AI Agents for Customer Service with Latane Conant

Russ Fordyce 🏆 (00:01.181) Welcome back to the Winner's Circle, everybody. I am here today with Latney from Parlo. They are an AI Excellence Award winner as well as a Customer Service Excellence Award winner, and they are doing some amazing things with agents in the customer service space. Latney, welcome to the Winner's Circle, and congratulations on both awards. Latane Conant (00:18.274) Thank you, Russ. I'm so excited to be here. Russ Fordyce 🏆 (00:21.053) Now you guys are doing some amazing things, what's interesting to me is you actually came from Sixth Sense, which was another AI company in the B2B space. What did you see that you made the move over to Parloa? Latane Conant (00:32.622) Yeah, so I spent all this time at Sixth Sense, which was predictive analytics and AI, or basically trying to decide who's in market to buy what you want. And it was very advanced modeling. And what I realized was marketers spend all of this time, energy, and sophistication around getting buyers to talk to them. And I'm a three-time CMO, so not just at Sixth Sense, I think about websites, ads, content. I mean, so much money spent to get people to want to talk to us. Meanwhile, if you look at that customer journey, that infinite journey, there's this whole other side of the journey that has a pretty big leak in it, Because on the other side of the journey, the name of the game is not talking to customers. It's about deflecting calls. And so I just felt like with the new AI technology, there was gonna be this dramatic new opportunity to fix what I think is a leaky side of the customer journey and turn what is somewhat thought of as a cost center into a revenue center. so I've, you know, every career move I've made has been about customer experience. So I'm just working my way around that infinite loop. I really think Parloa with our customers, we have a huge opportunity. Russ Fordyce 🏆 (02:09.383) Well, it's funny you say that because the CMOs almost never take customer service into consideration. It's kind of an afterthought. The good ones do, like you and me, because that's where actually the revenue is generated, is in the customer experience side. You can bring a million people in the door, but if you don't service them well, there going to be a million people out the door. So I really applaud you for looking at both sides. Now, if somebody doesn't know what Parloa does, I didn't before I met you guys. Give us a little bit of an elevator pitch and kind of tell us what's there, you know, what's actually landing with Parlo. Latane Conant (02:44.31) Yeah, so I mean we were founded to make every customer interaction as easy as talking to a friend. And I love that because if you think about most, especially on the other side of the journey, experiences, they are not as easy as talking to a friend. It's wait times, it's frustration, it's handoffs, it's lack of personalization. And we're able to now completely reimagine that with AI. And our expertise and kind of what we're the best at is voice. And I think this is really significant and important because the voice modality creates more connection. It is when people have a time of need, they still wanna call. Study after study shows the importance of voice. And I think a lot of times people have gone to like chat and you know email and FAQs thinking that they are. self-servicing the customer, but it's really just kind of a flat, you know, humanless experience. It's not making a connection. And at Parloa, we can completely change that. I mean, we can literally have a conversation with every single of your customers all the time in any language. So to me, that is the future. It's being there for your customers in their time of need, again, 24 by 7, 365, in any language, know, personally. conversations as easy as talking to a friend. That's what you would expect of a friend. Russ Fordyce 🏆 (04:14.249) And I love the human to human or that kind of voice channel. I've spent my career in voice or a lot of it. What I see though is a lot of AI companies are coming out with products that basically bash your call or your text in into a big LLM and maybe some data behind it. You guys are doing something different. Walk us through kind of how it's different than just kind of having an LLM listen to a call. Latane Conant (04:43.032) Yeah, so I mean we literally become the front door. So if you think about it, the front door is broken. So 96 % of enterprises still run on an IVR, which is the press one for press two for, know, and that technology originated in the 80s. So, you know, we need to like retire the mall bangs, you know, retire the shoulder pads, you know, whatever, even though I think. shoulder pads, but retire some of the 80s stuff and really modernize that. And so what now happens when a customer moves from that over to Parloa is every interaction is conversational and natural language and you know any dialect, any language, and so it's just a completely different experience. And I think what you're queuing into is voice is the hardest modality because chat you can have some delay. Right? You can go, you you said like plug it into the LLM. It does its thing and it comes back and, and you know, there can be some level of latency. In voice, we can't do that. Like it has to feel immediate. And so we've really honed and tuned and done a ton with our technology to be able to have a very, very secure conversation. Because that's the other clincher. Secure and reliable conversation with extremely low latency. And so that's kind of the secret sauce that is is Proloa. Russ Fordyce 🏆 (06:20.945) Now, in my experience, you can call these companies and you