And we could have lower headcount too.

(51:48):
But that's not really our priority.

Speaker 1 (51:51):
No. Okay. So what is your priority over the next three to five years? Where's PolyAI going? And if anyone actually wants to work for you or with you, how can they get in touch?

Speaker 2 (52:00):
Yeah. Well, look, I mean, our goal is, and what we'll be doing over the coming months, we've already opened up our model to the general public. Our platform is open and we'll be opening a hell of a lot more to people and agents. And that's really kind of the priority of our business is just to distribute this technology we've built and tested at such serious scale with some of the world's largest companies. We want to put it in the hands of more people. Because some of our most successful customers partners are ISVs, SIS people who know a space in a way that we never will. And they know how to provide value to people in those domains in a way that is really hard for us to learn in a few discovery sessions. So we're really excited about that kind of platform for future. So whoever's excited about doing that here or in the US and cares about the future of voice, I think they should talk to us.

(52:55):
And as you said, I don't think we're going to be hiring as much as we have in the past, but I think the density of talent is the thing that counts the most. So yeah, always looking for really great people.

Speaker 1 (53:05):
Thank you so much, Nicka, for coming on the podcast. I've loved it.

Speaker 2 (53:07):
Thanks for having me. Thank you, ma'am.