Rory Sutherland, Elfried Samba and Krish Ramineni on whether you can trust AI

[Music] There you go.

[Music] We're all talking and they just leave us in the room to just like figure it out. The podcast has both gained something and lost something from the use of heavy tech and filming, hasn't it? You know what's weird about it? What what I find fascinating about tech is what it ultimately reveals about what people really want. Yeah. Okay. Uh which isn't what we think we want or what economists think we want. As you think about it, when I drive anywhere more than 20 minutes in the car, I listen to an audio book. I listen to a podcast. When I was a kid, you had seven monkey cassettes in your car and you rotated them. Okay. Music. Okay. Suddenly, you've got access to the whole of world music in your car. You can listen to any song, opera, symphony you want to listen to and instead you listen to three people having a chat.

[Music] So we're obviously being we're getting lost in the whole like tech and art space and we're going to talk a little bit about that. So but before we get into it in its entirety, we're going to introduce our first guest which is Kish. We'll talk a little bit about what Kish does. Before we even start all of that, uh we obviously this is the debut episode of the bottleneck, right? And uh there's a reason why we called it the bottleneck. And I'm going to pass over to Rory to give us a bit of an explanation as to why we called the bottleneck and why we're here.

It was very simply we both bonded over our enthusiasm for bottleneck theory, which some people may have encountered through something called the theory of constraints. some other people may have encountered through things called butterfly theory, complexity theory, but bottleneck theory is a specific application of that. Um there's a great book called the goal by a guy called Eli Goldrat which is all about this whole idea of the theory of constraints that in any kind of dynamic system the performance of the system is constrained at one particular point which is effectively the narrowest point in the whole system and consequently intervening at that point will have disproportionately more effect on the performance of the system than anywhere else. And I would probably argue, this is where I get interested partly because it's a gateway drug to complexity theory, which is a good thing anyway, but I'd also argue that it exp I think a lot of bottlenecks are psychological bottlenecks rather than necessarily being constraints on your production line or your distribution or something of that kind.

It informs lots and lots of areas of study. You know, you could use bottleneck theory to explain very quickly why a roundabout I know you're American and you hate roundabouts, okay? But you could use bottleneck theory to explain why a roundabout is superior to either a four-way stop or indeed a a set of traffic lights. It always strikes me as weird by the way that America, the land of the free, has state controlled traffic through the vehicle of traffic lights, whereas continental Europe, which is supposedly derist, leaves it to the free-for-all of the self-organizing system, which is the roundabout. Okay. But you can use bottleneck theory to understand effectively that the glorious thing of a about a mini roundabout over a set of traffic lights is that there is only a constraint on your forward movement when it is necessary for you to stop.

Okay. But in marketing terms, I think it explains why I've always thought that marketing is fat tailed. By which I mean 10% of what you do makes 80% of the difference. And I think bottleneck theory partly explains why that might be so. that there are areas where I always give the example of a road. There's no point in widening a road if there's a badly synchronized set of traffic lights 300 yards further on. You'll just move the constraint down the road. Yeah. Okay. So, act on the greatest the narrow point of the system first and optimize from there backwards. And that might explain why I think in a lot of marketing we need to be much more experimental to find out where the bottom bottleneck is and much more imaginative in terms of how you resolve it. But it's a brilliant brilliant approach. I'm not saying it solves everything. Okay. But it's a very very useful toolkit to deploy if you want to concentrate your efforts in the area of highest payback.

Hey, hey, we were just talking about Tik Toks and one of the biggest Tik Tokers in the world right now is a guy called Kabi Lang >> and he basically is a human bottleneck almost like fixer because what he does is he showcases how people try to do things in a very complicated way. He does it in a very simple way and he responds with this. Right? So it's basically like as a human >> using like very simple like getting to the heart of what the problem is and able to demonstrate people that the demonstrate to people that the solution is a lot easier than they think. But uh before we step into that Tik Tok rabbit hole, right, we got Chris, our first guest on Bottleneck and we want to hear a little bit more about yourself. What have you been working on? What are some of the bottlenecks in your industry?

Well, we've been working for almost nine years now on building an AI assistant for people that can follow them around to their meetings, help write down notes for them, remember what to do, and be really, really good at helping you have perfect memory. That was the problem we wanted to solve. Now, we've had so many technological bottlenecks from the day we started. On day one, people would say this technology is not good enough. Right? transcription when we started was only 80% as good as what it is today. So at the time we had to change the framing. It's not about having perfect transcription or perfect notes. It's about is it good enough to jog your memory and help you recall information. Obviously now the AI today has gotten so much better that we can pretty much argue it does better than a human in terms of writing the transcripts as well as generating notes. But that was like a big bottleneck for us was changing the perception about what are you trying to solve and also the bottleneck around cost cuz the technology that when we were building it was extremely expensive but it was still cheaper than hiring a human transcriber annotator uh it's almost it would cost me about to do one hour with a human I can do a year's worth >> with Fireflies. So that was like a huge uh we had to change the the perception for people. It's like hey, it's not perfect but it's consistent and you don't have to worry about the cost like this is going to only get better over time. But then you run into another bottleneck which is I don't trust having AI on my call. This looks like a spy. Like >> that's a psychological bottleneck.

>> Yeah. What's going on here? Like I I don't trust this thing. Like, can you please remove it? It's crazy. the same people that said that to some of our early adopters, the customers, right? You always have these early adopters that are crossing the chasm that, "Hey, I don't I don't feel comfortable with this. Can you remove this?" And the other person saying, "No, no, trust me, the notes are going to be awesome. I'll just send them to you afterwards. Just give it a chance." >> To that person that comes back to me and says, "Hey, I was not a believer. I thought you guys were like some sort of spy, but now if I don't go into a meeting and don't see Fireflies, I get anxiety because I know I'm going to forget stuff. So, that was a complete 180. Uh, but it's all perception. That was that was the crazy part.

No, no, I can remember this having this uh AI agent turn up in Zoom calls and being basically slightly weirded out. And we ought to be honest about it, which is there are there are calls and conversations where we don't want to be recorded, of course, >> you know, just as there are times when we want to be off the record. And I think making that easier is probably a healthy thing. And you might argue that anybody should have the right to turn those things off for a minute at a time simply because there are conversations that will never take place if we believe that we've entered something onto the public record, but they're still necessary conversations. So there are, you know, there are technological solutions to psychological problems as well. Um, which I think is always interesting. But I'm also I'm also intrigued by the to what extent is it is it conversational the interface?

So to the current version of Fireflies, the way it worked was it's a silent notetaker. That's why the Fireflies fly on a wall. There's actually a funny backstory to the name Fireflies. We were actually working on drone delivery and at night drones look like fireflies. So, we said, "Let's just use that name." And this was me and my co-founder in college. We were too lazy to change the name cuz we tried 10 different uh ideas and many things that we tried didn't work. We finally fell into this like AI world >> and said, "You know what? I'm too lazy to change the registration. We're just going to run with Fireflies." And then one of our customers says, "Fireflies is such a brilliant name." I agree with you because the idiom is, "I'm just going to be a fly on a wall and take notes and be quiet." So they actually reverse engineered the explanation for your name rather than >> so the customer came up with this backstory and we we said okay that's great we're going to run with that.

I'm very much by the way intrigued. I hate to digress from Fireflies and we will come back to it very quickly but drone delivery and indeed drone transportation sometimes called the stenosphere or the low-altitude economy that interests me enormously because it's an area where 95% of the technology has been solved and the remaining obstacle is public acceptance. And so we undoubtedly have a both I suppose a news media and a culture in for example governance where small percentages of people who complain can make disproportionate noise and what fascinated me speaking to someone you probably know Bobby Healey in Ireland who does drone delivery with a company called MANA manner is that everybody's achingly conscious of the fact that let's say you make I don't 10,000 drone deliveries. Okay, that probably avoids two moped accidents, one of which could be mildly unpleasant. Okay, those non-accidents are invisible, whereas the second a can of beans lands on someone's dog, okay, it's front page of the Daily Mail and calls for an outright ban. And so, one of the really interesting creative questions for me was how you how you gain public acceptance. And we were talking about things like blue light flights where if you have a network which is publicly funded, okay, sorry, it's privately funded. In other words, it's a drone delivery network, but it automatically in a sense as part of its license has to give priority to urgent deliveries of defibrillators or epipens or urgent pharmaceuticals and that it delivers free to the elderly, for example. You could create something which the public accepted because now we tend to frame things as a private sector business or a public good. Okay. But if you look at something like a cardo, it's a bit of both. Obviously, it's a private profit-making business. No one's going to deny this. But on the back of the private profit-making business, you also have a really essential public service in terms of when my father was in his 90s, you know, 70% of his food was delivered that way. you know, it's it is actually a public service as well. And if we can get the public to understand that, you know, the post, you know, the postal service, if we go back to the penny post, okay, it was a commercial, well, it wasn't actually, okay, but it was a sort of it was intended to be a commercial profit-making business, but obviously it it conveyed things that were utterly trivial, which funded the network, but some of the traffic on that network was absolutely mission critical. Okay? And if we can get people to accept the drone network in the same way that actually okay mild irritation is worth paying the price for for the simple fact that you can now rescue people golfers who have heart attacks. Do you know this?

