AI and the Evolution and Automation of Live Chat - with LivePerson CEO Robert LoCascio


This is daniel fidel. And you're listening to a and business podcasts. We're going to be speaking today about the future of chat. What is the role of man and machine in online conversation. Was that look like for the future of all businesses regardless of industry. We speak this week with someone who has a unique perspective on this topic robert. Low casio's the ceo of liveperson. Some of you have heard of liveperson. They're a publicly traded company. Robert founded them back in nineteen ninety five and they are now currently linked in have something like sixteen hundred employees around the world and speaks to about the evolution of the company from purely humans servicing other humans to where machines are starting to fit in. And how live person is building out different kinds of automations in different kinds of chat interfaces for different clients. The process is not magic. There is a lot of setup. there's a lot of nuance. And that's exactly my job to extract. What's possible what's working but robbery. Someone who gets to see the future in the cutting edge of customer service across industries in his perspective for that reason is valuable obviously robertson tough to get a hold of person as well as ceo of a publicly traded company. But we're lucky to have him here on the show. I do want to say if you like the show. You want support this program. It would mean the world if you could leave us. A review on apple podcast. You can go to apple podcast droplets five-star of you if you love what you're hearing here and let us know what you like to learn. Most what episodes really struck you. We try to fight really hard to get access to hard to access people to talk about cutting edge. Transient applications that you wouldn't be able to find elsewhere and it really is your feedback that helps to drive us every time we get a new podcast review. That's something that goes in front of my team for our monday meetings when we discuss her what we want to be building out for editorial. See your feedback really does matter and it helps to support the show. You can find a and business on apple podcasts. And simply drops review there. It would mean the world but otherwise without further ado. We're gonna fly into this very exciting episode with robert la casio. Ceo of live person. Here on the a and business podcast so rob. I want to dive in to the future of a and chat in the enterprise and the little bit of your journey there. But before i even do even ryan this company for quite some time when you guys started online chat in the ninety s companies public firm for twenty years. Here when you look to your business something. Like i don't know how many seventy or eighty million conversations you're handling in any given month when you look down at your business tried to find where. Ai could layer value for your clients. How did you think through that process. Because it seems like that's almost an endless There's almost an endless number of answers. They're basically looking at the data center. We do about seventy eighty million conversation month. I kinda you know that data we could take and look at how are use cases around banking and telco and and and retail and healthcare. These are big customer bases. Come from these verticals. How can we automate those conversations and so we basically just dove into the data we aggregated. The data looked at across customer. And we could see that there were patterns that made really good conversations. So that drove is to say okay. Let's build a host of tools and a new platform called conversational cloud in which we could enable large enterprises market customers to scale their conversation. Automate them and with the onset of covid the there was a massive massive impact contact centers the contact center agents help and nope he was there. They couldn't take calls so so they went home now. They're taking calls there and so there's been a massive drive to automation now because having someone work at home and answer calls just not the way to go. How do you automate that through a digital experience. That's really what's happened and obviously with the onset of covid firms like yourself or now in this position to potentially catch the opportunity to major way when you look at serve. What's potentially automated man even there so much to get into. I mean there's for each individual business that you work with you know maybe a work in with a one. Eight hundred flowers may be. You're working with us. The citibank guys work with some of the biggest companies in america here. The use cases are relatively bespoke And i know you. You worked on building a tool. That's at least understandable for nontechnical people. We definitely need subject matter. Experts to layer context and to structure or flow flows for these conversational systems. But you determine what the bounding box would be because even that feels feels endless and you need to kind of focus in somewhere yes we. We're all this under the header of commerce. And i fundamentally believe that conversation. Commerce is going to be the next leg of digital so we had ecommerce and now we're shifting see commerce and e commerce is is a very interesting thing when you think about like show up at a website and basically every website looks the same right. I mean top nab. Yeah data and google set those rules instead of you don't do that. We won't index you and that's why everything looks the same so what you're seeing is digital each brand doesn't have its own personality now with conversational