AI Sales Agents That Actually Sell with Tom Cox, CEO of Humara

Russ Fordyce 🏆 (00:01.007) Welcome back to the winner circle. am here today with Tom Cox. is the CEO of Humira and something I'm super interested in because I have a lot of experience in this topic, but Tom has built an AI agent that is helping people cross-sell and upsell telecom products in a very complex industry of telecommunications. So Tom, welcome to the award show and congratulations. Tom Cox (00:23.778) Thank you. Nice to be here. Russ Fordyce 🏆 (00:25.797) Yeah, how about you give us a little bit of background for those who don't know Humira about what Humira does and how you got started. Tom Cox (00:32.92) So yeah, Humara is a highly specialist AI sales agent platform that can identify and resolve friction at every stage of the telco buy flow. do that, I mean, I'll go into more detail, but we do that by looking at how the humans do it, how the best human sales agent perform, what processes they follow, what sales techniques they adopt, what sales psychology capabilities they have. And we use that as a framework on which to build virtual agents that mirror that same skill set basically. So yeah, that's what we do. Russ Fordyce 🏆 (01:08.559) Very interesting. Now, telecom is, like I said, a super complex product categories. What made you decide on telecom to kind of tackle first? Tom Cox (01:18.134) Well, actually, okay, I started the company 17 years ago. We started life as a gift finding engine, something totally, totally different. Our name used to be 15 gifts. the whole principle was, having a conversation with, with a user and providing them really personalized recommendations. In that case, it was gifts. So that's where it started. And just after I'd launched it, we were approached by, a number of retailers, we got picked up by a tech blog that said, is this the beginning of conversational commerce or websites trying to take on a human sales element? Because you couldn't browse any products on the original 15 gifts. You could only have a conversation. And actually through that blog, we were approached by a few retailers. And one of those conversations we had was at the beginning was with Virgin Media, who said, actually, look, we've got this problem whereby it's hard. for customers to feel the same level of confidence online as they can when they speak to one of our agents. There's a load of complexity in our digital channel around broadband and TV and phone. And so we agreed to effectively pivot the product and they would be our first customer. And that's where it started. And we've just evolved over many years, proving ourselves in that vertical. And then at some point we decided that we wanted to stay there. At first it wasn't a... At first it wasn't by design, it was where it took us. And then at some point we realized that it was the perfect sweet spot because it has a lot of complexity, huge digital mix targets, very aggressive targets to try and keep people in digital, loads of complexity to the user around converged offerings and bundles and offers and lots and lots of complexity. And we decided early on, particularly with Humara and the generative AI, you can get so much better depth if you specialize. And so we decided to take a specialization route as opposed to trying to play everywhere. And so we decided to double down and that's what we do now. Russ Fordyce 🏆 (03:27.163) What is it about telecom products that are so hard? I I kind of know this, especially you get it on the consumer side. What makes it so difficult? Tom Cox (03:37.59) Well, it's a few things. First of all, they're getting more complex. So there's a huge converged push, right? So telcos are trying to get more and more money from every customer. That's the most obvious statement ever, right? But they're trying to be your provider of everything. They want your mobile, they want your broadband, they want your TV. They want you to sign up to multi-year deals. And so... They also want to provide you plans for your watch and for everything else. So they're just, it's all about converging and cross-selling and trying to sell more and more to the same user, which makes total sense. Obviously the problem is that when you're coming as a customer, that can feel quite complicated because you've got various different bundlings and converged offerings. So you might be coming in for one thing, but actually you're hit for two, three, four things. And that can be quite confusing to understand whether you should or shouldn't do that. And the other thing about Telco is We used to decide telco plans and minutes and based on the simple stuff, right? How much data did you get? How many minutes did you get? How fast is your internet? Now the truth is pretty much everything's unlimited, right? Most plans are unlimited data. Minutes are completely unlimited. And even with home internet, like now they're selling speeds that to be honest, go way above what anyone needs. You're talking about one gig, two gig. The average household doesn't need a gig of internet. Because the tech has got to a point