Absolutely appalling fatality rate if you have a heart attack on a golf course >> because they there's no one around. The ambulance can't get to you. You're miles from the clubhouse. Now, being able to deliver a defibrillator or something in, you know, a minute and a half, >> um, you know, would probably save an appreciable number of lives.

Yeah. >> Yeah. Drone firefighters as well. >> This is that's really, really interesting. Absolutely right. And and also the lives you save through things like inspection. So, you know, the crazy thing is you have this massive regulation if you use a drone offshore to inspect a wind farm. Okay? And you have to go through all these various safety regulations. Now the reason you should be a little bit lax with the safety regulations is the alternative is some guy goes out on a ship and climbs up a tower 300 ft high where the risk is so inordinately greater that it's crazy. You know in a sense one of the things you have to do with with considering drone safety is consider what the alternative is. Otherwise, you're trapped in status quo bias, which is it's acceptable to do something really, really dangerous and established, but it's not acceptable to do something much much safer and new.

Yeah. >> So, that was a distraction. It strikes me as a very very interesting case where we basically have the technology where I could I'm probably a bit fat but the thinner people among us could sit in a kind of you know effectively an autonomous drone and ride 30 m at 80 mph in you know low altitude at extraordinarily high speed with an amazing level of safety compared to helicopters insane level of safety. The only obstacle there really is acceptance. One of the things that I had to do when we go back to safety, my senior thesis uh was on making an argument that fully autonomous self-driving cars are going to be safer than having a hybrid approach where you have >> robotic cars and humans because that system will most likely have some sort of miscommunication gap. Whereas if everyone on the road is self-driving cars, if you've ever been to San Francisco, take it away, where it's a transformative experience. But >> there was this whole thing around network theory where if everyone has the same sort of information and signaling in that area, >> the self-driving cars are going to be able to communicate well with each other. Yeah, >> but when you throw humans, >> human error, >> human error will get amplified and it might actually end up being worse than just having only humans or only >> ironically you can make an evolutionary argument for religion in humans >> as a mechanism to make us int more intelligible and predictable to each other because people who shared adherence to a particular religious belief or pattern of practices. Okay, it's easier for you to navigate other people if there are certain givens to which you know they'll adhere. So it's you know religion could be seen as the rules of the road and obviously the rules of the road for autonomous vehicles are going to be so I mean there are there are aspects of it which are going to be terrifying like the most efficient way to handle a junction with autonomous vehicles is literally to have vehicles going straight through at 70 mph uh 10 ft apart at right angles to each other because they've got the skills to do it. Now most people I think if you're going to do that in Whimo you got to introduce brown upholstery first I think uh because otherwise you're going to have a bit of a problem. I mean by the way I was just listening to a fant reading well listening to actually a fantastic book called counter intelligence by a guy it's kind of history of GCHQ and codebreaking. There was literally a guy at Bletchley Park, mathematician, who believed that the safest way to navigate a junction was to go straight through at high speed because the likelihood of hitting someone else was reduced the faster you went. Now, I have to say he'd got something wrong there in his mathematics, okay? Because I don't think he was considering the worst case scenario, okay? Probabilistically, he may have been on to something. You're less likely to crash. Okay? But there are things and of course driving very very close to each other would be another thing that they can do at speed. We can do that at humans at about 30 mph once you get up to 40.

We've got to effectively allow for braking time and reaction time and a whole we've got to build a whole lot of buffers in which you won't need.

Well, traffic jams also you're going to need fewer cars especially you you won't need parking lots cuz they're going to the car is going to drop you off and go pick up a bunch of other people. That's the whole thing with Tesla's uh uh robo taxis which is coming at some point in time. Uh but that's it changes the way buildings are going to be built how uh city >> landscaping will be done.

Uh and yes, so instead of going 45 for us 45 miles an hour, >> uh maybe the acceptable rate in a city will be 70 miles an hour. But do you think that they that they're going to ban the fact that humans can have their own driving license? Because based on your theory, >> you can only drive on the racetrack for leisure. I don't know. >> Driving on the road becomes that thing people used to do back in the day. Right. We we've got to be very careful about this was one obstacle you face there is that outside London and even inside not everywhere >> uh not under all conditions but outside London um and in London after about 11 p.m. before 5 in the morning people really really enjoy driving.

Yeah. >> Okay. Now it's a sort of flow state. Okay. It's something we can do which involves a lot of tacit skills. It's a bit like playing tennis for lazy people, but it's it's it's a skill we really really enjoy exercising. And so at some level, one of the questions to ask is what are the kind of journeys where people would really like automation and long motorway journeys will be a great example where fatigue is a big risk etc etc etc. um you know certain long journeys. I think on the other hand there are going to be decisions which you can this business of removing the steering wheel strikes me as fundamentally misguided because there will always be occasions like parking in a field for a wedding. Okay, you know that kind of shit where ultimately you will always need to be able to override because there will always be some sort of contextual information that you need to factor in. like I I can park behind Dave's car cuz I know he's not leaving till late. You know those decisions. You arrive at someone's house, is it okay to park behind someone else's car and block them in? It all depends on your knowledge of when the other person's going to leave. Now, >> at some level, social context will always play a little bit of a part in what we do at certain, you know, certain kinds of driving.

Well, even for Whimo, there's a human operator with a remote control and like eight screens managing several different cars in the background. So even though it's a fully autonomous car, there's still a human override.

And also going back to the point about like Fireflies, right? >> Honestly, our company would probably not be in business if it wasn't for Fireflies, right? So So what was happening when we first set up our business was that everybody the whole company was in every call all the time, right? Because we hadn't post pandemic, everybody was still behaving like if we were still in an office. So for you to be able to get the information raw and direct, you had to be there for the experience, right? And the issue as well is that we're we're taking every meeting digitally. So if you've got an if you're actually literally taking notes, not only is there an interruption in the in the connection because you're on a screen, you're also like stopping taking your notes and there's actually no connection happening. And then Fireflies comes in and what that's did it means that everybody can just like focus on what they need to work on and you can catch up when you need to, right? Which is amazing. I think the thing I liked as well was the fact that it gave you tips on how to improve your speaking your speech uh level who dominated the conversation which makes it really really great for training right and I thought that that was really really interesting and we started to take that data as well and based on a conversation with somebody that's not that's external to the company we're able to predict whether the conversation went right or wrong and then you can also predict whether the likelihood of a deal happening was going to happen or not which is super super interesting right so Fireflies for us went from what the hell is that? We're going to take note takers. It's going to spy on us to the point where I was actually saying that I might actually create a film for our organization and call it Fireflies because every single meeting we've had from the start was documented which is super super interesting. Right.

By the way, I mean one tip I'd always give also if you're on a Firefly conversation is make sure that for 2 minutes of the meeting you discuss something really obscure like pangolins. Okay. And the reason for that is if you have this wonderful privilege that every all your past conversations are searchable.

Ah >> okay. And and you can remember that you can't remember anything. You can't remember the guy's name. You can't find his phone number. But you can remember that you talked to him about pangolins. Okay. So it's always worth it's like a hashtag.

Oh yeah. One of the funniest things is uh we >> can we hashtag this podcast with pangolins now. So now Fireflies, what it does is before a meeting sends you a meeting debrief. So if we met and we're connecting again a month from now, they'll say, "Hey, this is what you guys talked about. This is some of that small talk, the chitchat that included it." And then it also includes information about this is what we think this person cares about. So when you are trying to build relationships and stuff, Fireflies can basically like say here's all the conversations you've ever had with this person and this is like to help you just pick up where you left off even though it's been a month >> cuz that is the worst thing when you have two uh video calls back to back.

Yeah. >> Which is the second one. You basically it's like being landed in enemy territory in a parachute, you know, after dark. You have this sudden business where you have to spend the first two and a half minutes basically reassembling >> reassembling everything especially for recruiting >> and I ask candidates.

Yeah. I this is my third interview but I'm being asked the same questions that the recruiter asked the hiring manager asked like ask me something different like I used to experience that when I used to go do interviews and so we have this approach with Fireflies now where everyone gets the recaps and so when the >> recruit Hollywood or rather the streaming services have been doing that for ages with any series >> previously on 24 >> that's the that's the thing and uh >> and it is absolutely necessary because otherwise >> we'll give you a recap uh and And then we can also, what we do is have it turn an entire 1-hour call into 30-second sound bites, clips, highlight clips.