do, get those IVR experiences and generally those haven't been looked at in decades either because you get run around the mill on those too. And they don't really have any context. I kind of have a billing question, but it's not really a billing question. It's more of a... Latane Conant (06:40.414) Press 2 for thread and billing. I'm like, I don't know if it's billing. What I know is... Russ Fordyce 🏆 (06:44.369) Yeah, and you go around and around. I know you guys have done some recent studies and surveys on kind of what's really being done out there. Tell us, kind of walk us through kind of what those enterprises are doing poorly and then walk us through kind of what an interaction that looks like is that's good. Latane Conant (07:03.916) Yeah, so rest, prepare to have your mind blown. So one of the first things I did when I got to Parlo, it's like if you want to fix something, you've got to understand the current state. And so we actually used AI to mystery shop the Fortune 2000. Because I'm like, I don't want to survey, I don't want to know people's perception of how they're doing. Like, let's just shop them all and see how it's going. And so in order to do that, we needed to create some scale. So we built these mystery shopping agents and we went out and evaluated everyone's current experience. And I was actually kind of shocked. I knew it wasn't good. I didn't expect it to be like as bad as it is. And maybe you'll disagree. Maybe this won't be shocking to you. But 43 % of companies don't have a visible customer support number on their website. So again, that tells me, I'm like, gosh, I've been, my life's work has been getting customers who want to talk to me. I mean, even market research I've spent tons of money on, you know? And we don't even want people to call at all. Like, such a huge opportunity. I think you mentioned chat before, and I mean, the chats are just not really serving a purpose at all. We found that only 8 % resolved the mystery shoppers, you know, issue or whatever. We weren't like blowing minds with our issues. It was usually like basic FAQ stuff that we were asking. So only 8 % of the time it resolved the issue and also only 8 % of the time it correctly handed off to a human when we wanted to get handed off to a human. So they're kind of worthless. So I don't know what people are paying for those, but it's not really helping you out or doing much. So that was kind of interesting. Then when we went into voice, Latane Conant (09:02.051) we found that 96 % were on an IVR, which again, I thought would be less, but fine. And then kind of the clincher was... We ended up doing a bunch of more manual deep dives and using AI to truly experience people's IVR trees, which were multi-layers deep and this and that, and wait times were up to 90 minutes, so crazy. But in doing that, we found that only 1 % of companies are ready for agent to agent. And this, to me, is going to be a crisis. And so hopefully this is an impending crisis that will get people the impetus to change because I do think more and more customers are going to expect that their agent can go and initiate the return or do whatever needs to get done. And companies really aren't prepared for that yet. Russ Fordyce 🏆 (09:58.685) It's crazy it's only 1%. I would have thought that, you know, especially the Fortune 500, 2000 would be, you know, running towards this agent to agent relationship. I remember seeing, you know, I liken it to, you know, I won't say her name, the A word and the Google home word, you know, talking to each other and, you know, making reservations for each other. And if you don't have an MCP server or an agent running to contact an agent, then Latane Conant (10:21.943) Yeah. Russ Fordyce 🏆 (10:27.975) then your system's gonna be blown up, I would imagine, by all sorts of people using agents to try to talk to you, because that wave is rapidly coming. Latane Conant (10:36.515) Yes. Yeah. So. Russ Fordyce 🏆 (10:38.471) Now, so you've talked about these agents kind of being the relationship builders and not just kind of the trans actors. What does that mean in the in the Parloa world? Latane Conant (10:49.933) Yeah, so again, back to kind of making it as easy as talking to a friend. You know, we work with one of the world's largest travel companies and you want to go and... and change your hotel. knows that you call anytime, anywhere, any language. It knows the trip that you're on. It knows the snowstorm hit. It knows you're with your family or not with your family. And it rebooks you this, that, or the other. And so that's just one example. Just making travel, which is exciting and fun, but can also be pretty darn stressful, as we know. We've all experienced that as easy as talking to a friend. Another is we work with one of the world's largest credit card companies and you know need to pay a bill and you want to do it immediately or you need to ask questions about it and that is really about a security use case right so that's the other kind of differentiating thing about Parloa is we tend to work with the more regulated industries and we have a history in Europe which is more you know the compliance and things like that are even that much more important. And so being able to, you know, process credit card payments and have, you know, conversations about finances in a secure way is pretty critical. And, you know, you wouldn't want your friend leaking that information. again, making it as easy as to talk to a friend. Another use case that I love because it really speaks to the time of need and the urgency is we work with an insurance company that provides roadside assistance. I'm on the side of the road, my car broke down, like I need