where pretty much every plan kind of has what you need, they now have to rely on more complex ways of selling. Like, plan might be throttled at certain periods or there might be perks and benefits and this stuff. And so the simplicity of, need this many minutes and this many texts and this much data, right? That's kind of gone. And so for the user, it can be hard to work out what's the difference between unlimited Macs, unlimited pro, unlimited plus. Russ Fordyce 🏆 (05:34.129) Well, they all have unlimited in it and that's what confuses us. Tom Cox (05:36.908) They all have unlimited in yes, there's so there's other stuff. Yeah. Russ Fordyce 🏆 (05:40.433) I mean it is complex because you are are talking about different types of consumers, right? You're trick. We do not have bundles that that you know one size fits all we're all kind of different I know my my plan has got to be significantly Different than my mom's and my brother's and everybody else's just because of the way I use the services compared to them You know What where do you think the traditional kind of e-commerce and chat agents fall down where where you guys really pick up pick up the pieces? Tom Cox (06:10.166) I think the traditional and it's hard to use the word traditional, by the way, in this space, because it's so fast moving, obviously. But the majority of bots have at least started or seem to be mostly pointed towards service, customer experiences, right? So, you know, paying a bill, sorting out a problem, upgrading into another plan, you know, that type of area. which makes sense because it's logic driven, right? There's a fairly clear behavior decision tree that says, and that's brilliant for this type of technology to nail. Where no one is excelling, as far as we're concerned, is the true art of sales, like truly understanding what it means to sell in terms of how you build rapport, how you... how you deeply understand that customer, how you recommend products, how you handle objections, how you cross-sell, how you close, like all the things you learn as a human salesperson, right? The art of selling, and then you hone it over many, years. That's a much harder space for these agents to really perform and feel incredibly sophisticated. And so for us, think that the major difference and obviously there's lots of things technically that sit underneath this that allow us to do this difference. But if you boil it down, the difference for us is the laser focus on sales and then the laser focus of sales within a particular vertical in our case, telco. And that's where you get like real, real depth. Russ Fordyce 🏆 (07:59.993) Yeah, and it's interesting reading through your nomination. You guys kind of tackled this on a, it seemed like on a two prong approach, which is you kind of, as you mentioned before, you kind of had the sales expertise that you kind of built into the platform, but you also looked at the psychological traits and brought in the psychology of selling. us through a little bit of, you know, kind of those two elements in the one agent. Tom Cox (08:25.262) Yeah. So there's basically three, well now four now alive major training bits of training data, which allow us to create this sales, the sales intelligence. The first one was, well, we've had products live like guided sales products live for 15 years in global telcos. So we've got millions and millions of sales interactions. Our previous product was not. Gen I, Gen AI, right? It was machine learning. was deterministic flows with machine learning. but it was the same principle. was the same problem. And therefore we still had that, you know, millions and millions of these, these sales conversations. So that was a good start. We then realized that we, we had to, we had to see how the very best human salespeople do it. And so we hired a team of top percentile salespeople in the UK and the U S telco salespeople. And we have them simulate thousands of conversations with end users who who thought they were talking to AI. So their interface was an AI bot in the way you would expect. But on our end, we had one of our top performing agents. And so what that showed us is how the agents would successfully sell when the end user wasn't giving them any of the niceties that they would thinking they were a human. how, because that's how an agent, that's what an agent needs to do. It needs to humanize the conversation and sell even when the user is is interacting with it in a certain way. That gave us thousands and thousands of these transcripts. And then the psychology came from using the university that's local to our city. We partnered with the University of Sussex and their ability to help us to understand what the behaviors were that we were seeing in those transcripts. So whereas the agent would have just always said those words at that time. what they were able to do is to allow us to realize that, okay, that's anchoring or that's active listening or that's, you know what mean, labeling the behaviors that we're seeing. And it's through that labeling that we were then able to actually train the agent to