And so it'll go to the most action-packed moments of the call where they were about to get on a into a fight. Yeah. It'll be It's like a trailer. It's like and all those little It's like when you watch a podcast and you see the preview. So it'll give you 30-second bits of the most exciting that what the AI thought was the most exciting uh parts of the call. And I just can go through those like 30-second clips and then when I come into the meeting if I don't even have >> You're back in the room. I'm back in the room. Brilliant. Yeah.

Yeah. >> So imagine if you threw in a random hashtag in there and the AI catches it.

One fantastic thing. I don't know what Google's doing with Android XR, but ultimately augmented reality becomes an takes this same technology into the real world, doesn't it? >> Extremely well.

Yeah. I mean the thing with AI is context. So the more you use Fireflies or any other sort of AI tools, the better it understands what your goals are, what your objectives are, what you care about, cuz a salesperson >> wants to see something completely different in their notes. Like they want to know when the customer had objections or when they were angry or when the deal was going south. Whereas if I was a project manager, I want to know when the deadlines are going to be hit, why we like slipped up. So we had to go a level deeper and say we need to build this to be the most personalized experience for every type of meeting cuz not every meeting is the same and not every person is going to want to talk about the same stuff.

So that's what's really exciting for me in terms of AI or this like personal assistant uh is the more it understands you the more relatable it will be. And I love the example with the pangolins because what what I do sometimes is, oh, I can't remember what meeting, what date I spoke to this person. So now you can just chat with our AI, like we call our AI Fred. So you can just ask Fred, I swear I remember talking about drones and uh uh autonomous cars with Rory. And then it's like, oh, you meant this conversation. This is what you guys talked about. It'll pull that up.

Yeah. >> Yeah. And that's amazing. So, we're talking about AI now as well, right? And there's a lot of contrarian views and opposing views about how AI is going to impact the business world, right? You know where I'm going with this. So, from how we hire people to creativity and how we create content to even uh uh whether remote teams are working or not, right? So, I know that you have your own views from that perspective. Roy, it would be great to hear about what how you think that AI is going to impact business today and tomorrow.

I mean in terms of by the way remote working is also going to and and I think we're weirdly focusing on AI >> a little bit too much and focusing on the revolution that is remote working a little bit too little. I'm not when I say this I've got to caveat the fact that I think you know I'm not saying that AI is overrated or that it's hyped. Okay. However, what tends to happen in the business conversation is you get a winner takes all effect where it's I call this the sourdough problem, okay, which is I really like sourdough. It's a great addition to my bread repertoire. Unfortunately, it's become absurdly overdominant, okay, in the in the field of artisan bread compared to cornbread, soda bread, and lots of other for lots of other interesting things we can make. Okay. Are basically all being sacrificed on the altar of sourdough. And in the same way, I think there are about three or four discussions in business that should be proceeding in parallel.

Yeah. >> Um and actually the importance of flexible working uh and remote working and how to do it well and how to do it badly. Mhm. >> is is not a trivial conversation because we we have made a lot of mistakes in technological adoption which seem logical but ultimately fail at the same time if we use bottleneck theory if you're a business that relies on human talent demanding that your talent all collocate when I would argue after the age of 30 it is more or less impossible for anybody to move from elsewhere in the UK to London.

Yeah. >> Unless they get a job at Goldman Sachs or something, right? It's more or less, you know, particularly if you're married, if you have kids, you know, it's an impossible move to make. And so, you know, if if if generally you believe you're in the business of of of human talent, um then uh anything that actually frees up who you can hire and where they can be and also frees you up to hire people disproportionately women, I would argue perhaps, who have other responsibilities which mean that they need a degree of autonomy in their working hours in order to be most economically valuable. That seems to me to be important. It's actually it's unlocking a bottleneck.

Yeah. >> Okay. And we're not talking about that nearly enough.

Yeah. We when we set up our company, we had all the obviously the option to be able to have an office somewhere, right? But then we realized that actually if we really wanted to grow a company that was truly like diverse and you have people from different walks of life and you go where the best talent is then you can't have something that has as much friction as an office and almost like telling everybody that they have to come to the office every single time right there's people that I've worked with in our organizations that I've never met in real life right and the first time I met like my co-founder used to um live in Cancun and I lived in Dubai Right. Got it. So, so what was happening there was like we're running this company and growing it and we we're not anywhere near each other. Obviously, over time we got a little bit closer because of time zones, etc., etc. There's a system around it, but effectively I think that that can be done if you have a system around it and if you're not using systems of old to be able to like build a company of tomorrow. I don't know what you have to say about it.

Well, for me, I only had one year of corporate experience before starting the company. I worked for one year at Microsoft. They're pretty traditional. Everyone's in the office. This was before the pandemic.

And you're going in five days a week. >> Uh what's interesting is when we started our company, >> investors said, "I think there is a significant risk to you choosing to be remote."

Uh and my co-founder and I started out remote. This was in 2016. In fact, the way I met my co-founder was actually remote. We worked on some class projects. He went to MIT. I was at Penn. Uh so for two years we worked on remote projects and never met in person.

No. No. >> Yeah. And then we met in person like 6 months before we decided to say, "Hey, let's go all in and start a company together." So, you know, people talk about like online dating and meeting strangers, right? So, that's actually a really interesting thing. Uh they're asking a lot of wedding planners, how did people meet? And right before it was like 80% oh through friends of friends uh through family at work now it's like 70% are oh from a dating app.

Yeah. >> Yeah. So that's like so what was what I would say about remote work is in the past there was a stigma because one the technology was not good enough like you could not get a constant consistent webcam connection. internet before 2011 was quite slow >> and >> certainly the upload >> upload was really slow and then you're gonna always have like the static the issues of like actually having a proper interaction >> but now I >> you also had an interesting problem which is the call was only as good as the most technologically incompetent person on the call it to be going you know permanently on mute fiddling with their camera etc and the whole call was wrecked so it took the pandemic to get people up to at basically the threshold of competence.

Yeah. >> Yeah. Even though the technology was there, it always takes people a couple years to become uh okay with it. And then it has to be socially acceptable. It's technologically feasible, but is it socially acceptable?

My argument when we started the company was I looked at investors. I I talked to folks and said there is nothing that a knowledge worker cannot do remotely today that you can only do in person. Obviously, if you're a doctor, you're not going to be able to perform surgery remotely. Well, maybe a robotic surgery, but there is so much that a knowledge worker can do remotely. So, we're going to go all in on that. And then the theory for us was we could do hybrid, but the problem is the people in the office might know certain pieces of information and the people that are away are going to be left out. So, we had to make a decision where either you have to be all in person or all remote. And so the advantage of being all remote was we could hire people across 20 countries and that comes with its own challenges with time zone. So remote is not necessarily easier.

Being candid, I would take your real estate budget and spend it on an annual jolly possibly in >> Cancun. We do we go to a tropical place every year and we bring everyone in for a week or two. And what's crazy is there are people we have a rule in our company. This goes even extreme. You do not need to have your camera on. That's crazy, right? I've hired other than when I saw them during the interview, >> my camera or their camera. So, all I know is their voice and their profile picture. So, putting uh the picture to a face in real time and their voice was the craziest experience we had this past year when we did our offsite.

And you often get the surprise for them about how tall or short people are cuz you have no idea. Yeah. I mean no I mean I definitely I think you know what I wouldn't do is bank the real estate s the rent savings entirely. I think there's compensatory expenditure needed and indeed you know some people some of the time prefer collocating not necessarily at a head office but somewhere I think there are going to be property costs entailed somewhere but and and actually >> bluntly put I think you those two weeks are very very powerful because you bond in a space like that in a way to be absolutely candid you don't bond in an open plan office in the same way.

Yeah. What we realized um when we would bring people and this is where I've worked before and even in our company when you bring people in one place not a lot of like that well the efficiency of what you get done doesn't get better it actually gets worse because everyone wants to make the most of connecting with each other.

I'm very chatty in person and so I have to for me the open office space I'm going to get distracted someone's going to come up to my desk we're going to have a conversation. What I do love about in person though, you can have that spontaneity of let's go to the conference room, let's whiteboard. So I do feel like the transfer of information can be more spontaneous.

But the advantage with remote is everything has to be documented and written down. So you don't have to explain yourself twice. You don't have to have the same meeting twice. Uh when we're onboarding someone, we just refer them to the Firefly's previous recordings or the notes. So everything is like right there. But the one thing that I do think is so important. We we talked about self-driving. We talked about commuting. So the biggest challenge was the epiphany was I went to India. Yeah.

I went to Delhi and had one of my employees. >> We took the train >> and then he was explaining to me how long it takes to go from the suburbs where he lived to into the city. And we were literally packed like sardines. It was insane. And I'm thinking you commute two hours to the office. You're exhausted.

Yeah. >> You're going to spend like another 30 minutes in front of the the coffee machine. I don't know how much you're like work you're going to be able to actually do versus take that 2 hours every day, spend it with your family, go for a walk, live in an area where it's not as expensive as living in the city in like zone one >> and you can like it it's transformational for them. Yeah.