commerce. You develop your own personality and you develop your own way to engage your consumers too like chapultepec. We had obviously during covid. What happened was people didn't want to go in and get a burrito and make They don't want to sit there for five minutes. Make freedom so chapultepec turn to us said we want to build an automation that somebody could come into their mobile device to any messenger front facebook. I message app and we want to configure a burrito and or whatever they want and they built. We built this Automation called pepper. It's called pepper and you can communicate with pepper. And then you show up at the door and the handed to you. And so that's the poultry. One we've got David's bridal which also people didn't want to come and people get married but do you want to go and sit in the store and try and addressing the one of the largest bridal companies in the world. We automated looking at different things for your body type for your style making an appointment. So you're the only one that's in that store. Don't have a crowd in the store. We built all that. This is just examples during kovin but every one of our customers get the ability to create a conversational experience. That's unique to them. And that's what makes commerce Yeah i i could certainly see the argument that linear text back and forth actually has potentially less less opportunity for really robust customization than you know. Really fancy dancy website. But but i get where you're where you're coming from. I think that there is like a. There's a certain flavor of the brand that you get from talking that maybe you wouldn't get from from a website and obviously what you're saying is that we can tailor that of course with a website. There's a lot we can do with color and features and video and whatever but certainly there's personalization on the conversation side. I could kind of see arguments on both sides there with respect to working with these big brands again. You're you're dropping some pretty pretty big names here. Names of almost everybody listening in has heard of what does it look like to set up these unique systems for them because obviously chipotle's use case which by the way feels very accessible. The number of burritos you can build is not unlimited right as opposed you guys work with delta airlines. The number of things i can complain about to my airline is. There's probably two dozen. You could tackle off the cuff. But but there's gotta be another four hundred that are just we're handing this to a human being but cipolla feels. Wow that feels like almost like dominoes. Chat bot was pretty popular. Winning hundred flowers was doing some stuff. Because there's only so many purchase options but when you go into different clients you got big airlines. You got these restaurant chains. What's that process of really working with them. Because there's got to be a bespoke build out part. You got your core platform you guys are able to use. You have a tremendous amount of data and a lot of staff obviously which it looked like to hop into these big brands and build out. Something that really drives value for our platform conversational cloud. We really broken into three areas in and around the three areas of need when you're looking at scale automation. One is the intent. We called intent manager in its technology that ingests all the conversations and then organizes the intense. So we talk about intense anti-business. What is it consumers. They haven't intention to do something with you. They want to buy something. They wanna customer care question answered and that's an intention and we call that technology intent so we organized the intense and it turns out as we all know. We all asking and attentively. So i may ask a bill differently than you do. But the technologists differences aggregate them. And then what you get to see a list of all the intense in your business. And what are the contents and where the mid level. I'm usually what we say. Let's go solve the top intense bam one and is thirty percent of all the voice calls. You're having let's to solve that. The as the next thing do do have access to those systems that can support that conversation fulfilling that intent and then we have a thing called conversation. Builder that enables the automations the the actual conversation the to be built deployed and then managed and then we have the analytics behind it. But but that's kind of the thing so it's look at the intense. Crafty automation deployed improve. It manage it and then look at the data around that to keep the cycle going to be good gets ninety ninety five percent accuracy. That's where we try to get ninety nine hundred ninety two over ninety percent to goal that we're trying to give. Yeah that's one of the things about these customer facing systems. Righteousness you if you guys were just building. Some insider analysis tool based on call center transcripts. You could be right. Seventy five. Percent of the time in your. You're potentially still delivering a good amount of value to those internal teams but if you're facing customers we've gotta be able to be right more often so you're talking about kind of categorizing intense and i think the companies that do this. Well have to go through some phase of that you know based on volume what are people at at their core. What are they trying to get done. Let's let's take a look at our percentages here. It sounds like a. I can help with that. Obviously and then figure out what we can tackle. The there must also be a consideration of which of these questions that that occupy huge amounts of intense