be able to understand what they were and then to adopt them. So that's where the psychology comes in. It's the labeling and clarification of behaviors in a way that then can... Russ Fordyce 🏆 (10:31.857) Hmm. Tom Cox (10:49.23) can be used strategically, I guess, right within a conversation. Russ Fordyce 🏆 (10:56.027) So are most of these agents on your customers' websites? Is that where they're embedded or in the apps? Are they at the call center? Where are the customers experiencing? Tom Cox (11:08.718) It's all in digital, so it's not in call center. We're not doing any agent support. It's straight to the customer. They're currently in web, but the product is headless. So it could be anywhere basically. It could be web, could be app, could be iMessage. Anywhere a sales conversation is happening, the idea is that a brand can call Humara, recognize that this is a sales conversation. They can call Humara's specialist sales agent. And we can then handle that conversation wherever that conversation is happening. Russ Fordyce 🏆 (11:41.329) That's super interesting. Now, you mentioned again that you can have kind of multiple techniques, multiple sales techniques with one human rep or one AI rep, just like a human. Can you walk us through kind of the different types of techniques that you're talking about? Tom Cox (11:58.56) Yeah. So we, the whole product is built on this, what we call this like self sophistication framework, but it's basically a sales process. And in that you have various different stages of a sales conversation. So for example, you've got rapport building discovery where a human will also always make sure they deeply understand a user before they sell anything. Then there's the actual selling itself. Like how do you position a product? Do you position it with alternatives? And then there's, and then there's a sort of objection handling loop. So you try and work out like what's stopping this person buying. So they always, a good human will, a good human salesperson will always be trying to work out what's stopping this person buying today. And there'll be slightly nudging you to find that out. And if they can find out all your, all your objections, then they're trained to tick off each objection one by one. And the theory is that you can, if you get to the point where there's no reasons left not to buy, then the user will buy. Um, and so these are the types of techniques, right? So we built that into, um, into basically like an agent framework. So you have agents that are specialists that each of those parts of the sales process. Um, and within each of those, there are a set of techniques that, um, the, the first of all, we've learned from the humans, but now the agent is able to learn itself. So for example, in discovery, so discovery is a particular part of a sales conversation where you try and find out, you know, Russ Fordyce 🏆 (13:10.885) Mm. Tom Cox (13:27.66) what someone cares about so you can recommend the right product. We learned quite quickly that the best agents were following this methodology that's called SPIN. So situation, problem, impact, need. Basically, always start with someone's current situation and then always start with the problem they're trying to solve. Like what problem do they have today that they're trying to solve? Because if you can understand that, that's the biggest thing you can lean on to sell because you get the problem they're trying to solve and the impact it's having. And then only then do you talk about needs. So that's just an example of a technique which we can then sort of train into the model. So each of these stages, each of these agents that handles the stage of a sales conversation has a set of techniques. But now we're live, obviously we've got actual data. And the actual data, the thousands of users using Humara through reinforcement learning, these agents can obviously improve themselves. and they can improve themselves in a really focused area. So they can use that data really effectively. So yeah, that's ultimately where it will go now is Humara getting better itself at knowing how to do its particular job at that age. Russ Fordyce 🏆 (14:39.631) I imagine when you were doing beta testing or launching this in those initial trials, you started to see differences of maybe where you did things right and where you needed a little help. Was there something along that inside of the sales process that was a little harder than you expected? Was there one portion of it or one step or stage? Tom Cox (15:04.074) I think the biggest difference for us is that it was just so broad in terms of the level of intents and people that you see. So whereas before our product was effectively sort of deterministic decision tree, as soon as you open it up into a completely fluid, non-deterministic conversation, obviously you realize the rapid range of intents. So whereas, for example, that might make sense to say that discovery example I used, right? Russ Fordyce 🏆 (15:18.406) Mm-hmm. Tom Cox (15:33.922) that might make sense as a fundamental sales technique. But then you have to realize