Yeah. One thing that I've realized though to challenge all of our thoughts though is that was the purpose of an office simply just to do work right or was the other social um almost like benefits of being around different peoples and peers right from I I think I watched a podcast where uh Emma Greed behind skims and good American etc. She talked about the fact that actually because she was able to see leaders in front view and how they acted and how they spoke and how they moved and how they communicated, it allowed her to build up her own skills, right? Because she could see a live example of how somebody moved. So maybe if you're like a junior getting into the industry and you haven't had the opportunity to work somewhere before and you haven't never had the opportunity to actually see greatness or someone that's senior in your role do what they do. How is it how how do you grow? How do you gain those skills? How do you gain that experience? And also a lot of people's friends they meet um at work, a lot of people's couples meet at work, right? So maybe the the role of the office >> is remote and it takes a certain type of person.

I think you you went back to your phrase about context and I think generally >> um what we're looking for is wrong. We're looking for what is the right place to work. In fact, what you want to do is provide people to some degree at their own choosing, but to some degree not >> with as wide a range of contexts as possible in which they can work. Yeah.

So that there are moments where you want to be sociable. There are moments when you want solitude. Okay. I've often said the ideal office is half library, half pub. Yeah.

You know, and what we've got is an open plan office which is neither one thing nor the other. It's neither sociable nor is it a solitary space where you can get on with stuff.

Uh I also think you you know there are people who are introverts who we ought to nudge to go into the office more and there probably people who are extroverts who we ought to nudge to stay at home more.

Um Paul Dolan very good uh behavioral economist at London School of Economics always said his rule was go into the office one day more than you would left to your own devices. In other words, it's a bit like forcing yourself to go to a party when I've been when my wife has kind of bullied me to go to a drinks party. I'll be honest with you, at the time I just want to sit at home and watch a YouTube documentary about traction engines, okay? And I'm bullied to go and get dressed and go to this party. More often than not, I'm glad I went. Yeah. Okay. Now, that doesn't mean you force yourself to go to every party you're invited to, but some degree of just pushing yourself against your natural inclinations is probably pretty wholesome. I'd also say there's also you know if you take your colleague in Delhi you know two days a week why doesn't he work from home in the morning and probably at lunchtime the train's empty okay so he can work well work on the train conceivably you know there's also the question of just simple time shifting >> yeah because usually the 9:00 a.m. traffic and then the 6 p.m. Yeah. And so the other advantage of remote work, well there's pros and cons to this. The advantage of remote work is >> I feel that we operate at our best. We have these epiphanies when we're out on a walk or when we're doing something else and you're like thinking about it's not you sitting for 8 hours being monitored by your boss. So those sort of epiphanies happen when you're doing other things and it's working in the background. But the other advantage is you can also when you're working remotely you can choose to take like a couple hours during the day to do other things but then people will work in the evening so that you have that flexibility.

It's good use of it's optimizing the use of daylight. I mean if you think if you think of anyone you know who's based in Scandinavia they spend about four months of the year going to work in the dark and coming home in the dark. Okay.

When I was in Seattle that's what happened. >> Seattle is pretty much the same. Yeah. Yeah. Uh yeah. Get home far. Yeah. So it's about the same latitude as London, I guess. Seattle 49th parallel, bit further north is I guess. Um but it's um it's interesting in that respect in that um you know uh the reason it might very well be natural for human beings to work for 4 hours, take four hours off and work for another four hours. And it was commuting that effectively made that impossible. Not uh you know it wasn't anything to do with productivity or value. The fact that you had to cram everything into a kind of eight hour stretch is undoubtedly a product of, you know, wanting to get home at a reasonable time.

Yeah. And then you're exhausted afterwards and you want to do something else. You want to go to the gym or you want to you you get into like this routine and then weeks, months uh go by. But the other thing is a lot of the tech the big tech companies uh who were all in on remote work during the pandemic uh were slowly saying no you actually have to be back 3 days in the office and now most of these companies are 5 days in the office or you need to resign >> there I feel that are you doing that to get them to be more collaborative work more or did people abuse the system >> where they were working the bare minimum And now I need to see you be a busy body in the seat and I need to monitor you so that your you know your productivity levels can actually be tracked.

There is a problem there which is you are calibrating your level of uh stricture and general bossiness.

Mhm. to the people in the office who are least trustworthy and least motivated, which then imposes a massive cost on the people who are your most talented and valuable staff, who are people who are talented and are motivated. And so by forcing those people effectively to a lowest common denominator of general attendance, that seems a pretty dangerous thing if you want to hold on to talent.

Yeah. Especially the So we we have this policy at our company. I'm not looking at when you're clocking in or when you're clocking out. We actually barely have any recurring meetings.

Instead, what we're doing is we're measuring output. Whether that takes you five hours or eight hours, we know what output you're going to be able to deliver. Cuz I have friends in finance where no, I just have to act busy for my boss, for my MD, and I have to be there, warm the seat, and and we we said we're going to be the complete opposite of that. Uh it's not about you're not getting paid to look busy. you are getting paid to actually deliver outputs and we'll measure outputs and if we have issues >> around outputs we'll see like what the causes of that are. But yeah, a lot of times uh the people that are the most productive, the most creative, the high performers, they want more freedom.

And so remote work is naturally comes with a lot of freedom.

But to maintain freedom, you need to also have a lot of discipline.

Yeah. So it takes a certain type of person because there will be people that can easily fly under the radar, push their work under the rug and do the bare minimum. But we don't want to punish everyone else because of those like

low performers.

And going back to your point though, you talked about maintenance, right? So how does a company maintain a remote workforce, right? So what are some of the things that you have to learn and unlearn to be able to make sure that you get the positive outcome of remote working?

One is you can still create a really strong bond with your teammates. It can be lonely. We tend to hire certain type of people that are extroverted introverts uh where we tend to we we started something this past year where people in different regions can get together once a month and co-work together for a few days uh at a time. And so that starts to build like the bonds. We have the offsite where we fly everyone to a country away from the hustle and bustle. They can work together. Uh but a lot of times we're also constantly connected on the like Slack or any of these messaging platforms. So people are very understanding if hey I have to go pick up my kids right now at this time. Could you go do this? And people will sub in. So one of our sales people has to go they have an emergency someone else will slot in.

The other beauty of remote work is we'll have someone come in from our customerf facing team and they say, "Hey guys, I have a customer here. They only speak French. Can anyone come and do a demo for them in French?" And one of our engineers who's never done a product demo is like, "Hey, I'll step in and I'll take care of it." So that's like the cool part of like the diversity. When you have different people, everyone's like working together and helping each other out in skills that they might not be familiar that is total football, isn't it? which was a Dutch concept where every player can play in every position.

Yeah. And tying this to AI, I'm a big believer that we have gone from hiring specialists to now only hiring generalists. A very interesting bottle application of bottleneck theory here. without turning this whole thing into a total distraction which is always dangerous.

I'm a bit of a georgist which is you know the belief that actually I won't go into the whole single tax on the value of land business but the in the core insight of georgism or one of many insights is that there's no theoretical limit to the amount of capital you can have. There's no theoretical limit to the amount of labor you can or worthwhile labor that can be used because we can increase productivity potentially to an extraordinary degree. So the limiting factor is land.

Mhm. And so what tends to happen is because land is the thing that you can't make any more of, the people who own the land tend to acquire disproportionate power to extract value from the people performing useful work on the land which they own.

Okay, which is economic rent. And that's of course I time and land are fundamental limiters. Okay.

Land always appreciates over time and and and consequently, of course, they're not making it anymore. Land always appreciates and nearly all So, the argument is effectively in Georgeism is that it's as if, by the way, you own any improvements to your land in Georgeism. If you build a house or you make your farm more productive or whatever, you get to keep the gains. Okay. Um, however, land is treated as though the actual land itself were owned by the government and held in common for the populace. So the only tax you apply actually Texas is not miles from this. You have no income tax but you have uh land value tax property taxes. Um and that's actually technically it shouldn't be a property tax in Georgia's terminology. It should just be on the value of the land on which the property sits because if you build a barn on your farm that's the product of your labor and the gains are yours to keep.

Mhm. Um it's to be honest I think it's a very very insightful way of looking at the world but it is also it occurs to me it's something that actually derives from asking the question where's the bottleneck what's the constraint and I mean one of the interesting things with AI is we can ask the question what are the cognitive constraints okay like not being able to hold seven meetings in your head simultaneously okay what are the cognitive constraints which this can help resolve.

Um that's That's a really good point. When one of the things that we look at is we have a rule uh especially with Dunbar's law that we want to keep our company as small as possible for 150. 150. So we're around 120. We're t we're getting very close. For the longest time, we we kept it to uh 100 people. Yeah. Even though we've scaled 10x and right now we service about 20 million people. Yeah. And we only grew our support team by three people from one to 20. So that's because we use AI in every aspect to help automate half of our support tickets, our support conversations.

