SPEAKER_02 0:08

Welcome to the Future Ventures Podcast on Scaling with Clarity. Today's guest is Dean Fong, founder and COR of Air Task, one of the world's leading local services marketplaces, connecting people who need work done with people looking to earn flexible income. Since launching in Sydney in 2012, AirTaska has grown into a publicly listed company operating across multiple international markets. Facilitating more than a billion dollars in job opportunity team has become one of the most recognized voices in the future of work, marketplace scaling and evolution of the gig economy. Welcome to the stage, Tim.

SPEAKER_00 0:57

Thanks for having me.

SPEAKER_02 1:00

It's an absolute pleasure to have you on board and talk about uh your journey. Uh how did you found um Eric Tasker? What did you do to scaling up to uh to such excess? How did the IPO go through? Um so I'll turn it over to you and just uh tell like walk us back to uh to how it started.

SPEAKER_00 1:22

Sure. So um back in 2012, um my co-founder and I were working at um a company called Amesim. It was a uh mobile virtual network operator, a mobile SIM card business. And um I've been really lucky to be able to join that founding team because at the time I was working in an elite um fashion modeling agency, and it just happened to be that the owner of the agency wanted to start this business, and I was the only person working there, so he said, you know, come and work with me and I'll um and um you know, you know how to do Excel spreadsheets and and that. So um we were working in that business, and I guess like the the really amazing thing was we found that um it it um it proved to us that people could just come up with ideas, uh, raise a bunch of money and then go build up that idea. And so Johnno and I, who had been the first two sort of employed team members of that team, we were like, oh my god, we want to do our own thing. Like we'd caught the bug of starting startups. Um, and it wasn't a big thing in Sydney at the time, like it wasn't, you know, sort of like Silicon Valley is um today. Um, but we were really wanted to start our own thing. So we're looking at a bunch of ideas. Uh at the time I was moving apartments, and I asked one of my friends to come and help me move because he's got a truck that he uses to do for the freeze of his business. Um, yeah, which I think it's a pretty common thing, ask a friend to help you move, you know, like when you're in your 20s, you know, that's a you know, you don't want to go and pay thousands of dollars to some moving company. And so uh he did that, and then you know, we started thinking it's like, why do you ask your friends and your family to help with all these jobs when there's so many people who want to like earn an income and yet, you know, it's hard to know. And what we realized is there it wasn't really like a safe and trusted place where you could go and find someone, like there was Gumtree or Craigslist uh equivalents, but there wasn't sort of like an Airbnb of this or like a you know, Uber of that sort of thing. And you know, those startups were all sort of blowing up at the time, and we saw that um there's a real opportunity to do that because um probably showing our age now, but Facebook and LinkedIn and all these kinds of things were really taking the internet from like um a dodgy kind of place with pseudonyms and forums and you know, Reddit style basically to to the you know web 2.0. Um, and so we thought that we could create that. Um and so we jumped in and we did it, and it sort of took decade plus to sort of build anything um uh meaningful, but but um that was the the starting point for the for the journey.

SPEAKER_02 3:58

Interesting. Uh you and your co-founder technical. Uh did one of you write code, the other one was on the on the commercial side.

SPEAKER_00 4:06

Yeah, so I'm more I was more on the marketing and the business development and the um and raising money and all that. And and and frankly, like raising money is probably like kind of like core skill of like a founder in these early, in these early days. It has to be. Um, and my co-founder was more um operational and technical. Um, and you know, um I he did more the back of the house and I did more at the front of the house essentially. Um it was in in hindsight, it was a really good combination. Like you need those two forces, I think. Like I was the force that was always thinking of some new thing to do, and always kind of like looking at the new opportunity and selling a dream forward, and then you know, a lot of things weren't working and failing and stuff. He was probably more the hey, how is this realistic? You know, how are you gonna actually do that? Um, no, we've got to stick to that thing that we're doing. And I think sometimes I won, sometimes he won. And I think that balancing force is probably pretty healthy.

SPEAKER_02 5:06

Yeah. How big is Zeritasca today?

SPEAKER_00 5:10

So actually, as from a company perspective, we're smaller than where we were uh four years ago, which I think is it's really, really exciting. Um, we scale from sort of zero to 240 people. We're about 205 or so now. Um, but in that same period, we've probably five or six X star revenue and things like that. So I think that um we're getting much more efficient. Um and our business, um, maybe to step back, marketplace for local services, it's a pretty simple product. You can post a task, um, people in the local community put up their hand and they can do the task. The customer chooses who they want to do their task. Um, so it's a true marketplace in that in that sense. Um, and AirTasker doesn't define the scope or the price of the job. So it's purely defined between the customer and the tasker. So we're a real marketplace in that sense. We're not sort of like an on-demand kind of thing. Um, the upside of that is it's quite a light touch and efficient business model because we're not getting involved into all of these logistics and stuff like that. We're purely the platform. Um, and so where our expertise is mostly in um building that liquidity, building that network effect, and then designing how the system works, like the incentives and the um how you know reward different behaviors and disincentivize bad behaviors and all this kind of stuff. Um, but what that means is quite scalable uh as we go to new countries. Um, you know, we have a large team um in the you know the head office building the platform, yeah. But each new country that we're in, you're not hiring teams of tens and twenties and thirties of people, you're you're hiring a team of three people to go out and do BD and and get the marketplace um um ripping.

SPEAKER_02 6:59

Makes sense, makes sense. Um so you know you know I'll come come for the features and kind of the innovation that you're doing it there at Oscar, but um what was like how do you get your first customer? How long did it take you to build at least the the foundation of a marketplace? And then like I mean, if you know in a marketplace it's demand and supply, which side did you work first and which which side was hard to build or easy to build, and and kind of how did you how do you build it? I'm just walking to it because like 2012, I was I was doing that the the first years were the hardest and you would spend quite a bit of time building it out.