space are a tackle bowl into what level. Maybe maybe some of them. Well we could. We could answer the first volley of questions but almost inevitably the second one. We're doing smart routing to the right human being maybe in other cases we can potentially automate the whole thing so it sounds like you're probably considering what's taking a big pie slice of that intense face but then also you know how handleable is this thing picking pie slice you can actually get to that ninety percent with yes we we call this the tango which is a tango between machine human or agent and human and so when we go live a lot of times. Also we'll have agent assisted bots. That are as an agent's doing a human interaction of a message or chat that they are being assisted by abbad and usually start there so that there isn't a lot of error as in the bought will recommend something to the agent agent so that's correct and then they're training the bottom of what the right answers are so your listeners. If you're out there. And you're looking at szott. I'll tell you where we see the problems. Cool okay so if you're out there looking at physical transcripts like you literally are pulling a hundred chat transcripts and you're reading through them and you're trying to stab him with the intent is they. Try to understand what a good conversation is. this is problematic. So that's problem one because it's really hard for a human to ingest information at that level. You can go through ten twenty transcripts but you can't go through two hundred thousand and he usually need at least you know when we're looking at building automation using need at least two thousand transcripts to understand the commonalities of what is a good conversation around specific intent. It's very hard for human a look two thousand translating. Here's the pieces. I've made that good. The second thing is if you are if you're technical teams your data science teams are building the automations 'cause they're looking at these transcripts than they're creating something that's problematic because usually this group of employees don't talk to customers like context so so that's why we built the tools so contact center could rate the conversation in a natural like the writing a book like the writing a check they don't have to flow charts and they just they write a conversation naturally and then we have a there that tells them how they're how they should write the conversation they can pull an api but if your hand doing this this is where we have a lot of positive. Yeah i mean. I think teams that are. Let's say not wholly and unhealthily ignorant as to how a i should rate and i would hope that. A good percentage of our audience fits that bill teams that aren't at that level of wholehearted ignorance. I think would probably do a good amount of actual labeling of the intense and then they would take system like yours and they would say how well existing labeling a dense. You know how. How well does this lineup to. What we know are tend to be. You're right millions. Having a human team label million no no of course i would be obscene but having serve a head customer service person in maybe some of their crack customer service subject matter experts. Kind of come up with these major buckets and trees and then seeing hey can detect in bucket. These things appropriately. It would feel like. That's the only realistic way to to skin the cat. The part about having technical folks build out the system is also pretty tough so having having the guy straight out of carnegie mellon with a phd. Hey pal you work at delta airlines. Now why don't you go ahead and build our chat. But that would be one of those wholehearted ignorance things that we would sort of advise our listeners. Not to do the other thing. Though that's often really hard is take your subject matter expert and put them behind the steering wheel of something. That's that's all of a sudden going to influence the user experience of you know a hundred thousand folks a day you know delta airlines so when you work hand in hand with a big brand and you're working with their subject matter experts which is really necessary of course you need to talk to. It to get access to data you need to talk to subject matter experts. They know how these flows work. Howdy a team up with them to get the company to a level of confidence with they're ready to deploy imagine we got a sandbox and incubate the stuff a little bit because again it's a lot of responsibility for jimmy customer service even if he's been there for twenty years to push something live talk about that tango with you and the subject matter experts to get something rolling. You're absolutely right. It's crossing groups so you've got the it folks and the technical folks that have the knowledgeable of back end systems you need. You need that access you have done. The context rabs who understand about conversations and the diversity of stations. By the way if you talked to a context fine even for the same intent. They sometimes pick up the personality like we had a. We have a body called grill master which is selling outdoor grills for lows. Wow what an interesting use case and end flow end to end like no human in the loop. That's pretty a contact center. Agent group in the dominican republic built despite many others. And i remember when i was i went down. There is like one of our first about two years ago. I or may will scale and i talked to the contras at. Tell me what it's like selling a grill and they told me like well. There's three different types of people who talked of this the the know it all and they they ask an intent