instant assistance right now in any language. And I think that also talks to kind of an important dynamic of voice, which is authentication. Right? So in chat, if I type in my email and it's slippery snail 25 University of Virginia at whatever. Latane Conant (12:48.207) it's cut, you know, it's pretty easy to cut and paste it. If I'm on the side of the road and then I get distracted and this and that, it's more challenging. So making sure, you know, we're really highly tuned to that authentication key part of the process is important. But yeah, making it as reliable and easy to talk to as a friend from the roadside is pretty amazing. Russ Fordyce 🏆 (13:11.965) Well, and the model you guys have picked, this kind of agent model is really unique, I think. The thing to me with voice is it's always about context lost, right? I have to repeat myself and it's kind of like being at the doctor's office too. You get transferred around and you have to tell the same story over and over again or write your name 29 times on 13 different forms. You guys have really solved for that. Walk us through kind of that fix and what did it actually take to solve that? Latane Conant (13:34.617) Yeah. Thank Latane Conant (13:41.473) Yeah, so I think that it's the start. So it's knowing the context within that interaction, right, and being able to manage that context. So then again, you get handed off and you don't have to explain this, that or the other. So it's like retaining the transcript and all the data from, you know, one point A to point B. And then it's also being able to accurately call the whatever system of record. So if you see that I'm calling and I'm in your system, then you know, it's Latina. I know her phone number, right? I recognize her phone number. I even recognize the area code. where the phone number is from. I'm gonna start speaking this language. So there's things like that that just make the experience easier and smoother. The reason I said it's just the start is our aspiration is to really look at the entire customer journey and so being able to input, I would say, outside of that interaction information. So this person is a big influencer on Instagram or this person is you know, bought all this other stuff with us before, you know, so it's kind of like, how do we bring all of that together in a complete like context fabric? And then that will allow us also to do more of the proactive like outbound use cases that all of sudden it does really start to turn into that revenue center. So that's a pretty exciting innovation for us that's on the horizon. Russ Fordyce 🏆 (15:16.497) Now you say innovation and people kind of think, this is kind of a new thing. They're trying it out. Maybe it'll be deployed, but you guys are actually deployed in some pretty large companies doing millions of calls and transaction. Walk us through kind of the scale and how did you guys kind of handle that type of scale? It's not something you just flick on. Latane Conant (15:38.347) No, no. So what's interesting is we started in real-time translation at some of the world's largest companies. So that was like a good base. Again, it's like all voice and it was like helping agents, the human agents. So I feel like that was a good launch pad and you know, the business is not brand new. So that kind of gave us already great relationships and things like that. And then, you know, we really have worked with some of again, the world's largest travel, world's largest payment insurer, and they challenge us. I mean, they really do. And so it's interesting. One of the things that we do is we really co-develop with our customers. So the conversational platform that I talked about with that low latency and natural languages and barrage in and all that, that was co-developed with one of the world's largest global software providers. And so it's, you know, we just launched another feature that's really around making it very easy for an agent builder to be able to create new agents. So let's say a snowstorm hits the Northeast and all of a sudden every agent needs to be updated or, you know, the Northeast agents need to be updated or whatever. We essentially built almost like a fabric or a DNA that keeps all of the agents on the same page. So you can easily update one agent and it propagates to all of them. Cause agent drift is another problem. Like if you create this one and that one and whatever, then all of a sudden they're not in sync. And that was because of, you know, we work with a company that's like, Hey, if something happens, I need to be able to make an adjustment within minutes and your platform needs to be that easy to use. and really made for the customer experience organization, not an IT project. Because a lot of the AI initiatives are kind of like big IT science experiments. We're out there actually doing this. So those are just a couple examples. Russ Fordyce 🏆 (17:48.501) Well, and you're doing it quickly. It never ceases to amaze me at how fast we can do things now with AI. From the easiest, write me an email to review my database and pull in whatever information. It happens so quickly. Where we used to have to go, you'd probably have to write it up, write a request for IT to do this type of thing. They'd have to build a database. It was months of work to do stuff like that. And now you guys are turning these projects around in weeks. Latane Conant (18:11.364) Yeah. Russ Fordyce 🏆 (18:17.127) Walk us through kind of maybe a customer win that where you kind of didn't believe you could do it so fast. Latane Conant (18:23.983) One comes to mind. Yeah, I mean, so the key is actually in the sales, I think the sales cycle being very consultative. So one of