SPEAKER_00 7:43

Totally. I think one of the things about marketplaces, if you kind of think about like what is the product, it's in some sense it's a piece of software. Like that's the tangible thing that you can see, it's the app. But really, that the app with no users on it is is literally a useless construct, right? So the product really is the network, the way I look at it, yeah, or the marketplace. And um, and that means you've got to have liquidity on both sides of the marketplace. And so I think, in some sense, one of the things that was really good for us is that um because we'd worked on this successful startup before, we had um we did have some advantage in being able to go and raise what was a pretty big seed round back then, um, is a million and a half bucks. That was big money, honestly. I mean, inflation adjusted, it's probably, I don't know, three or four million bucks, but still, like um the you know, in in Silicon Valley, sort of the signs of the rounds has gone up faster than inflation. So back then that was a lot of money. Not a lot, right? And so, like, that was a lot of money. And so, for from our perspective, I think we did have the right idea, which was like, oh yeah, you're definitely gonna have to go at this for a while before you even have anything to do. Um, but because we'd raised money, we were like, Oh, we're in this, we can't get out of it. And and it I must say it was very difficult because I remember like looking at the the the um the cash burn in the early months, and like we'd raised a million and a half, and I remember in like month two or something, we had a cash burn of like 130 grand or something. And you know, as a as a as a a month and as a 20-something year old, you look at this and you just like, oh my god, like I've I'd never seen that kind of money in a bank account, which I control, like a million, a seven-digit figure in a bank account, I was like, What the hell? But then to see a burn rate of like a you know, you you do have the like we gotta make a yeah, but I was more kind of like you got about six or something, because really, if you have a million and a half, by the time you're down to 400 grand, you're kind of like we're kind of I raised, I'd risk overcropped problem. But but I think that that pressure actually was kind of healthy for us because it was like, well, you can't give up, like there's no chance you can just say, I changed my mind, I'm not gonna do this anymore. Because all of our investors said to us, Look, if you guys fail, we understand, but like you have to fight until the last fight, you know, you gotta give it everything you got and and and that. And so the first few years was really um throwing a lot of things at the wall and seeing what sticks. If I look back and I try to explain it strategically, this is not how we were thinking about it at the time, but like effectively the strategy was that we ended up making a product on experience that was 10x better for the supply side by by definition. And what do I mean by that is if you kind of have a look at how most people get a job these days, there's kind of two routes you can take. There's one, you go get a job at a cafe, like a casual job or something. And generally, when you think about that, that's probably you know, go to a website, see all the different jobs, go apply for that job, get a call back in a week, talk to a person, do interviews, probably takes a few days or weeks to do that, get a job, get a shift, get paid two weeks later. You know, there's a lot of stuff involved between today and money. And what Air Task Effectory said is well, you can just come onto this website, register right now, and right now you can go pitch the guy up the road that you can come and move some boxes for him for $200, or you can, you know, like the job opportunities are like on the platform. And so I think we sold for sort of instant gratification, no friction, no upfront costs. So the most you're gonna lose from doing this on the supply side is you're gonna spend an hour, you know, trying to get jobs, and you don't get any jobs. That's not a very big downside. The upside for you is you find a job and you get paid $200 like tomorrow. So, like that's pretty awesome. And so then it became um, because we'd kind of solved that inherently, it became mostly a one-sided problem, which was that first bit is only true if you can get a stream of jobs coming into the platform rapidly enough. And so that is where we spent all of our time in the first few years is just like trying to work out ways to get people to post jobs. And it kind of doesn't sound like that gnarly now, but it's kind of a gnarly concept to like post a task, like tell people what I want. Like most people are used to e-com is I'm the customer, I've got the money. So I go to a website and you tell me what you'll do for me, and I'll just give you the money. You know, that's what Amazon is, that's what uh Airbnb is, that's what basically all e-comm. But our platform was like, no, you're more like the job poster, you need to define a role description for this person, you know, come over to my house and clean up my backyard, and it's bloody hard to do that. Um, and so we just had to try a lot, a lot, a lot of different things. And and I think like one maybe big learning from this is that there was no single silver bullet. And that is both a good thing and a bad thing. It's a it's a very bad thing in the sense that it's hard to scale it, it's a very good thing in terms of like defensible motes, because it's if there's just one thing that you did, someone else can work out what that one thing is, and yeah, and um, and and that's not a very good business. So um, yeah, that was what the first few years were all about.

SPEAKER_02 13:21

Interesting. And and and now do you have the same kind of mentality? Uh, continuously I iterate, continuously test, and see what works, and make these refinements um to continuously improve the product, the service, the expectations, the experiences.

SPEAKER_00 13:38

Well, the good and the bad news on that is that we're constantly building new marketplaces. So we we constantly have this thing of like in in many of our Australian cities, in say London in the UK, and um um we have established cities where we're we're like, yep, great, we've built them, we've probably got the leadership position, and you know, we would be the you know, a leading choice for for customers, but we've also got brand new cities that we're launching, like literally every you know, day, week, month, new ones. So we're constantly solving that problem. And I think maybe like the philosophical thing about building a platform that I think is really um unintuitive, but like really important, is there's a hell of a lot more that you don't know you don't know than there are things that you do know you don't know. And so basically, um, like if you kind of think about like Slack as a platform, true platform company, you as the creator of that platform, you can't really try and design all the use cases. People are gonna use Slack, however, they want to use Slack. Like if they want to share stupid gifts and that's the main use case, well, you got to just go with that. Um, and if they want to organize lunch, you know, amongst colleagues at work, that's how they want to use it, or if they want to use it for business development, that's how they're gonna use it. And I think AirTask is similar, it's kind of hard because most people's natural um way of working is I'm gonna design this perfect experience, and then I'm gonna refine that experience, and once it's awesome, I'm gonna scale it. And what I think the difference is with platforms is no, you're gonna design a thing, enable a thing, watch what someone does, see what they do, and then if they do something like that, then go where the the action is, and that's a pretty that can be a bit daunting for most people because they're like, no, but I want to define it like that. I just can't do that, you've just got to be comfortable.

