The Truth About AI: Freelancers & Why Systems Matter More than People Think with Chris Dumpleton

0:00We think the language model is these huge, vast libraries. These companies are going into certain shelves, certain coloured books. They're far more interested in the outcomes and the technology, the labelling software, and they're less likely to want to be involved with the manpower. So we can see a natural relationship between companies that need vast amounts of subject matter expertise in a burst of activity as well, and needs them across the world in different professions. So freelancing is the best way of tapping into that type of resource.

0:35Call centers are at the front line of generative AI. Salesforce just announced thousands of roles being cut after implementing new AI tools, and every brand is asking what comes next. Few people are in a better place to answer that question than Chris Dumbleton, executive at Limitless Technology, who spent 20 years in telecom and contact centers and is now bringing innovation using the gig economy and AI approaches to global brands like Microsoft, Sony, eBay, and more. Now, the same gig experts aren't just supporting customers. They're helping train the next wave of AI itself. Chris, welcome to Expert Intelligence.

1:09Thanks, Paul. It's good to be here. So you can't miss a headline that says AI is replacing call center jobs. It just seems to be the kind of one place where AI-driven chatbots are going to replace humans and it's all going to make it okay. You've been in this space. What's real? What's really happening? Do you know what I heard the other day, which really landed, was, first of all, I think everyone's probably a bit tired of hearing this phrase, which is, AI is coming for the simple stuff, and then all the complex stuff is what's left, and trying to figure that out. So I tested that theory on a consultant I know, and he said, you know what, that's what I hear all the time.

1:50But what's real in the contact centers is the team that's being asked to use AI to remove some of the headcount. I'm like, well, I don't want to lose my team and I've got to still pay my wages. They don't want to do it. So there's almost like a paternal control problem with the people that are being asked to go and implement AI by removing costs from their front line to go and pay for it. And then the cost benefit payoff on how long it takes. So there's no doubt what it's capable of, but the actual applicability, and this is someone that I know really well. And he's one of these dudes that is in front of loads of contact centers, assessing what they want to do, implementing people process technology changes and all sorts of stuff.

2:37And this is the overriding thing he's hearing. It's widespread. It's across every vertical. Now it's about people actually struggling to implement it because of that challenge. One of the things I experienced in working with gig economy models in general of bringing external people in, kind of changing, to your point, like trying to change organizations at a fundamental level, was people's desire to try something new. You have a concept, GigCX, and one of the things that you've been working on for quite a while is bringing gig economy strategies to the call center, which is in many ways equally as disruptive as AI and sort of faces those same things.

3:16So when you talk about the Uber business model for contact centers, help explain to the audience what that is and how it works. So that's my classic elevator pitch. And I've even tested it out in an elevator and it worked. So the simplest way of describing what we do is taking the Uber business model to the contact center industry. And everyone kind of gets that straight away. More often now, we're talking more about sort of crowdsourcing CX rather than gig. but it's the same principle which is no one's employed we're using freelancers so we go to big organizations and we say that we've got this idea for you which is rather than using insourced or outsourced contact center staff with you know mega high attrition minimum wages you're constantly dealing with the seesaw of cost and csat where to apply automation where to keep people and trying want to keep the lights on is we're going to route contacts to your own customers that we're going to source and we're going to invite to become your experts and we qualify them verify them train them exactly the same way as in-house agents are but they are a freelancer and then the magic happens and this was my question right so when i first joined limitless i asked the same question as our clients ask now when we talk about this model i was like so we're not going to employ anybody that they're going to turn up for work and they're going to do just as good a job, if not a better job than someone that's in a contact center.

4:41The answer was yes. And I was like, okay, I'm in. And then six years later, this is me now convincing everyone that that's the way it works. And that's how it works, you know, at a high level. One of the case studies that I read was about eBay. And being an eBay seller is a very specific thing. You have to have knowledge on how to do it, how to use the platform, best practices. When you talk about a platform that has customers or people that are engaging in the platform, making money on the platform, being able to talk to other people who do the same thing. How is that different than me just hiring a contact center?

5:14The eBay story was one of our favorite case studies. So I think we can all attest to this, right? Which is that if you need to contact a company because you've got a support issue or accounts issue or whatever it might be, right? So unless it's a trivial thing like a moving house or administration or something like that, if you want to get some actual expertise, you want to talk to someone that knows more than you do, then I think it makes sense to talk to someone that's already got the product or service because they're best placed to be able to help you. That's what we go after, and that's what eBay did.