that two people will come on. One of them will be a really detailed orientated person, right? That likes to go into all this detail and give all this information. But the next person that comes along is someone that's in a rush. There's no time for that. Do you what mean? That just wants to get to the result, you know? And so, yeah. Russ Fordyce 🏆 (15:59.099) give two word answers and stuff like that. Yeah. Tom Cox (16:02.784) So, so then we realized that actually there are three layers here that without going too much too there were sort of three fundamental layers here for this product to work really well. One is yes, the core sales framework, the fundamentals, right? These, these are the sales stages and these are the techniques you have to follow. The next layer is intent. You have to be able to create experiences based on an individual's intent. and quite tightly control those experiences in terms of its quality. So for example, you might come in looking for a bundle, but you might care most about TV. I might come in looking for a bundle and all I care about is price or broadband. It's the same products. It's just that I've got a completely different set of things I'm worried about than you. And so you have to create intent driven experiences and you have to be able to learn and tune and improve. the experiences by intent, not just as an overall structure. And then the third layer is personas. That's like, let's say now we both have the exact same intent, right? We both wanna buy this product and we both care about basically the same stuff. But I'm someone that quite likes to get into the detail and understand everything and compare everything, whereas you're someone that just wants a quick snap result. And so you have to also detect persona. and be able to create experiences based on persona. So you end up with all these layers, but ideally, right, the perfect solution is that Humara understands your intent, understands your persona, and obviously understands the fundamental sales process and sales capabilities, and puts those things together to create this exceptional experience that matches your intent and your persona. And that's the layering where you start getting like really deep. sophistication. Russ Fordyce 🏆 (17:57.197) So I get that, that makes all the sense in the world, especially the persona part, because it really is. It's like, you know, I'm impatient and I want a price and I want to know it's going to work and I'm not going to get into the details. I'm assuming you well, you know. And so that's super interesting. How do you get to knowing, you know, an upsell versus a cross-sell and guiding those types of customers differently? Tom Cox (18:20.449) How do you mean sorry in terms of upsell and cross-sell? Russ Fordyce 🏆 (18:22.779) Versus like, you the different types of sales. Some of them are upsells, some of them you're trying to upgrade a package. The other one you're trying to add on a package or add on something. How are you getting at all those different, you know, different types of customer requests and kind of sorting them out? I imagine too, there's instances where, you know, somebody needs to be upsold and cross sold. Tom Cox (18:42.956) Yeah, I mean, well, to be honest, this comes back to the intents again. So you have to be really clear around what the intents are or what your main, what your major intents are of people on the site or on the app or whatever they're doing. And you also have to be flexible around people jumping intents, right? So you might start with one intent and then we might recognize that actually there's a possibility here. There's a clear possibility here to cross sell because they've said this and actually they're gonna be perfect for this type of cross cell. So therefore that would then become a slightly different intent, right? Which is the intent is, I'm thinking of super simple intents. They're broadband. I need faster speeds, I'm broadband focused, but actually something you said in there made us realize that you're highly likely to take a... Russ Fordyce 🏆 (19:26.459) I need faster speeds. Tom Cox (19:41.336) I'm trying to think of the right language without using any brand's language, a box which allows you to transmit your internet around five different bedrooms in your house, right? Because you said something in your discovery about having five bedrooms in your house and four kids. And so that would trigger like, okay, there's a slightly different intent now. therefore they get Russ Fordyce 🏆 (19:50.129) Gotcha. Russ Fordyce 🏆 (19:58.863) that's super interesting. Yeah. Cause you, with these products now you're getting more into the hardware space. I didn't even think about that because you know, all of these providers are now deploying, you know, mesh wifi and everything inside the home. So yeah, you're solving a lot, not just with service, but with actual hardware add-ons too. I imagine the content, the content too, and all these bundles now. mean, I don't know. I think I get all of my TV through my telecom provider and a lot of my DoorDash and everything