a certain way and as a person who asked he will ask me. I like do you barbecue steaks. Do you like you know. They'll they'll ask the agent and they want they want to hear the agents. Point of view is al. You can on this same antenna. I wanna buy Three or four personality types that if the baht started to work and take you down flow of a certain personality. Would you would lose the person you would lose the consumer's engagement so we asked some questions. And that's the beauty of conversational commerce. You don't have to guess the automation can ask a couple clarifying questions. Which then the flow in a certain way and then the agents watches so the agents deployed grow master. They watched it work. They see it's breaking like as in consumer broke something they hop in real time taking over so there's no transfer then they improve it and we within a couple of weeks you confidence rate of automation and again. I mentioned that that objective is easier for some use cases than others the grill. One is pretty interesting one because that actually does feel man that feels really bespoke really kinda like honky dory conversation you know what will like. What's the robot gonna say. Hey how would you like to grill on the weekends you know. Do you also like xyz kind of stake like oh man. What a weird bucket of questions. We gotta draw from there because at that point we gotta pretend to be human unless we don't But i presume. I i don't really know what the what the permission or whatever level thing is there. Some people don't really care for to bader a human maybe some people do but but that that actually feels somewhat challenging. I imagine in these cases robert. You've got to have a core team champions that you're working with because you can't just observed from the outside. Hey how well are these guys providing feedback to this thing. We gotta make sure. They know how to provide feedback to the system. Where to annotate it. Thanks to edit in the workflow. Make sure they're doing it right because a lot of time. They won't because nobody does stuff right the first time. And that's okay. I'm i'm sure you build software. That's good as it can be there but people aren't always going to do things right. What's it look like to to monitor and work with that team. What kind of a core set of champions do you need to have on on on your side really to consistently do that. At ration- to make this work we have a group of conversational designers that go out and teach our clients how to do these conversations and make them work and how to use the tools correctly. But like you said someone wants it to me like one of our customers that the way they look at automations is that they're like employees and the contact center agents. You just don't put him on front of a customer and give him some basic training because also the business changes so maybe the thing you trained him on week later there's a new product or a new offering and so the new marketing message that went out and consumers coming in. So there's a karen feeding that has to go on consistently with the automation. Some can be set some like we have on another airline qantas the first first automation. We did with qantas which the large australian carrier was. Check my back. They had a lot of people call. 'cause they lost their bags and they want check. It will the system to that. The first one we built was check. Your check your bag and that doesn't change much bag great. Give me your flight. Number union ended up. Okay great. Here's where your bags and it'll be it'll be dropped off to you so that something that was kind of one and done and set but you're right the automations need care and feeding over. Time are still gives you feedback one to do that. Had when you see a high failure rate when he starts high failure rates on the automations yeah in. Somebody's gotta be monitoring that i presumed to most of the firms you work with have some level of in house data science expertise where they can talk to you in a little bit more technical vendor language or the bulk of the time. Are you just talking to some. It folks and and really you're not you're not touching our in house data science fox it's mostly. It there may be some days easily now. There's data signs in the it group so they are our it groups essentially the it groups. It's kind of a fifty fifty split between we work with. It groups or we work with the customer care sales groups and so someone will be the champion on either side and they have to bring it through. It's hard whatever champion comes. Let's say it's hot customer care. They've got a really bring together the other groups of rural. Because i've seen this where we get a champion in cares. I'm all in automations. We're going to get the ended an. It's too busy whatever reason and you can't get too sophisticated automation. So you're staying up on that. Faq bought which we all kinda don't like. Yeah yeah problematic in. You're going to be a champion. If you're on the it side same thing just throwing it out there in front of a customer care group. They just will be getting involved in. This is not a good conversation. We're getting phone calls. And they're costly and it's a bad experience and see sat. Went down ramp down this. But here's here's the results. I been this business for so long. I've dealt with large customers in all sorts of customers. We can get to an eighty percent automation rate. We could do that today. We can automate eighty percent of your customer contacts now in order to do that we have to have access the back end systems. And we've got to have an alignment in the company looking