5:53So we tested a delivery model with those guys where the seller buyer sellers is a three-sided marketplace. So with sellers, they have an extraordinarily high drop-off rate with sellers who register to become a seller who then go on to become hooked on the eBay drug and carry on and become prolific. So that drop-off rate is – they probably won't like me sharing the number. so I'm not going to share the number, but let me just say it's really high. It was really high. So we got involved with them and said, well, you have millions of sellers. And this was in COVID that kicked it all off. Everybody then registered to become a seller on eBay.

6:34They were having millions of new registrations a day. But then when they run the numbers in six months' time, how many of those sellers went on to list five products, sell five products, stay for six months? The number was single digits from all of that. so what we enabled is for very experienced sellers to help the new sellers they're literally coaching them so when you log uh to become a seller on ebay platform you go through a seller onboarding flow and during that flow we put a little widget which said do you want to talk to an expert seller and the expert seller would coach them on their listing on their photos on where to list pricing everything like that and they saw you know some massive uptick in some of the metrics and it was a b tested so it was like yeah ebay is whilst it looks like an online marketplace is a data science business so behind the luck behind the scenes is some very clever people that are working out how best to optimize the platform and to gain the data analytics to show what people are buying where they're selling how much is good right and it was the data that led the results so they had a 24 increase in sales revenue from those that went through that process 20 more new or reactivated sellers and the cost of them handling those chats was just under 20 less than they were in in the actual contact centers themselves so it's a great metric it won one of those prestigious gold awards at the european contact center and it was the only year so i've been going to that award ceremony i used to sponsor it in my previous organization used to come out of my marketing budget and so I've been going to that for years and years and years and this was the first time I was kind of like shoulder to shoulder with a client who was up for an award and it was the year of COVID when it was remote so I'm there in a tux in my kitchen or this online thing and then it got announced and we're all just jumping around stuff so I didn't get we didn't get the on stage moment and the you know the Oscars speech and all that sort of stuff but um yeah it won you know won Best Customer Engagement Award.

8:38It was great. It was a wonderful time. It's good to get that recognition for the model. One of the interesting things as we start to talk about AI is outcome-based pricing. Yeah. SaaS is you pay your$10 a month or enterprise-level thousands of dollars a month for a SaaS platform. Use it if you don't, you don't. In a lot of the generative AI products, you pay for the token, and it's really outcome-based pricing. Tell me a little bit about the Gig CX model in outcome-based pricing versus, hey, I just hire a group of people to do support. And to your point, I pay those people to do that work. This is kind of the real essence of the gig economy, which is the difference between getting in a black cab in London, I don't mean any judgment whatsoever to black cab drivers, but the difference between getting in a black cab in London or through a taxi firm versus an Uber driver is the Uber driver is, he wants a really high rating because the rating is what drives additional work, preferential rates, further opportunities, continuation to work, better earning opportunities, that sort of stuff.

9:46So that's the same sort of principle as we have when it's all to do with outcome-based pricing. We don't exist as a business unless a question from a customer isn't answered correctly by a freelance expert that we've onboarded. If we can't make that happen, then there is no business model there is no limitless there is nothing this business we've built since 2016 doesn't exist so the difference between me rocking up to a company and going i'm going to sell you you know a thousand contacts into licensed software seats or tokens or whatever it might be something that you know you pay whether you use it or not doesn't matter you know i've sold you a license it's down to you do you you know obviously there's a a big desire to make sure the customers do because of retention rates and stuff like that but in principle i can sell you what i think it's going to do people go yeah i want to buy that and then whether you use it or not is kind of it's after the event outcome-based pricing is where you have to trust that there is no business unless there are positive outcomes so it's the most optimized way of commercializing a delivery, it means that we have to put all of the skin in the game in terms of the risk, if you like, because we're not asking people to pay for platform licenses and things like that.

11:07We're literally just giving people a price for outcomes. And it's those outcomes that drive the continuation of the program. I've built a couple of these systems, expert-based systems. And one of the things that I always got from people was the biggest pushback. So they're not employed. You're not training them. No one's telling them they must be here. But why would they show up? There's this idea that the only way to create that value, hey, here's a customer with a problem or somebody who needs something, and here's an expert that can help them. The only way to do that is a traditional model where you go to a staffing firm or some big company that hires them, trains them, and onboards them versus passionate experts that want to show up and help people.

11:53How do you overcome that when you go into a client who's like, hey, I'm interested in learning more? We have to basically start with a pilot somewhere. But we always say, look, firstly, we'll go back to that original point, which is that there is no business unless we can prove this works. We're not trying to get you to remove everything that you're doing today. But if you give this a shot, give it three months and you see the outcomes, what you're going to get, we're pretty confident you want to double down on that. It is a proof based whole delivery model. For me, the answer is like, so my first ever job was in a contact center.