SPEAKER_02 15:36

Follow the threads, yeah. Follow the threads, uh yeah, it's um and and so we're thinking in terms of Rails perspective, like you build the Rails for people to be on, and then they can see what what what what works for them and what doesn't. Um what have been some of the edge cases that people have used air task? And you're like, oh my gosh, we have we would have never thought about it.

SPEAKER_00 16:00

So so they actually fall into a couple of different buckets. So there's there's ones that are like just genuinely odd. Okay. We call them like um, we actually have a little website, I'll I'll share the link with you with your audience. It's not like oddly specific, because what we found out is like that a lot of these tasks are just like oddly, really, really specific. And it's quite funny. Like, so so there's an example of a task up there now, which is I need a costume of a German um cake, uh, uh a bunt, um, but but not just a German cake, I want it in the shape of a buntlet, because it's like a small German cake, and so you know, it's funny, like those things are genuinely funny. There'll be there's been lots of like romantic wedding proposals where people will say, I want to do an elaborate, you know, flash mob, you know, when I propose to my wife. So those things are like genuine edge cases, and yeah, they're great because they help inspire people and tell them the story. Hey, air task, it's seriously, it's anything. You can like ask for like yeah, face painter, you could get a tattoo on your arm, you could get you know, just any human thing that a human needs another human.

SPEAKER_02 17:09

Do you guys have a special category for this called oddly enough? Or do they?

SPEAKER_00 17:13

Yeah, I'll serve you, I'll sort of this. I'll send you like because so because this is what I always get asked. People always get like, oh, what's the funniest task? I go, like, I'll tell you what, there's 10,000 tasks a day or something. I'm like, there's probably about 300 of them, which are hilarious every day, you know, like it's only like 0.3% of the tasks, but they're freaking funny. Um, and and and so I think um there's that. But I think what's probably more interesting and and a big part of the business, and why this um this model hasn't been sort of the main model that has scaled before. It's that the more long tail stuff is normal is is sort of long tail but within a regular category. So it'll be something like cleaning, right? Doesn't sound very long tail, except that it's clean my barbecue. You know, I've just had a big party, I got, you know, my smoker needs a someone to come and clean the barbecue, and that's a little bit more specialist. Or it might be, hey, come and clean up after a you know, a wedding event. But the wedding event's gonna finish at 11 p.m. at night, so I need someone to come over and clean up at you know at midnight, you know, or it's um or it's clean out the back seats of my car because my dog has left hair all over the back seats of my car and I need you to clean it. And I guess like the thing about that is like that's not really a service that you know, um, you can just go to you know doghairemovals.com and and book because that doesn't exist. Um and so it's kind of long tail, but the category is still cleaning. Um, and I think Airtask enables like a lot of that.

SPEAKER_02 18:59

So at the top of the conversation, you were talking about Airbnb, uh, and and you and you talked about that what you really focused, the big unlock for you in the beginning was like building out the 10x experience for the supply side for the people that are providing the tasks. Kind of like and and Airbnb talks about the 10 star experience, even though the rating only goes to five stars. What was the what was the 10x experience that you created for the supply side?

SPEAKER_00 19:28

Well, on the supply side, it was the fact that you could get a job and not have to pay upfront. And actually, when you compare that to Angie's list, ThumbTork, you know, even Task Rabbit, there's always like this upfront money they charge you or like upfront process that you had to do. So I think the absence of that was was in itself 10 times better. You know, like I'm not gonna charge you $200 to become a member or to get a listing or something like that.

SPEAKER_02 20:00

So you removed any friction that would be standing in the way of onboarding the customer. It was like, okay, like go go go fill whatever minimal requirements you need to fill out so that you can become a customer, and then from there on, go post the job.

SPEAKER_00 20:14

Yeah. So on the the I I think it's actually kind of like deeply ingrained, sort of like the product philosophy. I think on most of those platforms, first of all, like they have a different customer to where like on Airtasker, the primary customer is the person who's getting the cleaning done. But if you look at how Angie and Thumbtack and all these platforms work, generally the person who wants the job done, they're not the customer, they're the product. They get they're the lead or they're the revenue that is sold to the supply side. You know, so like if you think about like how does Angie monetize, it's more like the plumber's paying for a lead. Um, and so you as the customer, Angie's not that interested in helping you, they're just interested in being able to turn you into like a revenue opportunity for the other side, which is fine. But I think then generally what that means is that um they will compromise the customer experience in the interests of the supply side experience. But I think that that's a little bit not the right way to go about it because that's a little bit like if Uber said, Hey, we're gonna make this really good for the drivers at the expense of like the writers, you know, and and um, or it's like if YouTube said we're gonna make this really good for the content creators at the expense of the audience who watch videos. And I think a general principle in most of these platforms is um you know that the demand side is probably the side that you've got to really um you know make sure is um is is is is working. And so for us, creating that experience on the demand side is is really challenging. But um mostly I think the difference with Airtask is that we don't claim or set the expectation that we In the customer experience. Actually, what we're saying is hey, we empower these taskers who are using our platform. We empower them to create a good customer experience. And you, as the customer, what you should do is only choose a tasker that you think is going to create a great experience for you. And so I would say the analogy to that is a little bit more like a dating website, like Finge or Tinder, which is like Tinder doesn't say, I'm going to make sure you have a great date on Friday. They would say, We're going to show you all these great people in your area that are interested in, you know, having a hookup or going out on a date or meeting people. And it's up to you. You should never ever go on a date with someone you don't want to go on a date with. You know, swipe, yeah, you know, only swipe them if you actually want to work.