at the data across telcos banks insurance retail. There's very few things that we need a human to talk on the phone about very few not being offender myself. I can't speak to the degree of confidence. You've been in this business longer than me. You're also vendor all things said you. You definitely said peace there though and i doubt that there's a good healthy chunk of what we all believe. Humans need to to be in there for that. maybe they they. Don't the last question. I want to run by you. Hear rob as we wrap up today a now that we have an understanding of what you guys are built. I also really appreciate context on the process. You guys have so many big clients. I think the process gets a lot more credibility when you guys talk about. It is really what you've done with your own firm. You know a lot of people listening in it might be interested in applying chat in their own business. They're also maybe a little envious. You know that a company that's been public for twenty years has been able to build an an wing of their business that involves artificial intelligence in leverages it in a powerful way when you think about the transferable lessons for the other listeners. Who might work in. You know the kinds of companies you work with big banks life sciences firm. Maybe a big retailer. What are some of the things you learned as a leader of an innovative company for bringing ai into a company. That really didn't have it in its dna from the get-go don't you pick up that other people should know. Yes so it's funny i. I was reading an article. Wants can years ago in the financial times about the best lawyers. Corporate lawyers in the country and one of them was the general counsel of spotify and spotify is one of our customers and i called them up and they wrote this article about how he had a he had a team. It said skye has a tech team. I'm like what does it have at least interesting right. So i call him up and i met with them and he said yeah. I'm automating like a ton of what we're doing in the legal department and nobody wants to automate legal. There's no technology automating legal like a legal process so contracts and all this stuff thinking and i started my own company in our our. Gna line is about twelve thirteen percent revenue so every year by told the thirteen percent. We're spending on finance. Hr and legal and this is kind of normal to twelve percents kind of what a normal public company will spend so. I'm like what we do there. So i looked into it and stuff and i started realizing it's just these big data repositories and then people rely on on analysts in these groups to look at the data and give them feedback. And so i said. Why can't we automate that right. I did did something which was pretty bold. You know and. I think it'll be come. The trend about is which dot my cfo. And i hired a data scientist ahead. Guy a guy who came at mit he. He built a hedge fund based on some machine learning algorithms that he built around consumer engagement on mobile devices and i found him. I hired him. His name's john collins. John join does and he is automating. A lot of you know what we normally have humans doing and such from mundane work like you know we we have. It's mid hundred million dollars. That comes in from every year in sales so people send money to our bank account and there there was a human who looks at like two hundred thousand and sixty four to twenty six cents came in. Oh that's related to something this customer's invoice loyd is humans to do this any any and so built in automation to remove that human. And there's a lot of other things. Now we're working on a crm system that's fully conversational so our sales people don't have the input stuff into salesforce but we capture the conversations. They're having on voice calls and emails also and we can tell where they are in the stage of selling and so he's built. He's building system like that. So all i'm saying is that we're now pointing ai. And all that towards us. And i decided to take our. Cfo make use it data scientists. yep and it's changing things so i think we could take down. Gna from you know. Twelve percent. I think we'll get it down someday to about three three percent so for companies like ours. That's taking fifty nine dollars spence and taking it down to five million so like that and that's all and so even though the navarre companies allied I kinda know that life people should be doing strategic work. They shouldn't be doing mundane work. Yeah this is the funny thing right. You're talking about eighty percent I can't back that number but You talking about eighty percent conversations not needing people that is the name of the firm off but it sounds like you're live people you know your vision here for live. People is hey strategy high level work. that's basically it. We shouldn't be doing monday stuff. Yeah and and there's there's a little bit of a debate on a abbott the concept of ai which is alon mosques. Keep saying it's an existential threat to humanity. My perspective is that i will give us more capacity as humans like any machines do and yes job loss will happen. I don't think it'd be as severe as people think because like our contact center agents who are answering phone calls now are becoming bob. Builders we've made that transition with them says a lot of need for even people with that skill set to move to conversational. So i think that it'll give us a lot of capacity and that's why you know if you look at china are big. Everyone talks about the biggest of china Country and our government is focused. Although our our large tech companies are most are the