else. Tom Cox (20:17.006) Yeah, there's loads of hardware out there. Tom Cox (20:28.194) Yeah. And actually I think that's the one thing we've seen. I think there's like 24 % more products that take on an average through Humara than through the site. But that product, doesn't necessarily, sometimes that's a service. In other words, like adding phone, but sometimes, know, sometimes it could be a physical hardware as well. Russ Fordyce 🏆 (20:48.325) Yeah, and that's going to be a service add-on as well on top of maybe a hardware. What surprised you most when customers started dealing with this AI instead of a human salesperson? Are they reacting in a different way? Is it different type of sale? Tom Cox (21:06.572) Well, it's super interesting, right? So there's a lot more people using it. So as soon as they realize it's AI and it's quite clear through the sort of, through where we position ourselves on the page and the use of sort of clear AI prompts, like the sparkles and the things that become synonymous, the open rate is huge compared to a sort of deterministic question journey type product. But the, guess the main thing that surprised us genuinely. And so I said before that we wanted our salespeople to interact with users who thought they were talking to AI because we thought they'd be quite rude and one word or two words and everything. And of course that does happen, but generally speaking, the, the way that the, the, the end users talk to Humara is not very different to how they do to a human. There's lots of niceties. There's lots of pleasantries. It's, it's, there's lots of thank you and please, and, know, telling them about some personal stuff. And, know, we, we have, we have lots of examples where they say, Hey, you know, will you get your commission if I come back? You know, it's like, and so it's been genuinely interesting, to, to see how close they are. It isn't that different, right? How, how, how customers are using it. I think that will change over time, to be honest. I think as people get more used to their own agents, right, and they're to be using it more and more, I think they'll probably become a bit more ruthless in terms of how they talk to it and how to get the most efficiency from it. So maybe that will change, but yeah, it hasn't been the radical change in behavior that we were expecting to see. Russ Fordyce 🏆 (22:50.937) Well, I would challenge you to encourage the good behavior. think some of the Amazon products actually do that, the voice response things that I won't mention because they'll start talking in my house. But it's interesting because if you're nice to those voice agents that we have in our house, they're nicer back to you. And so maybe we start building that into the frameworks. Tom Cox (23:01.985) Yeah, Tom Cox (23:12.408) Yeah, yeah, yeah. Tom Cox (23:16.738) Well, yeah, but they are nicer back to you. But in reality, okay, maybe with Amazon, it's different because it's in your house the whole time. So you're like, look, you almost have this weird connection with this thing and you don't want to, know, which in itself is weird anyway. But, but, but when you're going to a brand site, right, and you're buying a phone, a broadband, a TV or whatever, you don't need to build a relationship with that, with that agent, particularly, right? You're there for a, to do a, to get a service like Russ Fordyce 🏆 (23:43.985) task. Tom Cox (23:44.888) to solve a problem, to solve a point of friction. And so it's really interesting that they will take the time to write long sentences and long paragraphs that say exactly what they want to do with lots of pleasantries in there. When, of course, it doesn't matter. But yeah, it's hard to know how it's going to go. But it's been really useful fundamentally. It's a good thing. It gets you a huge amount of data and it allows the product to work incredibly well actually because it has a lot to then build from and build rapport from because the name of the game is rapport and confidence. And it's a lot easier to do that when the end user is giving you the opportunity to do it. Russ Fordyce 🏆 (24:31.993) And you guys have seen some amazing results. know, again, like I said, I've been in this exact space of upselling and cross-selling telecom customers, and our conversion rates were pretty low. They were still good, but they were pretty low. You're seeing conversion rates up to 23%. What's driving that lift? Tom Cox (24:53.506) Well, so for us, look at, we call it breadth and depth, right? So, breadth is the fact that we can identify and resolve friction at any stage of the buy flow. So whether that's trying to understand, do I want to buy from this vendor? Do they sell in my area? Mid funnel is kind of what product is right for me, real technical product questions. And then lower funnel is things like credit check and installation. So there are like a hundred points of friction that stops someone buying a product in telco and telco people have quite a high confidence threshold. In other words, it's quite a meaningful purchase, right? You don't want to get it wrong. You're