the major things that we do in the sales cycle is we've actually built our own proprietary system that helps prioritize use cases. Right, so picking the right use case to start, And being able to really sequence that in the right way is absolutely critical. And there's sort of a secret sauce to not picking something that's so easy that it's kind of lame and doesn't show value and not picking something that's so hard that it's like, you know, way out there. And so, you know, one of things that we do is we just work with the customer on, okay, use case, use case sequencing. We have a great partner. We do some of our own deployments, but we also have a great partner ecosystem that have done this over and over and over again. And so, you know, getting, being able to stand a huge, huge customer up in 60 days, we didn't think we were gonna be able to do that, but actually they just went live and, you know, sent them, sent them the cupcakes. So that was a really kind of great win. You know, our average time to go live is about 90 days, which I think in an enterprise environment is absolutely, like I told you weeks, you know, Russ Fordyce 🏆 (19:48.902) It's insane. Latane Conant (19:50.541) That's weird, right? So I feel like it's a fast turnaround for that. But that's again, because we really lock onto the use case. We've also done a lot of work around like, know, pre-built integrations and pre-built skills. So we have this whole layer of skills that we built that customers can reuse. So an authentication skill, a routing skill, you know, those are things that every customer is going to need. And they're going to have to be a little different, but we've kind of built the the base of it for people. Russ Fordyce 🏆 (20:22.439) Now I imagine you don't throw these things out and just go like, turn it on. What sort of testing are you doing? Are you doing any simulations, anything like that? These are really kind of noisy environments, especially with voice involved. Latane Conant (20:35.671) Yes. Yes. So it was interesting because like You need to move fast and innovate, but reliability is the ultimate currency here. Like ultimate, it's how you get adoption, it's how you get customers to use it. And so we're always thinking about reliability and reliability in the lens of like a customer using the agent. And so... You know, there's things that we do in the agent build, know, best practices, actually taking a really, really long prompt and breaking it up into small tasks. So I like to joke that an AI agent is a lot like my husband. If I give him five things, maybe he does one. Maybe. And if he does two or three, they're probably wrong. forget about it. So, you one of things that we do is we take that long prop and almost develop like mini helper agents that make sure that, you know, the agent does exactly what it's supposed to do and does all five things. So there's a bunch of things in the build phase that help with reliability, but then, you know, just like you wouldn't take a human agent and before training them and practicing with them, throw them out, you know, to make call, like answer calls. You don't do that with an AI agent and The cool thing is you can simulate like thousands of conversations, right? I mean, we can create a angry Polish customer and a happy Italian customer and every edge case and everything in between. And that significantly increases reliability and getting agents ready for prime time. So that is kind of the second step in our agent lifecycle is the simulations part. Russ Fordyce 🏆 (22:32.445) Now you guys keep iterating on this product, improving it, improving it, improving it. How are you dealing with the flood of startups that have this kind of AI wrapper around them coming and probing at your customers going, hey, we can do that piece better, we can do that piece better. Latane Conant (22:48.119) It just, comes down to reliability, you know? And I'm like, hey, put us to the test. Like, let's test authentication. Let's test, you know, tool calling on their voice agents. Let's test latency in a production environment. I joke that like, voice is so sexy to demo, especially if you're doing like some of the lightning fast models, right? And I joke it's like Matthew McConaughey, like, all right, all right, all right. and he's joking and this and that and then you put it in production and it's like Jack Black. I know what the hell happened to that. So I think it's just making sure that customers understand you don't want Jack Black. And so the things you need to test are authentication. You need to test tool calling accuracy. You need to test latency, but latency in production because it's easy to have a demo that sounds like Matthew McConaughey. Russ Fordyce 🏆 (23:45.001) Well, he's one of the most, I think, sampled people in the world. I think I need to call that Fortune 2000 and see if anybody's putting in the McConaughey voice. I think he just patented his voice or something. He was one of the first to either treat. Yeah, I think he did. was either a patent or a trademark on his voice. And I think it was him and another actor. Now, we're entering this world and you guys are helping where. Latane Conant (23:45.551) you Yeah, yeah. Latane Conant (24:00.301) He should, because it's a very recognizable voice. Russ Fordyce 🏆 (24:13.341) we're getting to this kind of agent to agent interaction, which scares me a little bit. I can see it from the enterprise side of, is going to be coming at me. What is it going to look like for the consumer side of me being the consumer of one of these products and calling in? And what do you see that market looking like in five years as we kind of expand this kind of personal agent