12:25So I think I might have just turned 17. I was definitely 17. So I remember I got promoted when I was 18. So I was definitely there when I was 17. And what I can say is it was a job, right? I was 17. I can't say I had glorious career ambitions at that stage. I was interested in acquiring some beer tokens, and that was about it. It was a job that allowed me to do that. In the end, I absolutely loved it. And here I am, only a couple of years later, still growing up in the industry. But I didn't have a passion for the brand. It was a job, and I enjoyed my workmates and the banter and all that sort of stuff.

13:01So I was working for an electronics company in the UK, massive department store one. And if you bought some stuff and then you needed to buy some additional stuff or things went wrong, warranties and all that sort of stuff, and you called in, you might have been lucky enough to speak to me. and my job was to you know sell you some accessories or figure out warranties or even deal with returns and stuff like that that was that was my promotion by the way i was then in charge of people returning stuff that was my big gig um no pun intended the point i'm trying to make is i didn't shop in that brand i shopped in one of their competitors because that was my preference and my question to everyone that leaves contact centers today even those outsources if you were to get everyone to go right stop what you're doing i'm gonna ask you a question who here has actually got our products or services or uses what we do how many hands do you think would go up and obviously nobody knows the real answer but it's not as many as people think they are because people are doing it as a job they want to do a good job but they don't have the relationship with the brand first we go the other way around which is we'll go to a company and go right we will source these experts from your customer lists they've already chosen your products on their own volition, on their own free will.

14:16We will now invite them to work as a freelancer to support your other customers. And only those that qualify to do it will be able to do it. And they fit it around their lives. We survey the thousands of experts we've got working on the platform. They are stay-at-home parents, they're retirees, they're students, they're commuters going in between jobs. They're working for two hours a day. So they're on an hour on the train going in, going home, and they're working on behalf of the brands that they've already got a relationship with. So it's not like you're asking them to do some outside job where they haven't got a real sort of feeling for it.

14:51And the feedback we get is sensational because it unlocks people's earning opportunity. It allows them to fit it around their schedules. They've already got a relationship with the brand and they want to help people. it's kind of like if you had a bpo and community forums and they had a love child that's what you get with gig because with forums and community forums you've got people that are really knowledgeable because they're the ones that are going on there they work on their own free will and they want to help people right they're working digitally yes they're working digitally they want to help people and they've got product knowledge but it's not a channel that you can scale or rely on because it has no slas has no kpis and it has no operation control whereas a bpo you might have you have all the slas and management insights reporting technology and that sort of stuff but you you may have a bunch of people that are just trained through the sheep dip system that are going to get their head turned for 50 cents a dollar more another organization and attrition is 50, 70, 80 % in contact centers.

15:55So crowdsourcing, gig, economy, what we do fits in the middle of all that. It takes all of the best bits around community forums, but professionalizes it. So it does become a dependable channel that you can then scale. One of the big secrets about generative AI, we hear about transformers and all the technology, scraping the entire internet, that's a whole different conversation. But what really created generative AI was the humans that trained it. Especially you go back and you look at Anthropic and how they got so good at coding. Most of the Vibe coding platforms are run. The underpinning is the foundation model from Anthropic.

16:31And it was trained by expert coders, people that write code for a living. Limitless has gotten into recently the AI training business. Help me understand what was the insight that said, hey, this is a good sort of product to invest in and what you're seeing from customers who were interested in exploring that with you so it came you know natural evolution i guess you know we could see the huge uptake in it and we're like well the difference between generous AI is that with contact centers you're dealing with customer service you don't have you don't need things like knowledge in mass or post-grad qualifications and that sort of that sort of stuff.

17:12What a generative I required was that level of expertise, a huge subject capability. So we evolved up thinking around what we do. And all we've done really is rather than the question coming inbound from a customer who needs customer support, it's not, it's a task. And that task is to collect data, to evaluate data, to label data, rewrite data, create data to train a language model of some description. So 90 % of what we needed to do was already there because we've got a platform that enables messaging, routing, payments, really, really, you know, something we've built since 2016, runs everything from a freelance perspective, all of the onboarding quality.

18:01And we've got thousands and thousands of experts. It's just that those thousands of experts are doing customer service work rather than doing AI training work. So we can see the explosion in Generative I.O. We can see the need for people and need for more specialist skills. And these are the people that aren't going to come to work in a contact center or aren't going to go and sit down somewhere because they might be teaching, they might be in jobs, you know, there's a different type of access. So we always have this concept which we have in the customer service world, but it's never been more true in the AI training world, which is rather than ask the people to come to work, is we take the work to the people.