reason that countries china are taking trying to take a lead is. They think we're going to open up more capacity and if we open up more capacity for the people that are in our country that do better things like we're going to give him time back and usually when people get back they fill it with. Sometimes you know new innovations and we've got to take a leadership role and not keep talking about it as like we're going to replace human brains and we're going to be creating weapons with ai. That's gonna kill us. They're gonna turn on us like by Terminator yeah beyond the science fiction. This this is here to help us do our work. Better like my. Cfo doesn't work better because he's got machines are making better decisions because of our machines and everyone should be thinking like that in. That's gonna make us a better better place. Once we more capacity or freed up to do better things with our lives yes strategic work. I'm not entirely sure if so. I'm definitely with you on the idea that humans can move to higher level strategic work. I think sometimes we're not gonna be able to find a role for everybody but you know some. Some of the time definitely are will be able to repurpose staff. I think making such a statement would not make me feel like a misleading. Anybody i think that there are some folks that are thinking about the existential side of ai. Twenty thirty years out. And i think some of that may very well be still be super relevant. And i'm not gonna take away from actually some of those broader concerns that said today for a cfo. If you're reading nick bostrom on the job you basically should be fired And in you should be reading something that has to do with you. You actually delivering results to your to your company. So i won't. I won't disparage bostrom or folks in that camp. But i also will totally agree with you that. Hey there's a lot of good to be done and let's go ahead and get down to it in terms of transferable lessons as we close out here rob things other people could learn from is one of them. Hey take leadership and bring in folks that can breathe life into a i in key parts of your businesses. That one of the and do you have any other takeaways as we wrap up this episode. Yeah i mean. One of the things. I decided to do. Last year was everyone is in the company is going to be an ai. Native and so everybody from a sales person to person in finance top seed. Technologists may have that skill. But everyone is. What will our kong native so we put people through a program. Now a series of programs ends a continuing education series. Everyday i can listen on on something around back. What i felt was if a really embraced his as a company everyone in every corner of the company has to understand at some fundamental level. You may not be able to program. I get that you may not be able to create an automation yourself. But you gotta understand. Would a feature says you have to understand what you know machine learning as you have to understand some basic stuff around this in not just what. You're reading in newspapers. And so i would say even though we are learning i company. We have data. Scientists are headed day size. Help with you don't have to go that way like we're even here. If you want to reach out to me. I cannot tell you where it can go for resources. We created a curriculum which we would even be more than willing to share with people. And i think if you can get more people in your company someone in finance going to try to use a to change their job versus their job gets changed because they i and that's where i found this to our employees. People are not technical. People would tell me. I need to learn about this like i know. This is going to change my job. Yeah i wanted to help feed on it. I don't wanna be run over by it. And that's the greatest gift you can give to people in your organization got okay so two things here on the one hand there's some level of hey smart leadership in the parts your business that have opportunity who who can share that. Ai vision that's going to help drive some change also from the bottom up in a week and decide to breathe ai fluency into the training of all the departments of the business so that people get it and can participate in this transformation. I like it. I like it. And you're you're you're doing it while you're moving and shaking with a rather large firm now rob so hopefully. Some of the folks tuned in are taking notes. I let's all we have for time. But thank you so much for being able to join us on the show today. thank you. So that's all for this episode of the i and business podcast big. Thank you to robert for taking time away from being the ceo of public firm to join us here on the show. And thank you to you for listening all the way through to the end if you enjoy this program and you wanna make sure you don't miss any future episodes or any of the trends are use cases we cover. It emerge the be sure to stay subscribe by email. Probably already subscribe on podcast but you can stay subscribed on our newsletter as well. That's emc rj dot com up at the top right is abundant called subscribe. Every tuesday and thursday we send out a newsletter with all of our latest interviews use cases in our here to merge if you wanna stay ahead of the curve in terms of use cases cutting edge trans and best practices for the roi. This is the place to stay plugged in you can go to. Em yard j. dot com click subscribe and stay connected to us there otherwise. Thanks for listening all the way through. And i look forward to catching you for thursday's episode here on the a and business podcast.

Coming up next