or digital twin kind of concept? Latane Conant (24:43.983) think it's highly dependent on what the service is. More transactional things, change my hotel, this, that, and the other. Already booking a reservation, changing a reservation for dinner, that's already happening. So I think those will be no brainers. you know, hey, I need to talk about my finances and saving for kids college. I don't know. You know, you know, we'll see. You might do one or two conversations, you know, with, with an AI agent or have your AI agent pull a bunch of data, but then, you know, want to talk to someone. So I think it's going to be a blend. I do think that it's sort of like to, and this has been happening since the internet is customers are more and more informed and opinionated. I think we need to watch out that they're not misinformed. And that's part of marketing's job, obviously, getting the word out. But they're just going to be more more informed. So I think that when it does come to a human conversation, the degree of skill level of the folks on the other end handling that is going to have to be high. Russ Fordyce 🏆 (26:04.551) Yeah, I've talked with enterprises about AI and they say the same thing, which is you can kind of get rid of that layer one, level one kind of customer service with chat bots and agents like you're talking about. But when a customer has a more complex problem, they need somebody to talk to. They almost want that reassuring voice to talk to that I am talking to a human. They do have access to the systems. They are seeing it and I'm not talking to a computer. And I imagine that will shift up to tier three soon and tier four and Latane Conant (26:23.983) Yeah. Russ Fordyce 🏆 (26:34.717) you know, who knows beyond that. When you're, go ahead. Latane Conant (26:37.111) Yeah, we like I never thought I'd order shoes online, right? It's like I have to try the shoes on. You know, now it's like, why would I go to the store and try on shoes? So I think it's just like, you know how it is, you get more and more comfortable with something, it works well, and you're like, all right, let's go. Russ Fordyce 🏆 (26:57.277) Well, and the companies are struggling with that. I remember walking into my brother's house and his wife was an online shopper, busy mom, and she had that same problem. don't know, it medium or small that's gonna fit? So she would order twice as much stuff and then return it all. And then those companies are, you hear horror stories that those companies basically dump that stuff because they don't know if it's been worn or not. And it's just like, that is a big waste. That's got us stopped. And I know a lot of the companies now are actually charging kind of return fees to kind of. Latane Conant (27:10.351) More! More! you Latane Conant (27:19.69) yeah. Russ Fordyce 🏆 (27:26.437) start mitigating some of that. When you're out talking to these, know, prospects, we'll call them, and they say, you hey, we've got to get AI in our customer service. What are most, you know, kind of, what are most executives thinking wrongly about deploying AI into their customer service operations? Latane Conant (27:27.119) Okay. Latane Conant (27:47.001) So I hear a lot of like, check the box AI, if that makes sense. you know, well, we put our knowledge base, you know, we built a custom GPT on our knowledge base, or we've got agent assist tools, and it's making it faster for them to be able to get the information that they need, and like, we're so AI, you know? And so that's kind of like, All right, we'll see how that goes, you know? So I think it's like, and I get feeling more and more comfortable and I think we should, you know, obviously give our human workforce all the help in the world. But I think you really do have to reimagine. The other thing that I hate hearing is like they keep the IVR front, then put an AI agent, then a human agent, and I'm like... I think the front door needs to, yeah, and I think about that as like an experience, and again, it's still that cost savings lens, like how do we contain, you know, they call it contain, how do we contain, contain, contain, and I just think that there will end up being winners and losers once people start to really. roll out these amazing experiences and they'll reap the rewards from Loyalty, CSAT, and ultimately lifetime customer value. Russ Fordyce 🏆 (29:12.547) Well, I think you're right on. remember hearing a story. I was in voice services for a long time, wholesale and retail. And we would always hear, and I don't know if it was ever true or not, but we would hear that if a customer called in and had a good experience with customer service, that their loyalty was actually longer than if they had not called. Latane Conant (29:32.771) Well, it's always an opportunity. You know, it's funny, my mom, you know, growing up, you learn everything from your mom. But she would, every time I thought I like messed something up, she's like, this is an opportunity. This is an opportunity. Like, go put it right. Go apologize. Go, you know, do the extra credit or whatever. And that's how you, you step up and you show that you're the kind of person that fixes things and step up. And that's a character. showcase and I think that companies can showcase their character and their values and how they step up. Russ Fordyce 🏆 (30:09.597) Yeah, I couldn't agree more. This is such an interesting space. I can't wait till we have all these kind of personal agents out there working on our behalf just autonomously. It's going to be amazing. Latney, thank you so much for the time on the Winner's Circle, and congratulations again on all the awards. Latane Conant (30:25.519) Thanks Russ, talk to you soon.