In fact, the next phase of Fireflies, uh, which I haven't announced in many places yet, is why have Fireflies just be a silent notetaker on a call? This goes back to your thing. Can it be interactive? Why not have the AI just go take the meeting on your behalf? Especially if it's a recurring meeting, if it's like a redundant meeting. So, the first AI we're going to roll out is an AI recruiter that will go screen and interview hundreds of candidates simultaneously for you. And it knows the questions you as a recruiter ask, how you how detailed you get, your persona and you're essentially cloning your best recruiter to go take on that stuff.

Now, does that mean like, oh, our recruiters are not going to be able to do other work? Absolutely not. Like, it's actually going to give them more coverage. uh and then to be able to do the more analytical part and decide the right candidate. But if your bottleneck is I can't be on 10 interviews every single day interviewing candidates, I need someone else to do that so that I can go find the best candidate. We want to be able to allow people to do that. So that's one bottleneck where we've solved.

That raises some important ethical issues though because one, we need a system. A good recruiter would occasionally hire a wild card. They would also vary their criteria every now and then simply out of uh because cognitive D. So if you optimize the danger would be you optimize for one thing and actually what you need to be optimizing for to a degree in a team is optimizing for complimentary skills and variety.

Yeah. Um so so there are some really t I mean there are some very very dangerous questions we've got to ask. We don't want to have the AI decide who to hire. It showed this it's it's basically doing if you think about like the robotic task that a person has to do where they go through a list of questions ask the same questions.

Can you explain that to me a little bit more? Uh the nice thing is we've trained our AI where it'll go through the list of questions. It's just purely doing the functional screening interview. But if one of the answers isn't very detailed or it feels off, it'll automatically noted.

Can you go a little deeper there? Can you explain that. I feel like uh this is different from what you said earlier. It pays attention to the little the nittygritty stuff. So, I'm guessing if you take Fireflies, it'll learn about you, right? And your patterns over time. And I'm guessing you're cloning your best recruiter. So, they've probably done like 500 of these interviews. So, it knows when it pivots. It knows when it gets a certain um almost like cue to be able to like almost like counteract uh what the computer says it should do. You give this AI agent an end goal, right? I want to know these criteria or these pieces of information. You figure out what's the best way to ask those questions to get me that information and then you do it across a 100 candidates and then you analyze all the candidates and you give me the data.

Yeah. I will ultimately make the decision. But now I can like look at everything and make that call. Right. So like it's there's no shortage of data. So this is something that I I I I believe like paradox of choice is only going to get crazier in AI because there was a limit to how much I could research when I were to buy something, right? Like I I want to go buy a new iPad. I am one of those crazy functional buyers where I will look at every stat, every spec. There's a clear reason why I'm a Samsung user and not an iPhone user.

Uh I am as well. Explain it. Explain that. I'm also a Samsung user. Explain the uh health healthy loyalty to the glorious ecosystem which is it's so funny. There's something there's something about Apple that all of my friends as they start getting married. They switch from Samsung to Apple. The green bubbles, the blue bubbles, the peer pressure to get on that, they they all switch. So I'm like one of the lone survivors that uses a Samsung. I'm like objectively I get my stylus. The screen is bigger. you know, the camera is like it folds. Yes, there's a foldable version. Uh, but yeah, I I have for, you know, I also have a suspicion which is that if a Samsung phone is reasonably popular, it is popular on its intrinsic merits. Okay? Whereas I always get the impression that Tim Cook could suddenly launch the new Apple turd on a stick and there'd be a load of Apple fanboys queuing around the block to go, "Oh, I love the way it tapers towards the end, you know, doesn't it smell glorious?" you know, okay. And my suspicion is that I don't fully trust peer pressure and social proof with Apple, whereas I do with Samsung. If a Samsung phone is popular, there's reason to believe it is actually a pretty good phone.

If you're a functional buyer, we do this. Now, the you've said this yourself in the past that the reason marketing works is because humans are irrational buyers. We we appeal to emotion. We appeal to what feels good. And especially back in the ' 50s and uh you know the bad men era, it's like where a good ad campaign could completely change the perception about a product.

Also, do you remember when that there was this huge craze for that Twitter thing where you could all go and chat in a room together? What the hell was it called? It only lasted it was a bit of a flash in the pan. Oh, was it Clubhouse?

The clubhouse. Now, now Twitter built their own version. They built their own version. Okay. But there was this thing clubhouse which was available. It was available on iOS but not on Android. Now, what was so funny is it came out during the pandemic at exactly the whole part time as things like the Me Too movement and the whole question of kind of, you know, prejudice and division.

Yeah. And I said, "Sorry, does nobody on iOS have a fundamental objection to the fact that Clubhouse is iOS exclusive?" Okay. Because you know if I if I went and opened up a white Sony golf club, right? There'd be a thousand people outside really objecting to this. Okay, not unreasonably. And yet these people go, well, obviously it's only on iOS. I'm kind of going, hold on, this is this is not acceptable.

More Android users than iOS users. Not more Android worldwide. Of course, the US is slightly unusual because I think it is Apple penetration in the US might be quite high. We we always see violent violent ventures Koreans. By the way, we see every like tech company first release on iOS before Android. And I I was like the one person that said, "No, we're releasing either both at the same time or we're releasing Android." I completely agree.

Yeah. Cuz it just didn't matter. And now obviously there's the whole lawsuits and stuff that are going on about Apple and is it allowed to force you to only pay through like Epic, the Fortnite uh game creator. So there's there's a lot going on around like who controls the system. But what I what I think about with AI and this functional buyer that I am, the majority of people are irrational buyers where we we want something that feels nice. Like we've already made that decision and then we decide, okay, I want to buy it. I'm going to justify it. But in the world of AI, a lot of people are going to go to Chat GPT or Perplexity or another AI tool and say, "Go find me the best toaster to buy and look at do all the deep research and come back with specs." Now, will that AI agent or assistant be influenced by this creative, beautiful marketing ad, or is it just going to be like, I'm going to look purely at the facts and the stats and then tell you what to buy?

It's interesting that because one of the assumptions is people want to make decisions fast. Yeah. And actually there are a lot of decisions dating, real estate, etc. where you want it to be iterative because you use the process of search to refine your own preferences. Okay. I'll give you an example. If you're buying a toaster, you can now buy a glass-sided toaster. Now, it's once you've seen the glass-sided toaster exists, it's clearly superior to ordinary toasters because you can eject it at exactly the right moment.

Okay. Yeah. And so, you know, there is this element of we we embark on a decision without really fully knowing what's really important to us or or indeed how to compare various attributes because most products have incomparable attributes. You know, there's aesthetics versus performance would be an obvious one. They're both important. Um, you got to be careful you don't get overly obsessed as a performance buyer, as a rational claw buyer, because there are things that are important that have a number attached. And there are things that are important like usability, for which there isn't a number. And and the danger people like you will always fall into is getting obsessed with how many kilowatts the toaster is rather than how pleasant it is to use. And yet, in the scheme of things, both matter.

Okay. Yeah, the thing I was going to add to that as well. Sorry, as you jump in there, I was like some of the times the reinforcement of how people act around you or how they treat you based on your purchase is also the thing that drives your decision making. The iPhone in India is a luxury per it's a status symbol. Yeah. To have it. So, so even though you might go with a rational purchase, when you receive it and somebody says, "What are those, right?" Or somebody doesn't give you the almost like the social reinforcement that you want, you might actually adjust the prompt, right, to say that it's currently popular.

Yeah. Do you know what I mean? It'll still in some shape or form fall almost like foul of like social pressures. Yeah. We go out to a social event and we take pictures on different cameras for friends and then my friend on the iPhone will say, "I don't really like my pictures. Like, send me your pictures." Wow, this is so great. Which version of the iPhone is it? Oh, that's that thing. I don't use an iPhone. Like, this is why you should consider maybe using another phone. They're like, "No, like uh I want people to know that I'm holding an iPhone and I don't want to be the odd one out in the group when I send a text." I ruin my family group chats, my friend group chats with the green bubbles and they're like, "You're the one person ruining it for all of us." Drake even had a lyric that said, "Text go green." Right. It's actually my text go green.

You could argue they shot themselves in the foot with an exclusivity and that FaceTime had they opened it up. Yeah. Might have become Zoom stroke team or WhatsApp. Yeah. Or WhatsApp. Okay. So, their their own assholary sometimes I think backfires. Um I have a particular grudge because there was I think it was called Primeonic. very very good classical music uh music app because Spotify is terrible for classical music because it doesn't understand you know movements. It just you know it only understands tracks and albums effectively. And so Primeonic was a fantastic classical music app and Apple bought it uh claiming they were going to turn it into something wonderful. On day one they shut off the Android app which I'd been happily using for some time and then they proceeded to do absolutely nothing with the iOS equivalent. In fact, I think it's now just dead. Okay. So, occasionally they do these things that struck me as so gratuitously malicious to buy an app and immediately close off their use to Android users that I I've had a kind of grudge against them ever since for that small thing.