18:40Through digital means, means that we can federate out the work to crowds of people that are onboarded, they're qualified, they're verified, they're skills tested, and then they can choose to do the work and they can use the knowledge that they have built. It accelerates this brand ambassador concept, which holds loads of water if you're a Sony PlayStation expert and you're helping another Sony PlayStation person. If I've got a PhD in astrophysics and I get asked to contribute more into that world or create stuff in that world, I'm probably going to be far more likely to want to get involved in it because I've spent the last however many decades sort of learning my profession in that space.

19:21So it opens up a whole new pool. And what we could see was that there is also a massive explosion in the types of companies in that space. So you've got the custodians of the language models, the ones we're all familiar with, the metas, open AIs, etc. Then you've got this big subset of companies that are selling into those language models in a way that, in my simple brain anyway, if we think the language models are these huge, vast libraries, these companies are going into certain shelves, certain coloured books, certain parts of the library and going in a very direct way to create or improve the integrity of data that exists.

20:03so that when someone types something, it goes to the right book on the shelf and presents it back. They're far more interested in the outcomes and the technology, the labeling software, the data evaluation, tagging and that sort of stuff. And they're less likely to want to be involved with the manpower, finding the people and dealing with them because it's probably not as sexy as what they're there for. So we can see an actual relationship between companies that need vast amounts of subject matter expertise that needs them in bursts of activity as well. So it's not a general hum of work and needs them across the world in different professions.

20:42So freelancing is the best way, whether Limitless are alive or not, is the best way of tapping into that type of resource. And we've just got a mechanism that happened to already 90 % of the way there. So it was a tweak in the top end to say, well, rather than this being someone tapping in a problem about, I can't get my headphones to work, it was a task around, we need this piece of work done or this piece evaluated. Otherwise, everything else is still the same. So that's where we are today. It's interesting because when you think of labeling, images would come in and people would, that's a cat or this is a dog.

21:15It's like Mechanical Turk, a couple of cents per task. And now when people talk about labeling and expertise, they talk about lawyers, doctors, scientists. To build these models with all of the promise in medicine or in legal scenarios, that has to be done by people who understand the subject matter. Where are you seeing the most initial interest in the offering? We're seeing it in a couple of areas, definitely with certain skill sets. So a lot in STEM, a lot in maths, a couple of other subjects. And we are also seeing still, which is this enormous need for people to do, still do that labeling work, still doing the annotation and label work.

22:01So that part still seems to be just as prolific at the moment. And whether that's going to change or not, time will tell. I guess what seems to be happening is that the more advanced the language model has become, so does the requirements around the skills that underpin the reinforcement learning. So I learned one of these many initialisms a long time ago, RLHF. And when I first heard it, I was like, what the hell is that? And it's reinforcement learning through human feedback. So basically humans in the loop helping train and stuff. So the world of RLHF is absolutely vast. But it's also the Wild West.

22:37So we regularly get requests to find people very, very quickly for huge spikes. So it's not like a contact center where it's like you get a 20 % volume increase and that stuff. We are talking like, you know, we need thousands of people to, in this space, this level of qualification or in this part of the country or that language skill set to do this piece of work for a short period of time. And that's the change in all this. So the requirements are there and they're only going to get more and they're probably almost definitely going to become more complex. So it's going to become harder to find the people.

23:15And like I said before, whether we're here or not, digital engagement is the way to bring people on to be able to do this. And freelancing is the best way for both to work. Because if you want to try and do this with a work from home BPO arrangement, whatever that might be, you still have shifts in schedules. you still have expectations of work that just grate against the very fabric of the demand that's out there for the more advanced stuff so if you have a general you know we always have this requirement that this type of work this type of skill set this type of language then it still makes sense to to use outsourcing and find the people who are there that you can rely on but that's not what's happening in the ai training world the demand is all over the place all over the time, all over the subjects, all over the levels of qualification.

24:10And by design, it requires these bursts of activity around trying to find people. Freelancing is just a fabulous way of finding that. And flexibility while people are, to your point, still trying to figure it out. Chris, thank you for taking time on a Friday out in the UK. I know that it's been a long week and you're just getting your kids down for the weekend. It's nighttime there. Thanks for your time. You guys are not just adding features to your software, you're rethinking systems. And that's one of the things I've really been interested in following the journey from 2016 from Limitless is not only pushing the gig economy model, but really thinking how technology and now AI integrates and how you can create value for companies and for customers from those systems.

24:52So thanks for your time. And as always, everyone, thank you for tuning in. Keep questioning, keep experimenting, but most of all, stay curious. Thanks, Chris. Thanks, Paul.

25:08Thank you.