SPEAKER_01 22:51

Yeah.

SPEAKER_00 22:51

When you're confident. And so that's what we had to do. So we're probably less sort of like in it than Airbnb. And by the way, I think the Airbnb thing is really interesting because I think that like, I know, I kind of feel like Airbnb has kind of got a huge network effect now. And so now you can talk about all this experience stuff and all that. But I'm pretty sure for the first few years, it was more like we just got to get some liquidity and have a network going. Because until you know, we've got at least like a hundred apartments in each city, yeah, we kind of got nothing to even sell in the first place, let alone having a great experience. You know, liquidity is kind of first.

SPEAKER_02 23:28

It's it's interesting when knowing you unknowingly, you have adopted a lot of the Airbnb practice Airbnb practices. So Airbnb launched in New York, uh, even though the the Brian Chesky and his co-founder were based out of San Francisco. And so what they realized is that the it it the the principal's thinking is very similar to yours, is that the that people are drawn to apartments or uh the or or accommodations that have high quality listings, high quality pictures. So uh I can't remember who uh maybe it was A16 or Sakoya, one of the big guys were one of the original investors, and said, Well, they they they interviewed him, like, Well, where are your customers? And they said, Well, our customers in York, and they're like, Well, why are you here? So they went in in the beginning, they were the ones taking photos of the apartments because they realized that the photos made all the difference as to how often the property gets booked. So they were really uh and again that they come from the uh from a design school, so that was their background, it's like, well, we're gonna make the listings beautiful. So it's very similar in terms of how you grow in scale, uh, Eric Asker. Um, and it and it's all like we need to give people the 10 10x experience, and and now uh that that's a natural transition. Well, you said it uh uh customers don't uh that uh the supply side doesn't have to pay anything. Um I'm assuming that you have driven some kind of innovation in transaction that provides certainty because you talked about the ink factor with the early days of the internet. Uh so what are the things that you have built as an innovation between like in the transaction process between a customer and um and this and supplier in terms of like higher certainty, higher surety, uh maybe um the guarantee of a payment, whatever it is, like kind of you don't want any kind of negative experience. So I'm sure I'm I'm assuming that over the last 20 uh the last 14 years you have built quite a bit of innovation in that space.

SPEAKER_00 25:38

Yeah, I think it can be summed up, and I think most of these marketplaces, in a big broad sense, the bucket is trust. Like you're trying to establish trust between other users so that yeah more transactions can flow and all that. Um, what are the big innovations? So at the beginning, I said the biggest innovation was it's not really innovation, but it's just like find a way to get liquidity into the marketplace. And that wasn't just product growth. I think product growth is really hard to do in a network effect business because you don't even have a product yet. Like I I would describe Airtasca with less than 5,000 jobs per week in any one geography, you don't even have a network. So the the first bit, just go build a network, just work on liquidity. Um the second thing I think that we innovated on was payments. And this was actually really interesting. Yeah, I did credit my co-founder for sort of pushing this because um we wanted to do payments because we knew that we were gonna raise a series B. And we were like, there is no way we're gonna raise a series B without transacting all of the payments through our system. Like, because in the original version of Air Tascar, it was sort of like a somewhat of a classified. It was like you could post your task, you could answer the task, but we didn't actually transfer the money between you. What we would do is like get you to agree that somehow you would pay each other $200 and we would charge one person, um, the tasker, 15% of the agreed price. So we just charge them 30 bucks. Um but that we knew that was not a good um product. And so from a business lens, we're like, no, we're gonna have like escrow payments. Like that's the model that people want to have. And so we built is like a model whereby the customer could sort of post a task for free, the taskers could make an offer for free, but when the customer wanted to say, I pick you, they would actually, you know, put the money into our account and we'd hold that money until the job was done, and both sides says yes, and then we'd release the money out less our less our fees. And that was actually quite a big innovation in the local services economy. Because if you have a look at what happened, like we didn't know this, but like basically customers are always worried to hand over money to somebody until the job's done. Yeah, at the same time, the worker never wants to go do a job without getting paid because then they're chasing the other person, going, like, oh, you owe me $200. And the person goes, Oh no, yeah, now my job's done, and you know, expaces. Yeah, it's like it's a bad scenario. And so that innovation, although we did it for business reasons, ended up being like this massively better customer experience. So I think payments was a big one. Another one we did was um a um cancellations and reliability um uh um product. And what do I mean by this? Basically, in the first version of Airtascar, you could cancel, and it was like there's no downside to cancel. You cancel all the money just went back to the people. And this was like a really crappy experience because we would have like a double-digit percentage cancellation rate. And so we realized like if you want to make the product more reliable, the only way you're gonna do that is if you actually have a cancellation um policy, a system that can um determine who is responsible for a cancellation, and then a fee essentially, a disincentive for that cancellation. Now, this always sounds like counterintuitive, like most PMs are like, oh, I don't want to work on that. That sounds just like a business thing. You're just trying to like make more money. I'm like, no, this is not even about business. This is actually just like it's kind of like the roads. If you don't have speed speeding tickets, that is bad. Everyone's just gonna be speeding around. Like, you think people are gonna stop driving fast on the roads because you asked them to? Like, that's not how a big platform is gonna work. You must have like a $50 speeding ticket to make it bad. And not only that, uh, in Australia and yeah, we also have um demerit points, which is like even if you want to pay $50, if you're rich, you can't just like speed a hundred times and no consequence, you have to have multiple disincentives. So that was actually quite a big innovation. And one of the things we saw is that like cancellations plummeted. So if you want to have a um a good case study in like consumer behavior, it was like a double-digit amount of drop by about 40% when we introduced even just a small fee. I think it's about $20 or something on average. Um, it's enough to just go, oh, you know what, it's a bit more annoying to cancel than to not cancel. And so, you know what, I'm gonna put in that little bit of extra effort. You know, if if before I was, you know, because what we'd find is taskers would say, I'll go for 10 jobs, and then um the highest paying one I'll definitely go and do, but the lowest one, forget about like if if the if another one comes in that's better, just get rid of this one. Now that you're chart getting charged $20 for that, well, you're gonna think twice about doing that, you know, and and and so I think it's it's that was uh actually sounds like a small thing, but it was actually quite a big innovation.