I don't know if you know this, but the founders of Siri, you know, they they sold Siri to Apple later they came out, built another company almost exactly identical called Bixby and sold it to Samsung. Yeah. And uh it's literally the same product. You played it twice. So that's uh it's it's very interesting. Again, not to harp on Apple. They make fantastic products. Uh but this is the thing that I think about in the world of AI is how transparent do we need to be. So in the past, the saleserson had all of the information power, right? When once the internet came out, people have done 50% of the research themselves. you know, at least something before going to the car dealership, right? Uh to now where you can go buy a Tesla online.

Yeah. And Tesla doesn't do any marketing. Well, Elon's one tweet is equivalent to doing an entire marketing campaign. So, by the way, I would argue just just to defend the advertising industry here. They would sell a lot more cars and at a higher price if they did advertising with people from a Silicon Valley background, it's an article of faith.

Yeah. It's not really a business decision. there's a kind of engineering purity which believes that your product should sell itself on its own merits. Okay. Um and actually there are quite a few I think there are quite a few digital businesses. I mean Amazon is actually one of the world's largest advertisers. We shouldn't forget. Okay. But there are quite a lot of products out there which would sell more and make a larger margin if they did advertise. But it's kind of a tech bro um signaling mechanism. I absolutely was the same person for 5 years and we've now started doing marketing and I realiz I realized wow there are so many people cuz we think that oh everyone that needs to know about fireflies already knows about it either you're using it or not and then you realize there is half the world that is not aware of it and so when we started doing marketing whether it's digital marketing billboard marketing uh the the things we get is like I didn't know this sort of wizardry existed like I I was just like looking for some some voice transcription memo tool and you're giving me like full-on notes and you you're having it fill out stuff. And so the constant education part like for me marketing is like a huge part of setting the context and educating people that don't that don't even know that something exists. Like the awareness and brand marketing is like been a game changer for us the last like

I'll give you a tip too. You could try and I I genuinely believe this from a behavioral standpoint. uh try what I call um actually a few one or two people do it. Kaggi, I don't know if you know Kaggi, the meta search engine. You pay for it, but it actually gives you what you're looking for rather than what somebody else wants you to see. Okay. Um and I suggested to them and they adopted it an honest subscription model which is in any given month if you don't use the product we refund your subscription because I think one of the problems you face one of the mental bottlenecks you face is subscription fatigue which is a lot of consumers out there are once what once bitten twice shy that in other words they've suddenly realized they're paying for things they never use because our bank accounts are now too chaotic.

Okay. And there's no software that'll help you uh do notpay.com which I think is an AI cancellation tool if I'm right. But actually if you if you simply have the the proposition that you don't use it, you don't pay. In other words, it's you make very very clear up front you're not an inertia seller.

Yeah. Okay. Yeah. Your business model is not to get a load of people to forget to cancel after the trial period ends like a gym. Um I could we will never be able to well we probably could do a test to prove it.

Well we actually give our product out for free. There's a free version of it which people use uh and then it's only after they've acred a ton of value they realize oh I'm running out of storage I need to buy more or I want to unlock these new integrations or advanced capability. But for us, our entire marketing funnel for five plus years was you as the user are the champion of the product. Every time you bring it to a meeting, you're advertising Fireflies for for us.

How does it interact with say email and other messaging tools? So, right now it's going to send you the e meeting recaps. Are you saying like can it actually read actually read email or is that everyone asked this question like why why stop why stop just at meetings? why not do my emails and Slack messages. So for now we're doing meetings. Uh but I think it's something that's on our uh radar is if fireflies had the context if you think about relation they are actually relevant aren't they? Because every every meeting is probably the you know, the product of an email exchange. That's what I was partly thinking.

So everything so every communication stream right whether it's an email conversation a slack text messaging conversation uh uh a video call voice call. Yeah. If you can bring all of that together, like the meeting might only be 70% of it, but you need that email exchange and the text messaging exchange. Now, Fireflies can be your like personal relationship manager. I can tell you, it can tell me every conversation I've ever had with Rory, every email I've ever had, and help me prep in a much better way. So, and you even gave me an example when we're speaking of error about the fact that it can also help you with content creation, especially in an era when we're going into personal branding. The reason why a lot of people are not posting is because they don't have the the headsp space, the time to even think about valuable content for other people. But if you have a backlog of a lot of the the stuff that you've spoken about, you might be creating like nuggets and moments that you don't even think about. Is there a way you can like go into those archives and be able to like bring it in front of you to see?

You can use you can have Fred essentially search back through. So like I'll give you an example, two examples. One is like for sales, I can't sit and listen to every one of our sales reps calls. No. So, but I love learning and listening to everything that a customer says. Uh, so I to this day I read every single feedback NPS score that a customer uh fills out in Firefly. So, every day I'm reading hundreds maybe thousand plus messages of when someone churns, why did they churn from Fireflies? Or when someone gave us like a 10 out of 10 versus a 8 out of 10, what what do they want? And so we've gotten to a place where it's just so much information. Now we're having the AI go and read everything and classify it for me. Uh but one of the things that I can do with Fireflies is I want to know how our sales rep is doing. So I'll say what are like the biggest objections that customers had this past week when they had a call with Kim? And then it'll list out the top 10 objections. Hey, everyone wants these new features or everyone wants a discount on this new tier and that informs uh 50% of our road map today.

This is extraordinary because I I knew someone very very astute. I better not name them but when they were product manager for Microsoft back in the I think it was the early 2000s they took the unusual step of putting the call center in the middle of the product development team. Yeah. So it was a closed loop straight into voice. Most people would have put the call center 500 miles away and based it on price. They said, "No, no." The great thing about a something like a call center, by the way, is really important because what you learn from a call center is the things that people can't tell you through another interface because the call center is generally the, you know, the the last resort. You know, it's the last port of call. If everything else fails, I call the call center. Those calls are disproportionately valuable because they're actually telling you not about trivial problems but about insurmountable or really really stubborn.

For the first two years of our company, we had no product roadmap other than taking everything from support tickets uh and using that to drive our road map because people are going to scream the loudest about certain things. And the problem with product managers is we have our own theories. We like to sit in our own vacuum in our room and think this is what's going to be right for the customer and then you go build that and you still hear realize people are screaming about something else. Like why is the upload uh taking 10 minutes instead of 5 minutes. So if we just all we did was listen to customers and build what they needed, we knew that our support tickets would go down and then we could actually breathe at night because in the early days it was just me, my co-founder either writing code, designing or answering support tickets. So we need to do anything to keep the support tickets uh from like uh piling up.

a lot of the times cuz obviously um used to work for a company called Gym Shark that was very like social driven and a lot of the times people thought that the reason why the company did very well was because of what we were posting and putting out there. But actually it was a massive listening aid, right? It was a hearing aid because we would put things out, the customer would tell us what they wanted, what they didn't want. We would see people buying hoodies and almost like cutting them up and then we started to make crop hoodies, right? So, we'll just listen to what the consumer says. We'll stay close to the customer. We'll see. I actually describe social as behavioral psychology at scale, right?

Yeah. Because if you just see how people act in the wild, you're able to understand the patterns and you're able to act off the back of that. And what happens there is customers always believe that you know exactly um almost like where they're going. You make them feel seen. You make them heard. Actually, there's a company called Nandos. I don't know, obviously, Rory, you know Nando. Uh, and uh, one of the earliest almost like examples of this was when I saw a tweet um, uh, because what used to happen uh, back in the day uh, when Nando was uh, for the youths. Uh, uh, we used to go to Nando and people didn't have enough money to actually buy a Coca-Cola. So, they'll buy a a water and they'll go to the machine and get a Coca-Cola. And then the Nando's handle tweeted saying, "All of you that go to the machine with a with the bowl of water and you take a Coke, we see you, right?" And the audience realized, "Oh my god, we're being seen. This is a brand that cares about us, etc., etc." So going back to your point, um, uh, uh, using this technology to be able to listen to the consumer and get to the actual bottlenecks and challenges is actually a great advantage, right? But then going back to that point as well, I think we're talking about the importance of hiring people. I have this analogy which is you should aim to be more Quincy Jones than Michael Jackson these days, right? Conducting an orchestra as a leader, right? So it's like actually you have all these tools at your disposal. How do you conduct your orchestra? And then you can bring in people for um almost like individual tasks and then bring them out for for when you don't need.

this. A generalist idea because everyone is going to be a manager whether you like it or not. whether you're managing people or you're managing a bunch of AI tools like our designer now tells six other tools to do stuff and can create 10 20 different graphics whereas if he were only able to do uh it by himself so you need to learn to manage other tools like that's inevitable. I have this belief that the future is going to be if it's a 20 person company you're going to have 50 AI agents AI teammates and then 20 humans. So you're going to need to have like literally a workforce of just AI agents that need to be managed by someone.

where are bottlenecks good. Now one of the things we got to be alert to in AI is there are certain things which is uh where friction is desirable. For example, litigation. If you literally create a world where everybody has an [ __ ] lawyer in their pocket, okay, which is potentially being created. So, in other words, whereas normally if your neighbor was mildly annoying you because their rotten fruit was dropping onto your lawn, you'd settle it through a decent conversation, you wouldn't go to a lawyer because it would cost you $3,000 and mean you'd have to talk to a lawyer, which is like disincentive in itself. Okay. Suddenly, if you create a world where, for example, everybody can dispute everything. Yeah. effectively not necessarily solving the problem but using the tools of a bureaucracy to make everything unbelievably slow and turgid and and problematic.