stuck with it for a number of years. You're signing a contract. and so, and so it's not like an Amazon purchase where you can just buy it in a few seconds. Like you do not want to get it wrong. And so the combination of having this high confidence threshold, and lots of points of friction along the buy flow that can knock you under that line, means that Humara has a lot of opportunity to lift conversion. Because if we can unlock those points of friction, we see it directly go to the conversion and the uplift. So that's the breadth. And then the depth is just simply our ability, because we're so specialist in sales and so specialist in this market, we've been able to build a product. that is incredibly good at persuading people to buy, that can follow that sales process, that can build rapport, can discover and then handle customer objections. And so all these layers that we're building into the product basically mean that we can get better and better at the art of persuasion. So the combination of breadth, as in volume of customers and points of friction we can unlock, and then depth, how good we are at actually persuading that customer and unlocking that point of friction, all delivered through the umbrella of specialism. That's what's driving the performance that we're seeing. Russ Fordyce 🏆 (26:59.845) Are you finding any need for any human handoff? I mean, I would imagine that at the end of some of those conversations, some of those personas might be a little reluctant to sign on the dotted line with an agent. Tom Cox (27:14.966) Yeah, well, we, so we hand back to the car, you know, to purchase. don't, we don't do the, the purchase itself. but yeah, in short, yes, there's absolutely a need for human handoff. There's no, I don't think any AI, agent in the world can, you know, customer facing can't have a human handoff. So for us, it's basically, well, three things, they're asking for a product or a service. just can't, humar isn't, set up to support. The conversation goes in a direction which makes you Mara think that it can't complete this resolve this problem, or quite simply the customer wants to and just asks to speak to a human. So the human handoff is absolutely key. The biggest part of the human handoff is how do we provide the human with enough information from the conversation so that it then feels seamless for them to have that conversation, right? What's the right synopsis of the conversation so that an agent can quickly read that and then carry on the conversation. But I think it's about 8 % of our conversations at the moment go to human and around 92 resolve within within Humara without a handoff. Russ Fordyce 🏆 (28:29.251) Yeah, that's amazing. So you're talking about kind of this upsell channel and cross sell channel. When do you think it'll become kind of primary, you know, sales channel for some of these companies? Tom Cox (28:42.466) By primary, what do mean, sir? Russ Fordyce 🏆 (28:45.169) You know your first your first time in you know you're coming in and you're buying a phone you're buying a plan you know I think is your agent more targeted on upsells and cross sells I thought or are you actually doing kind of the initial buy Tom Cox (28:59.126) No, no, no, we're doing the initial buy flow. And actually that's our specialism. That prospect sales first time coming to brand, that's absolutely the sweet spot. We do a lot of cross-sell and upsell as part of that, because obviously you come looking for one thing and there's an opportunity to sell you three things. But no, no, the primary goal of Humara is to... Russ Fordyce 🏆 (29:01.157) Gotcha. Tom Cox (29:26.922) is to increase conversion for that primary sales flow, that prospect sales flow. Russ Fordyce 🏆 (29:34.129) Did you find any kind of trust issues as you started this and how is that changing as customers and consumers are getting a little bit more AI friendly? Tom Cox (29:44.408) Trust from the paying... Russ Fordyce 🏆 (29:46.063) just from the consumer into the platform. Tom Cox (29:49.088) are from the consumer. Weirdly, no, think the trust from the consumer point of view has been pretty amazing. In the sense they're very happy to share whatever you ask and to have that conversation. I think the biggest trust thing for us is trust. We only work with basically tier one enterprise telcos globally. And so the biggest thing is building the trust of those brands to allow a product onto their site to sell to their customers, right? That's a massive step of trust. Partly that's guardrails and accuracy and how do we make sure that, you know, that this product is going to be, you know, the absolute most accurate it could possibly be and have the highest possible guardrail metrics, which we've focused on hugely and achieved massive results there. But it's also brand. Is this going to represent our brand in the way that our brand wants to be represented? Is this going to talk like us? Is this going to sound like us? Is this going to represent our products in the right way? So that's the biggest thing that's been right. Building up the trust through