SPEAKER_02 30:53

That's awesome, that's awesome. Um I mean, we live in a trust economy, so um, everything is trusted. I mean, even now, like if you look from an AI perspective, is like these all of these target that have started in terms of like how do we validate that the agents, the AI agents, uh can be trusted or they can trust one another. Um, so I I could I couldn't agree more with you in terms of uh the way that you're thinking about it. Um and and and this is no different than like if you're booking accommodations. Um if you can cancel to the last minute, boy, you're gonna hold uh maybe two or three bookings. But if you know that you want to be charged, well, you're gonna think very carefully in terms of what you hold.

SPEAKER_00 31:38

I think um one of the things we talk a lot about is like um uh transparency and accountability. And generally, and these are generally like not popular or intuitive framings, but like in order to have accountability, what really does mean is that there's reward and there's punishment. You know, like without the punishment side of things, you can't really have accountability. Like you can't just only reward people. That's usually kind of odd because the way that like the e-commerce platforms have trained customers is there are no punishments. Like, for example, like my my I have a big um disagreement with the concept of free returns on e-commerce, like that that's just like a bad system, right? So, of course, I'm gonna order 10 things, return nine. You know, not not of course, but like I can do that. The fact that there's no consequence for that, it's actually sort of a broken system. It's like, well, now as a customer, I don't do that. But if the person next door does that, I'm kind of subsidizing all of those shipping things happening, and I'm paying for that. Like, I'd actually rather if this website charged five dollars return fee, you know, just like a small amount of money, something small, to discourage that behavior. Because who wants to have literally aeroplanes flying across the world filled with shoes that are definitely getting returned? It's kind of crazy. And actually, I think it's quite an interesting point you bring up about AI and agentic commerce, because like that is something that you know we're thinking a lot about at Air Tascar. But I think one of the downsides of a gentic commerce is it the cost of the agents running around is so low that now you've kind of lost this accountability factor. Like, if you kind of think like, why do you want to not have an AI flying the aeroplane in your pilot? You kind of want to have the fact that a human is very incentivized to not crash the plane because they also have aligned incentives, like I'm gonna die as well if the plane crashes. Whereas if that's like an AI system, there's a little bit of like, I don't know, maybe Delta Airlines just says, Oh, you know what, that plane's gonna crash because um, you know, um the computer said so. Um you lose that sort of symmetry of reward versus punishment. There is no punishment for the AI, you know.

SPEAKER_02 33:54

I agree. Uh I mean what what we're seeing, because we we work with a ton of AI companies, what we're seeing is on average, AI um or AI-based product, let's say uh uh autonomous driving, because that's the most common-sided example. Autonomous driving on average is uh to a magnitude of uh several times safer than human driving. Um in terms of statistical probability of occurring. Now, the challenge that you're describing is it's those edge cases when something happens, what are the guardrails to prevent it from happening? So the consequences are dire if and when it happens. And so because of that, it's kind of that's the reason why it's so slow. That's the reason why it's it's uh this this inherent distrust that uh that you can rely on AI.

SPEAKER_00 34:53

Yeah, I think if you kind of apply that to like maybe some more you know um e-commerce things that are around today, you know, talking about like an example where you could talk to an AI agent and say, hey, I'm you know, I'm hungry, you know, go go sort it out for me. And you know, brings you a meal and all that. I think the fact that um, you know, that's not you making that choice. There's like there's a lot to unpack in that. You know, like what if the agent sends you something you're allergic to that thing, and then you eat it and you die? Well, um, you know, you didn't take on that responsibility. You sort of thought you were giving that to the AI, um, but you know, you you've lost agency in that in that in that process. So I think it's gonna be a lot of interesting things that happen because if the cost of doing is now zero, that's a bit of an issue. Like another example is like, you know, when you look at like um Sora, that that social network with all the AI videos, it's like if the cost of content creation is like literally zero and I can just make like an infinite number of videos with an infinite number of variations, um, I guess you're gonna end up with a different kind of content ecosystem versus an ecosystem where effectively it used to take me two hours to like make this video and all that. And I'm not saying like one is better than the other, but like you're gonna get some weird and different kinds of outcomes, I imagine, the two cases.