Okay. Um you know, there are large elements of behavior where actually the fact that it's difficult is a is a feature not a bug. Yeah, absolutely. One of the things that we've seen with our own product is Fireflies is taking these notes, automating things for you. And there was another competitor that came up where their core summary tech isn't that great on in an autonomous way, but what it does is it lets the user write their own notes and then cleans up those notes a little bit. Uh, and then those people would like shout out the top of their lung, "This is magical. this is the best thing in the world. But it's usually the 1% that are actually going to pull out like a sticky note and then type everything and we're like, is something going wrong? Like why are these like people like saying this is so much better? And then we look at it, 90% of the adoption is still from the lazy folks like us who don't want to write anything, they still want to use it, but the people that are the loudest on social media, it's cuz it's like the IKEA furniture thing. You're putting in a little bit of effort. you you feel like the reward is going to be greater.

I see. Oh, so it's the effort reward heristic. Yeah, because I'm typing my own notes. I'm still putting in all the effort. It takes away from the magic of not having to type anything and just listening on a call. It's rather like the psychology that you the people's enjoyment of a play is sometimes proportionate to how much they paid for the ticket.

Yeah. Cuz you feel a greater obligation to actually enjoy something if you've invested in it. Yeah. Cuz it's like your own notes that you you've written. There's a terrifying theory that that's explains 90% of high fashion. That people like the clothes not because they're particularly good. To be honest, you can get something from ASOS for 15 quid that's just as good. But having paid 400 quid for the dress, you've kind of got to like it.

I recently saw the consumer to go and kill myself. Okay. 75% of high-end luxury fashion is actually bought by lower and middle class. 100% because it's about status signaling.

But but then going back to the conversation around Apple, right? So there's two levers there, right? There's that the status symbol, obviously the fact that it costs more. There's obviously the storytelling element. So the marketing is able to distill something very very complicated in a simple way. But then the the the the third one that people don't talk about is the ease, right? And not necessarily the ease of use. And even though it is easy to use, it's more so it's just so easy for it to be attached to a wider ecosystem. And actually, the reason why you don't want to get out of it is because it'll be too much effort to get out of the ecosystem, right? Because it's attached to your iPad, it's attached to your laptop, it's attached everything's so attached that you're like, for me to be that person that's going to break out of it. It's just so much effort. And I think that's how they've kind of like got you into that ecosystem. The habit changing things that people tend to have.

Salvation lies within. So, a quote from the Shaw Shack Redemption there. There you go. There you are. Samsung Galaxy Fold is turned off actually on podcast. I'm sure Apple's going to come out. We can rescue you from the uh the cult deprogramming. Yeah, before you're all lining up in front of Tim with his bloody Kool-Aid, we can rescue you. Don't worry.

the the thing that when you said about friction and if people can sue others as easily as pressing a button on like Uber and you just end up having that because now the AI is going to triage all the legal cases you can eventually have a slippery slope where a human can't be the arbiter of the final result. So you submit the case, your AI lawyer, whoever uh looks through it, and then an AI powered judge makes the decision and says, "Okay, you need to pay him 100£100. You need to do this." So essentially, there's just going to be too much volume for a human to do. They're just going to delegate it to an AI. That would be a crazy dystopian world. But that's how you like transfer power over to AI.

What doesn't it understand? I I mean what are the qualities where its empathic abilities are limited? Okay. So I mean to what extent I mean obviously it doesn't know how what anger feels like or unfairness feels like but it sort of understands it at a verbal conceptual level. Okay.

Um no I'm I'm just concerned because when you look at automated decisions it always strikes me okay let's take something like car park fines which have been increasingly automated. Okay, there's no longer a traffic warden. It's automatic number plate recognition. Something of that kind. Now the Okay, let's look at six different parking infractions. One of them is you pay for 3 hours and you stay for 3 hours 10 minutes. Okay, is it fair to find someone £100 for that? Most people would say no. Under the terms of the rules for parking, yes, you can. It's not fair. Okay. Equally, you park honestly in a place, you know, four days a week, every week of the year, and one day you forget. Now, they have the data about, you know, you've previously paid them £600 for parking at the railway station, and unsurprisingly, on one day of the year, you're just going to get distracted and forget. Is it unfair to charge them? I would argue yes.

If someone persistently parks illegally and they get caught for a second offense, £100 is probably pretty fair if there's no record of them paying a hell of a lot, honestly. But the the extent to which there's a very nuanced human view and another one would be classically which happens literally which is uh people literally are fined £100 for a ticket which had they bought it legally would have been £350. Yeah. Is that really proportionate? No. Okay. And so there seem to be things which are innate to us as humans which are, you know, a sense of what is proportionate, fair, and just.

sounds like a social score like a Black Mirror episode where you're like, you know, you're not that bad of a human being, so we're going to give you this sort of fine. Well, your track record looks really bad, so we're going to give you this. Yeah. Actually, I mean, that sort of fuzzy logic, I guess, is is actually how humans have leared to operate. Yeah.

And you know, I have this fundamental issue with things like speed cameras, which are automated decision- making. Okay. You can barely call it an artificial intelligence. Okay. Um GPS or satnav would be another example of a very very crude single function artificial intelligence. And we ought to spend a bit of time discussing what's good and bad about those simply because they're they're now long established. And I would argue that, you know, a non-ontextual fine for speeding infractions. Okay. What we have now is in London, they brought a lot of the speeds down to 20 mph. Now, who gets knobbled when you have random speed changes from 30 to 20 mph throughout London. The people who get nobled are people who drive a lot in London, okay? Who don't drive the same route every time.

What they're called is taxi drivers. Okay? Now I can't remember a taxi driver told me the number of taxi drivers who are losing their license or getting temporary bans every month as a consequence of the 20 mph limit. Okay. Now for a taxi driver losing your license has disproportionately painful uh consequences financially. Okay. Is it fair to apply the same rules in that in that environment to a taxi driver or to someone driving at 3:00 in the morning at 25 miles an hour on a dual carriageway? Okay. versus someone driving down Oxford Street. No, we all know it's not fair. And yet, for some reason, what we've had we've got is a world where we have all the data to make nuance decisions, but instead we calibrate the whole system to the strictest possible interpretation of everything.

Yeah. Without with no benefit of the doubt. SATNAV interests me or GPS because yes, it's very, by the way, it's very creative sometimes. I mean every now and then your GPS will suggest a route to you that literally never occurred to you or which is counterintuitive. Equally I do like to look at alternatives because there are facets which the satn about my own personal context which it doesn't understand. You see it might not understand I'd rather take the scenic route or actually I don't want to get there as fast as possible. I want the route with the lowest variance. In other words, I don't want to miss my flight.

Mhm. So I'd rather go on a route that always takes 70 minutes than go on a route that mostly takes 50 but occasionally takes 120. Yeah. So there are I mean I mean the same thing goes I mean my PA goes insane with me because I I will never fully delegate I will never fully delegate a flight.

Yeah. Because there are too many weird preferences and I have which are some of which are tacitly understood not explicitly understood. Yeah, without which I can't fully make a decision. Okay. I mean, you're probably the same buying a new piece of tech. Okay. Uh yes, you you're looking at performance, but you are also looking at a whole load of factors uh which you know, for example, with an airport journey, it would depend on how long I'm going for, therefore how much luggage I have. There are a whole load of variables like what the weather's likely to be like, which would determine whether I take a taxi to the airport or a train or or drive myself.

Yeah, like version has this thing where uh if you want to upgrade to business class, how much are you willing to pay for it if there's an availability for the upgrade? And so everyone has a different threshold, right? It could be based on the duration. It could be the time of the flight or is it an overnight flight? Do I need to sleep? So all these, by the way, weirdly, although economically or rationally, logically, we should say for a given a flight of a given length, I should be prepared to pay X for an upgrade to let's say upper class or midclass. Okay. Weirdly, it also depends on how much we paid for the economy ticket, doesn't it? To an economist, it wouldn't because your upgrade has utility. Okay. But if we've paid £700 for our economy ticket and the upgrade's 300, we're probably Yeah. Go on. Whereas if we paid 300, okay, you know, we're going that's twice the price.

