What's Next for AI in the Contact Center - Art Coombs, CEO, KomBea - Project Voice X

This is the transcript for the session “What’s Next for AI in the Contact Center” presented by Art Coombs, CEO of KomBea, presented on day one of Project Voice X.

The transcript below has been modified by the Deepgram team for readability as a blog post, but the original Deepgram ASR-generated transcript was 94% accurate with only a 6% Word Error Rate (WER).  Features like diarization, custom vocabulary (keyword boosting), redaction, punctuation, profanity filtering and numeral formatting are all available through Deepgram’s API.  If you want to see if Deepgram is right for your use case, contact us.

[Art Coombs:] Really quick. My name is Art Coombs. I thought… I wasn’t gonna do this, but I changed my mind. I’d like to tell you just the the very first time I ever heard the term artificial intelligence.

In nineteen eighty two, I was hired to be an… computer operator at HP Labs. Now the address of my office was fifteen o one Page Mill Road. If you’re a Silicon Valley savant, you’ll know that’s the headquarters of Hewlett-Packard. Bill Hewlett and David Packard… Packard’s offices were directly above me. I know I’m showing my age here. They were directly above me. While they didn’t come into the office all that much, they did come into the office. I work in the basement. In the basement, so just to set the scene, in the basement, we had a couple of colleagues from Stanford that would constantly come over to our our offices late in the evening, Sandy Lerner and Len Bosack. I don’t know if you know those names at all, but I remember vividly about nineteen eighty three, nineteen eighty four, Sandy and Len were working on a black box about the size of a lunch box. Maybe a little bit bigger, a bread box. And I said, Sandy, Len, what is that? And they said, this is a router, and I said, a router, what does a router do? And they said a router is gonna allow one computer to talk to another computer. The next week, we all went to a place called the Velvet Turtle. If you know Silicon Valley, that’s a very… it’s no longer there, but it used to be a a very nice restaurant, and they told us they announced that they incorporated Cisco that night. So Len and Sandy were the founders of Cisco. In that environment, there was a bunch of people working in the in the basement, and I remember the big excitement, and we all went over to their office, and there was one of the engineers who entered in a sentence on his computer in English and it read it back in Spanish. And we were all blown away, absolutely blown away. That was the very first time I heard the term artificial intelligence. Now, for a cocky twenty two year old at the time, I thought, what a stupid name. Artificial intelligence, that makes no sense whatsoever. How can you have artificial intelligence? That’s an oxymoron. It’s the lamest thing I’ve ever heard. Now here we are, forty years later, and we’ve been involved in some form of artificial intelligence for many, many years.

It helps if I turn this on too. So I’m gonna take this. I come from the call center industry. I don’t come from the AI industry. I come from the call center industry. So I’ve been in call centers, I know it’s scary for some of you, for almost forty years. And so I’m gonna take a look at it from that perspective, but I wanna I wanna sort of set the stage. I think it’s impossible for for us to truly understand where we’re going until we sorta look back at where we’ve been and what we’ve done to this point and some of the major events that have taken place. Here we are in this very simple, simple forum. We have a human agent on this side, human customer, and I know we’re going to have AI customers. I know. But for right now, let’s just stay human customer, and we have AI agents. For many, many years, this type of communication has been the only communication we’ve had. Now what the huma… what do customers want? This has not changed in fifty years. Whether you’re standing at a grocery store, getting your oil changed, going through and getting your clothes dry-cleaned, or calling customer service, there’s three things customers want, and three things only, respect my time, be professional, make it easy and accurate. That’s all consumers want. Now we adopt AI and technology very, very quickly as humans. If we are convinced that those three stools of the customer service chair are strong and met.

Let me give you an example. When the first, I don’t know about you, but when the first grocery checkout self-serve stands happened, I don’t know about you, but I resisted them. I I kept going to the human checkout. Why? Because I had a bunch of vegetables and fruits and all sorts of stuff, and I knew that was gonna get all messed up, and I didn’t believe it was going to save me time, and it wasn’t gonna be easy. Now if I had a couple of boxes of cereal, a jug of milk, and I had barcodes… ok. Sure. I go over to self-service. But consumers are not gonna move over here until those three things are met. Now we start talking…

Now all of us in this room are trying to get some of these attributes over here. I don’t know if we can get ’em all over there. I know that there are some who believe that they can make a chatbot empathetic, but so far in our industry, in in our research… and we’ve done over, about twenty years, we’ve done over north of five hundred million calls, and we can go both ways. We can have the consumer believe they’re talking to a human, or we can have the consumer believe they’re talking to AI, because your brain switches. And I think that has already been addressed in this audience, but you literally think differently if you believe you’re talking to a human or if you believe you’re talking to an AI. But if you believe you’re talking to AI, having that AI be empathetic is condescending. So be very careful with how you’re designing your AI conversations, because you need to understand the dynamic of this human person over here and what they’re really after.