SPEAKER_02 36:17

For sure. I mean, just just just to um this is personal experience. We got into the AI game really heavily, probably about two years ago. Um, and so we have a number of different 20x accounts with with Cloud. Uh, we we use all DLMs. Um, what we're seeing is things are consuming higher amounts of tokens progressively all the time, whereas now we're going to crazy kind of uh gymnastics to to compact messaging to clear uh caches to make sure those uh uh that that it doesn't consume as as many tokens. And the other aspect we're seeing is the cost per token, even though the the ability to generate the tokens has gone lower, the cost per token is going up in terms of what's being charged. Um, and this is, I mean, Clock has said that they're gonna be uh breaking even or starting to be profitable in 2028, which is really literally just around a corner, 18 months. Uh OpenAI has said they're gonna be profitable by 2030. So the cost will go up dramatically uh in terms of doing anything that's AI based. Um and so I feel like I personally feel like we're in a bit of the gold rush at the moment. So like we and my team were just like like kids in a candy store, we're just so aggressively trying to capture as much as possible because I think it was like past 20, past 2030, of course those opportunities are gonna be there, but it's kind of like the early days of the internet. What was possible then is not gonna be possible now. I mean, like you give me an example about a restaurant. I use agents to book my reservations or to book a rental car or to book now. They have my history, they have my information, they have my credit card, they can go and do this. Um, what I find absolutely shocking because you give the restaurant example, very few restaurants have the ability for an AI agent to go and make a reservation. And so the information is not exposed in a way where you can make it for reservation, or it's like callers for reservation, or the menu is not shown, uh, it doesn't have the caloric information, it doesn't have the ingredients taste. Like it just it becomes really limited. So it's like it I'm like if you want to play in the agentic economy, you need to make it easy. Like you give the example of friction. You wanted to remove friction for the supply side. Well, it has to be the same side, uh the same for any business that relies on agents in terms of bookings or customer acquisition. Like you need to remove the friction and you need to make it you make to make it such that it's a it's possible with AI to book something or acquire a customer.

SPEAKER_00 39:12

Well, that's interesting, isn't it? Because if you kind of think of like maybe AI as its own network effect, I guess OpenAI has got to make it, you know, or whoever that that that um agentic service provider is, has to make it easy for that restaurant to upgrade their website to be easier to be read through. Because like in in some lens, you know, the guy who owns the restaurant, he's busy making spaghetti and meatballs. Like he kind of doesn't have you know time to go in and you know, upgrade my website, put in like a you know new protocols, upgrade all these things. And I think it's a great point because you know, in some sense, things are moving so fast, but in some other sense, if the rest of the world is just sort of left behind and they're not really into it, I mean, that's actually not really an open AI or anthropics sort of interest that no and and and that's probably why it made sense to go down coding as a thing, because everyone who's doing coding definitionally is sort of gonna upgrade themselves and get into it. But I think you bring up a great point. Like the guy who owns like the spaghetti restaurant down the road, uh, who has a website that he uses to take some bookings, might be take a bit longer for that person to come across to that.

SPEAKER_02 40:27

For sure. And and it's some of it is mentality is like, oh, I don't want to post my my menu because my innovation comes from the dishes that they create. And I'm like, that's great. But you you you you kind of like it comes down to trust. You you you kind of need to expose someone that thinks that that uh that they're necessary for an agent to take action. And what I love about you is that you immediately you went to a solution. It's like, hey, we gotta solve the problem on the customer on the supply side. They don't have the right platform, the right problem.

SPEAKER_00 40:59

I'm I'm I'm I'm uh I'm a middle ground on AI because I'm a little bit like I think it's amazing. Like as a technologist, I think it's amazing. Yeah, at the same time, it's probably 50-50 as to like, is the world a better place post all of this change? You know, like I think there's like a hell of a lot that you're like, you know, I think a lot of these technologies are like this, you know. You kind of like, oh, some of the things we do now, I kind of like them, you know, like and it's this odd thing where it's like we can automate all of these things, and they're usually things that we say that we don't like, but I often wonder as to whether people actually did like them. Like we have a lot of engineers in our company who are like, oh, it's kind of sad. Like I kind of like writing code, like I find that interesting, and I find like, you know, it's like the carpenter. I I I liken it to like the carpenter, you know. Yeah, I used to make a table, I used to love like chiseling the wood and all that, and then IKEA comes along, you know, he he's like a million tables for like five dollars cheaper, no need to make any tables. And some of the carpenters would be thinking, oh, that was quite enjoyable. Like they probably used to complain to their family, oh, I've got to make do all this carpentry all day. They probably complain, but really in their heart, I think a lot of people are oh I love love making, like making things with my hands. And so I think that there's gonna be quite some and it's it's very, very interesting. I'm sort of 50-50 on on um it, but it seems like inevitable that you you you've gotta go with it, like every technological way you've got to go with it.

SPEAKER_02 42:35

So well, I mean, what we're observing from uh from from talking to um the different founders and and different customers is we're seeing that AI in the hands of the right person becomes extremely powerful. But what we're seeing is uh of course there's gonna be the orchestration layers with uh agentic workflows, and we're using this within our companies as well. But um it's kind of like music. If if if you are a musician, you can use a AI to produce amazing music. If you are a developer, because you gave me the example of a developer, you can use AI to produce code. Like when when we write code, especially since uh Opus 4.7 was released, we've seen the code quality go down significantly with with Claw. And so there's no substitute for a human to go and review and make the corrections that that clock has made. Um, codex at the moment is a little bit more precise, but like I don't see, at least I'm not seeing at the moment like a possibility where it just in highly automated, highly structured, highly predictable tasks. Yes, AI can take over, but I see AI purely as an enabler rather than as a substitute. I mean, look at the number of code number of developer jobs have gone up, the number of radiologist jobs have gone up. So it's not This place can it read a mammograph? Absolutely. And and and like a friend of ours, she's a partner in the radiology clinic. Um, she's like, Yeah, we run it through AI first, but then we have a human read and interpret the results. So the number of radiology jobs are going up, even though the prediction originally was like 98% of radiology jobs are going to disappear because AI can read the pictures.