Yeah. It's always comparison, right? Yeah. And that's that's one of my theories which is I think everybody will get AI wrong because they'll try and come up with a single right answer. And what we actually want is a a short list or possibly an iterative conversation where you go have you got anything a bit you know, have you you've got anything a bit more modern or you know, is there a hotel that's a bit closer to the sea. Um because we're fundamentally refining our utility function.

I do see like a convergence like with AI where it classic example of this is when I go on LinkedIn and I start reading posts, they all start sounding the same and they all are using the same phrases that chat GPT would use cuz I think what most people are doing is going to chat GPT and saying I need to post something on LinkedIn uh and so write me uh a post and just the and I'm a human and I can see that okay the sentence structure here. Oh, it's not this. It's revolutionary. Like the long hyphen.

Yeah. The words, the hyphens, the grammar. I can just see it and I I I lose interest. You have to put a joke in to prove it to humans. And I realiz or maybe you need to put in typos cuz it's everything looks the same. Swear words. Yeah. And what I realized is AI sure can write like good content. It can take uh someone with like 70% skills, bring them up to 85. But ultimately at the end of the day day today at least people that are going to use AI to the max potential you need to have good taste so that the AI has good taste. Otherwise it's just it's just going to regurgitate information and it's going to give that information to 20 other people. There's nothing different from what you post versus what they post and it all just starts looking the

There's a client of mine which I won't name although I admire them greatly in every respect and they give all their employees very very good presentation training and it's very very good at the individual level in that you put one foot in front of the other you you make certain hand gestures extremely good except if you have to have five of these presentations in a row from the same client where you go oh here we go again just so this is the feed everyone looks the the post content looks the same. Whereas what I think about is use AI to curate, short list, but you need to have your own unique ideas and then use that as a way to reference or clean up, but not have the AI do the creative thinking for you, at least not now because it's just then you start to wonder how many other people have you told the same lie to. So that's that ends up becoming uh a

reason why it looks so monotonous. How many people? That's a beautiful phrase. How many people you tell the same lie to? Exactly. Yeah. I bet you say this to all the girls is the equivalent of Yeah. Yeah.

The role that AI plays. But, uh, we obviously talked about AI, we talked about remote teams, we talked about Apple, we talked about everything. So Chris, what's the one thing that you would love to like leave the audience with? Like, out of everything that we've said, what's the most important thing that we should know about the world that we're going into?

I think we're going into a world where human connection is going to be a premium. I feel that now you're going to have AI that sells to you, that markets to you, that does everything. I really feel that there's going to be a premium put on human-delivered experiences. You know, the funniest thing when people leave like Goldman Sachs or banking or some, sometimes management consultants say, "Okay, they always like buy a vineyard or a brewery." Because after 30 years of working with intangibles, the joy of just seeing a truck leave the factory full of beer is just, you know, kind of all you know, or indeed, you know, a cellar full of wine. Finally, something tangible and physical. And I've often wondered the same thing about this, which is that the, the only, probably said this before, the only really great thing I've ever heard from a futurologist is the phrase, "There are no trends, there are only vectors." And where you get a trend, you always get a counter-trend. And the example I'd give is the thing I love is farmers' markets, okay? Which are basically the way you create structure. Farmers' market, they're totally illogical economically because you buy direct from the producer and yet you pay a premium for it, okay? Cut out the middleman and pay more, okay? And it's totally inefficient. It takes place out of doors, was you've got to pay nine times. You've got to have a conversation with everybody. Okay. But we, so a farmers' market in one respect is a Tesco Metro that you've made absolutely [ __ ] Okay. But on the other hand, we, we like it precisely because of what it isn't.

Mhm.

And so that you're, I think this is absolutely true. There will be this huge premium on, um, uh, on organic content of some kind.

We tried to automate sales and said, "You know what? We're going to have self-service. Everything is going to be on the website. You don't even talk to us. Just go buy it." And there will be people, usually in the older demographics, 40, 50-year-olds, I know everything is great, but I just want to talk to a human and reaffirm all of this information. So that forces to build a sales, or a lot of times they've already convinced themselves. They know they want this, but they just want the reassurance from a human.

Can I give you the most boring advice? Boring business advice, but I could actually start a business school where one term will be devoted to the single sentence, "Answer the [ __ ] phone." Because the extent to which business is now, right? You ring up a restaurant and because it's all bloody Gen Z, if they're busy, you know, doing whatever Gen Z do, like photographing their own genitals or whatever, okay? Right. Okay. You know, sorry, this probably going to get cut. Okay? They just don't answer the phone. Now, I grew up in a small business. My dad ran a whole series of small businesses. He was a small entrepreneur. It was a capital offense not to answer the phone. Even when I was 11 years old, I had to answer the phone and say, "Mr. Southerntherland isn't here, but I can take a message." Okay? Because you were throwing away business every time you didn't answer the phone. The, the inference of not answering the phone is either they're shut, they've gone out of business, or they're not interested.

It's the worst thing you can do.

Negative brand perception.

And yet everybody, you know, you mentioned the thing about, I think it's an article of faith in Silicon Valley that you don't do marketing. Okay. Because your product superiority is supposed to speak for itself. Like that's ever happened. Okay. Okay. Right. And the other thing is, you know, there's this article of faith that you don't use physical direct mail. You don't use the telephone because anything that involves organic matter is somehow an inefficiency. Okay. It's a terrible, terrible. I mean, actually having a really, really great call center is still one of the greatest things you can do.

Yeah. The, the AI counter to this is if you're a real estate agent, you're on the field or if you're busy and you're a small business.

Oh yeah.

Yeah. This is where I think AI can actually also help take the phone call on your behalf. So just like how you have an AI recruiter, we'll have an AI receptionist and an AI SDR where for we cannot have sales calls with the smallest of customers today, but we will have an AI that can at least have that vocal verbal conversation with you. "Are you open right now?" You know, literally is probably, but the f, the point is that everybody interprets a no reply as as the answer. No.

Yeah.

So, we want to answer everyone. And that's where we're, we said, like, we really care about our SLAs's and, uh,

The, the worst is sometimes people just need to know that, okay, it, your issue is being solved. You've been heard. That's all they need, like, from support. And so we said, let's just have the AI have that conversation with them if a human is busy rather than saying, "Sorry, we can't take anything right now. We don't take phone calls. So this is the, you're just going to have to come back in."

By the way, you are being heard, okay, is possibly the most important sentence in that the rise of populism, okay, in politics. The, the great book to read about this, Dan Davis, the unaccountability machine, but he wants to revive cybernetics, which is, you probably come across it, the 1970s kind of discipline which looks at business through the lens of information flows rather than just money flows.

Which is what we tend to do at the moment. And his whole point is that if, if you literally don't have a feedback loop between the customer and the organization, or indeed the elector and the government, okay, uh, the only, the only signal you can give over that is a scream. Okay. Everything basically resolves to a scream. And that fundamental point about, "By the way, we heard you."

Yeah.

Um, the refusal to actually acknowledge, uh, an opinion, uh, it doesn't matter if you, to be honest, even if you don't act on it, just admitting, "We know it's crap, but yeah."

Okay.

Is is probably 80% of the value of that call.

Yeah. Yeah. And if you think about it, things like the Domino's app, when you order a pizza and it tells you how,

Or the Uber and the map, you know it's coming. It's that assurance that, okay, at least it's happening. And it's the same thing where I wonder in politics, how many people elect an official but then actually are able to reach that official or like, you know what, we should be able to like send them a submit a form saying, "You are, I need you to do this," or "I request this," and that person should sit one day and read all the requests from its his or her constituents and say, "I've heard you and I'm going to address A, B, and C. I'm not going to be able to do these things right now." No one's gonna have that conversation, right? So that's,

This is absolutely fascinating. Yeah.

I think the, the folks, the royal family in Dubai, they, they do this where they sit down and they like go around and get everyone's problems, like just come and listen to it. It, it's like the voice of the customer, like who are you representing? Uh, and then you go and try to address it. So a lot of times you think stats and data and you look at the trends that's going to be enough to address it. But many times you can just go sit and listen to a call, a customer call or a support conversation and the change that's you're going to be able to make is massive. Like Amazon did this with packaging.

Where people were t like sending Jeff Bezos emails about, "Hey, that plastic packaging, it's a pain to cut open, like I'm hurting myself. It's so painful."

And so that's why they switched to the paper packaging.

It's also environmentally friendly, way easier to open, and then you can bulk package multiple things at once. Yeah.

And that's comes from listening to a customer. You can look at as much data as you want on why is this person's ratings five-star, not four-star, but you need the qualitative stuff.

I actually had the Larry David moment once where I bought a pair of scissors in order to open some impossible to open packaging, and the scissors itself came in packaging which was impossible to open.

Yeah. You've got to listen to the customer. That's a great note to end on. First episode of Chris.

Great end. Absolutely superb.

Great shouting. Absolutely.

What a pleasure.