the guardrail metrics, through the accuracy metrics, through our ability to ingest the brand's tone, the voice, and reflect that back through Himara. So that's the big about. and one that we focus on a lot. But no, the end consumer is pretty good. They're ready to roll. Russ Fordyce 🏆 (31:23.841) They're ready to roll. Russ Fordyce 🏆 (31:27.997) I'm wondering, did you learn anything about human behavior and and changing of human behavior when you're building this AI sales system? Tom Cox (31:36.45) Well, we've always been like, ever since I started the company, the whole thing has always been around the simple premise that the humans, i.e. the human sales reps already know how to nail this. Like anything we're trying to build now, they've done this forever. It's an industry that's always been very performance driven, right? Because it's always been a commission industry. And so there's always been this appetite and this urgency to get better at it. and to find new techniques. And so I've always just been fascinated by this simple idea that the problem's already solved. This isn't about solving a problem of how you do it, right? How you build rapport, how you build someone's confidence and trust and how you make sure that you provide an exceptional customer experience. It's already there. It's just... It's just how do you harness it? obviously for us, adopting Gen AI so completely is purely because the technology was the exact right technology for the problem we were trying to solve, which was to mirror how these best human salespeople behave. so we're always learning new things in terms of how how they behave and what they do, but that's the whole premise. Russ Fordyce 🏆 (33:06.405) Have you had any of your customers come back and say, hey, want you to take that, those learnings and all that expertise that you're finding from my customers and start working to train my non AI agents, my human sales people. Tom Cox (33:20.43) yes, yes, it's an interesting loop actually, because the first thing they want is for us to interview and meet their salespeople because they want to make sure that their salespeople are represented in, in their agent, like the way that we do it here. but then of course what happens is that Humara starts performing and starts to, you know, get better and better. within its own capabilities. And then of course, yeah, the natural request then is, how can we take the learnings from this and make sure that the human agents are now following the same process and the same technique? it's a bit of a loop. Russ Fordyce 🏆 (34:05.413) Yeah, I would imagine you could almost give a of a state of the consumer report, either back to the industry or back to your customers to kind of say, hey, here's what's new about your customer mix and here's what they're seeing. From a product perspective, it would be very helpful to have all that kind of first person feedback of, you know, what are they asking for and what do they like about our bundles? Tom Cox (34:28.482) Yeah, and also practically like, you know, we can now say to a brand on this page or at this point in your sales flow, these are the top five reasons people are not buying. And if you address these five things in the way that you position your products, the information you have on the page, the way you show your pricing, whatever it might be, you will dramatically increase your conversion because these are the five reasons people are dropping out at this point in the buy flow. And so if we can provide that for every point of the by-flow, so say for example, credit check, like these are the top five people reasons people drop out before they give you their credit check information. This is what they're worried about. And so again, if you can mitigate that in the way that you ask the question and the information you have on the page. So yeah, ultimately in the end, we should probably do ourself out of a need for us. The insights should do that. Russ Fordyce 🏆 (35:24.165) Yeah. Well, it's super interesting. I'm now going to go check all my provider sites because I am in the market for switching providers. And we'll start looking for the AI agents to help me build the right bundle because it always is very confusing when you get on there, especially when you're dealing with multiple accounts and multiple lines. Tom Cox (35:46.678) Yeah, yeah, that's a US thing, by the way, that mental thing of where you have, you have all your family on one line, people in England still can't understand what that's all about. But, yeah, that's, that's an extra layer of complexity that we don't, we don't have. Russ Fordyce 🏆 (36:02.149) Well, congratulations again to you on the AI Excellence Award. Obviously, this is a huge need in the industry. These things are getting more complex. And I think you're helping to reduce that complexity and make it easier for people to buy. So congratulations on the win. And we really appreciate the nomination. And we look to see what you guys do next. Tom Cox (36:21.07) Cool, thank you Russ, nice to meet you. Bye. Russ Fordyce 🏆 (36:22.533) Thank you, Tom.