SPEAKER_00 44:23

Yeah, I'm full, I'm fully with you on that. Um, I imagine that radiologist is gonna be really, really busy too. And I I guess what must be happening is more people will say, you know what, I'm gonna get that scan.

SPEAKER_02 44:36

Exactly.

SPEAKER_00 44:37

Maybe in the old days, but they'll be like, yeah, the costs going out. Um, I also think that genuinely that radiologist is probably gonna be really busy. They're probably gonna be like, my life is so busy, and so I do wonder as well, you know, like it's kind of not making our life easier, it's kind of making our lives everything's hyper-productive, I guess, but that's hard.

SPEAKER_02 45:00

I mean, this is the the other common trend that we're seeing across uh uh at the moment and across AI is that everything is going from one to many or from macro to micro, one-to-one. Um, like a high degree of precision, like for example, uh, I don't know if you're there's an LLM model called Pi. It was the company was started, uh Inflection and Pi was started by Reef Hoffman, the founder of LinkedIn. Pi is your personal uh intelligence, it's it's your own personal model. So if I want to have a conversation as if I'm talking to a psychologist or an advisor, this is what I use. It just me, it just knows it knows just me. And so this is an example of like highly, highly personalization. You talked about like radiology tests. We're working with a life sciences biotechnology company that has developed this highly innovative uh blood test to detect uh cancer with four times higher precision. Well, before, in order to go and let's say to get prostate cancer screening, it was highly invasive. So uh it would take a lot of effort. The the false positive rate was really high, so people wouldn't do it. Well, now you can go and get um a blood test for $150, a test, and you can do it every quarter if you want. So you move from like React, I'm gonna catch it, and hopefully as early as possible, to know like I'm gonna catch it as early as possible because I'm gonna test myself like every quarter.

SPEAKER_00 46:25

Yeah, that's really interesting. I guess, and then that probably has its own flow and effects, right? Like, um, you know, either you're gonna need more psychologists because more people are gonna be like, oh my god, or you're gonna need more doctors to fix more things, or for sure. Yeah, I'm sure there's a lot of flow and effects to that as well.

SPEAKER_02 46:42

For sure. So, I mean, my my personal perspective is that we're going to see an evolution of the roles that exist here. And and I'm not saying jobs, I say roles because I see jobs as a collection of of tasks. And and so that the the task that you a role holds now will be different from the task that a role holds in the future. And maybe some jobs are completely going to disappear, but new will be creating their space. And so uh when we're talking about reversal basic income or other things, kind of like to save the humanity, I think that it there's gonna be pain, for sure. There's gonna be pain in terms of how we evolve, but then it's going to be better, there'll be more jobs, there'll be higher quality jobs. Uh, but I'm curious from an Airtasker perspective, what have you seen? What has been the impact of AI or none? Because a lot of things that you're doing are human in nature.

SPEAKER_00 47:40

Yeah, so um for AirTasker, 95% of our jobs are in-person physical jobs. So they are like, you know, come and move some boxes for me, paint a fence, fix a leaking tap, you know, things like this. And so you see much, um, we don't see much disruption directly. I haven't looked at the Dana about like, say, Fiverr and work and these sort of like a more digital jobs, but I'd imagine it's more impactful in that um in that space. Um, yeah, task, I think, is like quite well protected from that perspective because you know, until humanoid robots come, and I think that's a quite my guess would be that from a technical perspective, it's quite far, you know, quite a lot of work to get to that level. But then probably more importantly, it's like an actual adoption level. Um, and if I kind of look at the task of driving, seems to be fairly sort of two-dimensional, somewhat standardizable thing. And it probably took us 10 years to get there technically to your point to have a um, you know, car that's facing a human, but it's probably gonna take another 10 years to get humans comfortable with even on scale and at mass and then localized and all that. So I think humanoid robots are quite um far away from becoming like a regular replacement for human uh labor. Um, I guess we also look at it as like as a marketplace. Um could will there be like a genetic product that um that replace like the concept of like a network effect and and a marketplace? Um and I think that that's pretty hard. Um you know liquidity and network effects are kind of like a pretty powerful moat um in a business, I think. Um but we are certainly on the lookout for what is that sort of killer app that might even, you know, might one up, you know. And I suppose Airbnb must be looking at this. Like, what's the killer app that could be like, you know, finding a place right now in Rome, you know, uh and just replace the idea of you know browsing homes and finding the one you like and booking it and all this kind of stuff. Um so we are um uh we are um looking into that. And then I think the um the third thing for us that's been really, really valuable is that we aren't a software business, like we don't sell software, but we do have to make a lot of software to activate the value in our business. And that just got a hell of a lot faster, cheaper, more efficient. And so our ability to be able to like roll out features and features, testing products, it's it's awesome. And and the great news is when you build them, it's not like this commodity that you know, we're not sort of like a Canva or a um Salesforce or something where the software is the product. Um we're we're more kind of like software enabled. So yeah, that that's actually been really, really good for us. I guess interestingly, like one thing that I've just seen is the people who can use the right tooling are going absolutely terrible lights. Like it's in like we've got quite a few of these like sort of 10x employees, you're just like, oh my god, like this person's just yeah, so good. Um, but it can create a bit of a hard situation inside an organization where you've got some people that are sort of not quite up to the speed, and then others that are. And I guess it's our job as leaders to try and get everyone up to that speed, but you know, it's not an easy thing to do. I don't I don't think so.

SPEAKER_02 51:07

So, Tim, let me uh ask you a different question. I mean, like you you talked about trust in the marketplace and how vitally important it is. I'm assuming that some of the other things that you have done is kind of rewards and incentives for for the both sides. Um companies have built an incentive program for their employees to maximize the use of tokens. What are you doing from an AI adoption perspective?

SPEAKER_00 51:36

You know, I find that quite fascinating. So um, first of all, maybe what we are doing. I do think that it's important to have sort of like the horizontal bottoms-up approach of hey, here's all these tools, they're all available to you. We've security checked them, we've cleared them from that perspective, and go for your life. I do think that's important. We also think it's important to have sort of like a top-down and verticalized approach within the org for AI too, which is usually against my principles because I much prefer sort of everyone's enabled version. But I do think through this sort of like high flux transformation period, it's probably gonna be better to set up a team of three or four people whose job it is to go through the org and go, hey, finance team, this is what you're gonna do. I'm gonna set it up for you. Now you need to do it like this. Um because we just noticed that, you know, I'm I'm not picking on our finance team or anything, but like um they're operational, they're doing things every day. They don't have time to be also exploring new tooling, you know, setting up systems to enable that new tooling and that, because they're just trying to run the business. So um I think setting up a verticalized team is like a pretty important one. And then I must say, in terms of like um encouraging people to just spend money um or like uh you know, uh ingest tokens, and maybe I'm a newbie on this. Um I'm a little bit like, wouldn't that be awesome if you were anthropic or open AI and you had this whole group of companies out there who are all going, basically, I'm incentivizing my company, my my employees to spend more money in your business. Yeah, I'm like, that's amazing. Because I would have thought it's more like outcomes driven or outputs driven versus like tokens driven, you know.

SPEAKER_02 53:23

Um completely agree, but but I I completely agree because I mean ultimately it's not how busy you are, it's how productive you are that matters. And so I I completely agree with you, but like the the example that you're bringing is really important. So um three three years ago, I was part of a public company, and at that time was part of the AI task force. Essentially, the the governance group that was rolling AI into this company and and what we're going to do. And you give the example around finance. A lot of people, it's kind of like um the reason why houses are staged is because people have a hard time imagining themselves in them. And it's it became the same. We have to stage the company or stage the house uh in such a way people can imagine. Like you gave the example around finance. Around finance, it's like, okay, well, we're gonna use AI to create to create the management discussion and analysis document, create AI to use the to create the corporate sustainability report. We're gonna use AI to like take our peer group and and run the peer group through AI so we can see what kind of risk disclosure they have and compare it to us. We're gonna use AI to drive the investor investment disclosure. So we just all right, so case, case, case, case, build it up, show them, and then it becomes like, oh, I see in California they're doing this uh regulatory reporting from an air emission perspective, we can do that too. So in some cases, people that know they started to write their own Python call them, but that's amazing. Like, but but we had to show them some of the possibilities, like, okay, well, let's build this case, let's build this case, let's build this case, and just to be able to kind of plant the seed, plant the kernels into for the for them to like, okay, I know what's possible now. And then then we started to build like the the uh a groundswell around the the champions, the people that for the early adopters, the people who were like champion this, and it's like okay, well, let's wrap things around because success we get success. So the more cases we have from within, the more likely is that it's going to get adopted.

SPEAKER_00 55:30

Totally agree with that. I mean, you gotta inspire people essentially. Yeah, one inspire to give people permission. Um yeah, and I think I think that is right. Like, you know, without being able to see it, it's kind of hard to believe that it's true. Or you know, maybe maybe there's a difference between believing it's true and then actually really believing it's true, you know. So um, yeah, you just gotta get some action and some examples out there, I guess.

SPEAKER_02 55:55

Absolutely. Uh Tim, I know we're almost at time. Any final parting thoughts? Um, what are you what are you what are you doing to scale yourself, to scale um your leadership team, to scale the company? Um, how should people think about this? Like what's next for Atasker?

SPEAKER_00 56:12

Well, um, what's next for Atasker is we're really focused on growing um uh some of our international markets. So in Australia, we're you know, sort of built in market leadership position. Um we're um you know halfway on that journey in the UK, and then we're early on in that journey in the United States. So that's really like where my focus is. So it's almost back to square one and relearning how to be a startup again in many ways, like unlearning some of those corporate things that um sort of build up over time and doing that. Um and then um uh I lost the audio team. Oh, can you hear me now?

SPEAKER_02 56:49

Yeah, I can't hear anymore.

SPEAKER_00 56:51

Oh, you you still can't hear me? Oh, it might be your pods because I can hear the uh maybe it's on my end. Yeah, I can still I can still hear it.

SPEAKER_02 57:07

Perfect.

SPEAKER_00 57:08

Okay, okay, awesome. Yeah, so um really focused on expanding into to new markets and leveraging um the platform uh that we've built in Australia and and trying to explore that. So learning how to be a startup um again. And then I think you mentioned a lot of this like um technical um uh revolution that's you know, everyone's talking about AI. I think we are taking a pretty measured and intentional approach to it. I think there's a big upside in sort of operational efficiency. Uh the big thing for us though is more the sort of existential product risk of like, what about that thing that comes that just does what you do but does it better? Um and I think that is something that I've definitely um, you know, is in is in my um peripheral vision as we're trying to run a business, is also like, you know, where does that um that next killer um app uh come from?

SPEAKER_02 58:03

Fantastic. I absolutely love the conversation, love getting to know you. Um thank you so much for sharing your lessons in terms of how you went about building AirTasker and continuing to build it and expand it geographically to what I'm assuming is like you're looking for global donor uh global domination uh in terms of uh tasks.

SPEAKER_00 58:26

One step at a time. One step at a time.

SPEAKER_02 58:30

Fantastic. Thanks, Dave.

SPEAKER_00 58:32

Okay, thanks for having me.

SPEAKER_02 58:34

My pleasure.