Now let’s put in this… and, again, I’m not gonna go through this whole thing. We’ve only got just a few minutes, but Rockwell will tell you that the first call center came around in nineteen seventy two. They’ll tell you that they created the first ACD. That’s not the case. Call centers were around in the late fifties and early sixties. Now during the seventies and eighties, we were in the loop. We may not have called it in the loop, but we were in the loop. Agents were primarily talking and controlling the entire interaction with consumers, but we were assisted by technology, AI in a sense. In the two thousands, we started to get on the loop in a big, big way. I’ll give you an example of how we’ve done that. And two thousand eighteen, nineteen, twenty, twenty one, absolutely, we are getting out of the loop. Does everybody understand what in the loop, on the loop, and out of loop means? Yes. No. Ok. Everyone, pan-faced. I’m assuming you all do. In the loop, agents are in control, but AI is helping with that interaction. On the loop, it’s a combination of human and AI. Often, the human is in the background. AI is taking the front, unless the human needs to get involved. Out of a loop, the agents out of it all come… completely. Now, in the seventies, we had ACD, CRM, IVRs, knowledge bases. I know all of this stuff is is… you guys know all this, but these were major moves in the call center industry that changed the call center forever and changed how we interacted with consumers. The reality, though, is back in the seventies, eighties, and nineties, all of this was first done to reduce cost. Second, it was focused on CSAT.

Let me give you a story here. I was about to start KomBea in nine… in two thousand one, two thousand two, and I wanted to get a feel as to what had happened for the last twenty years in our call center and how we had improved this product, the conversation. So if you think about a conversation with a consumer as a product, I wanted to get a gauge as to, had we moved the needle at all? Had we improved that product for you, the consumer? And I called my buddy, Jeff, who ran the American Express call center off two fifteen in Salt Lake City. He’s got about five hundred agents there, and I said, Jeff, can you tell me the average handle time of a lost American Express credit card in nineteen eighty five? And he said, yes. I can. And I said, what is that? And he said seven and a half minutes, about seven and a half minutes. Seven and a half minutes, it took a consumer to call to say, hey. Once I got through to an agent to say, hey. I lost my credit card and take care of that transaction. I said ok, Jeff, American Express has probably spent more money than just about any company on the planet on the best CRM, the best IVR, the knowledge bases. You guys are doing screen pops and skills-based routing, and you’re doing all sorts of stuff. Onshore, offshore, everywhere. In two thousand two, what is your average handle time for a lost credit card? There was a long pause, and he said, Art, it’s seven and a half minutes. And I was like, Jeff, isn’t that kinda depressing to you? Isn’t that kinda depressing? You’ve spent hundreds and hundreds of millions of dollars. We’ve spent twenty years with with, you know, Six Sigma analysts and all this stuff, and we haven’t moved the needle on saving customers’ time at all? And he said… he made a comment to me that I’ll never forget. And he said, Art, it’s more depressing than that. For the last decade, our CSAT has been dropping. He said, we have taken that seven and a half minute call in nineteen eighty five that cost just under thirty dollars, and in the year two thousand two, it costs just under six dollars. So while our shareholders are happy, our customers are unhappy.

So now let’s go… let’s keep going. So in two thousand, early two thousands, again, I know you guys all notice, smartphone, social media, consumer-friendly apps, they hit the market. They changed our lives forever. As soon as Steve Jobs held up that PDA, bang, the world changed in a big way. That created an explosion as to how we, as consumers, wanted to communicate with customers. Over the last decade, you have this explosion of cloud services, machine-learning tools, and powerful AI capabilities, which you guys are all involved with. These things are gonna change the customer service world even more radically than anything in the seventies, eighties, and nineties. The interesting thing here is based on what we see, when we talk to customers, is the first focus is on CSAT. Second, it’s on cost reduction. Now we just… I’m gonna… I wanna just play you one conversation of an agent communicating with a customer on the loop. Now this agent happens to be sitting in the Philippines, and it is a female. But I want you to listen to this interaction. If I turn up the sound… how do I turn up the sound? My sound’s all the way up. Bradley, SOS. Is there sound here I’m supposed to turn up?

[SPEAKER 2:] Let’s try it again.

[Art Coombs:] I’m not hearing it. You should hear that. Peter, it’s nice when you have your vice president of engineering in the audience. Am I supposed to… is this turn on and off?

[SPEAKER 3:] I think.

[Art Coombs:] Volume up?

[SPEAKER 3:] It should be. Let me try to plug it back in there.

[Art Coombs:] I hear I hear something happening. I don’t hear any volume. While he’s doing that, I’ll tell you a story as to why I started KomBea. I was working for a large BPO company in India. They had hired me not to own or manage a p and l, but what I learned was, I was literally supposed to be Yoda. I was the gray-haired guy, American, that was supposed to give our customers comfort with outsourcing to India. We had a customer, Blue Cross Blue Shield. I can say this. There’s no dawn disclosure or anything like that. But we had a customer, Blue Cross Blue Shield, and we were doing millions and millions of calls on their behalf. And we had a big meeting in London, and we… we’re all sitting around the table, and the vice president walked into the office… or into the conference room, and he said, I don’t want anyone to say a single word. Have you ever been to one of those meetings where the client walks in, he has a little attitude, and he says, essentially, everyone shut up?

[SPEAKER 4:] Same time work for you, though?

[Art Coombs:] I can hear it. I can hear it coming up.

[SPEAKER 5:] I’ll be working on Wednesday at this time. Can you call me at nine AM?

[SPEAKER 4:] — next… back at nine AM on Wednesday. Thanks again for calling, and we look forward —

[Art Coombs:] So let’s see if we can do this again. You want me to finish the story? I’ll finish the story. So he called… he walked into the calls… into the conference room, and he says, I don’t want anyone to say a single word. So everyone hushes up, and he said, we’ve recorded randomly twenty introductions of your agents introducing themselves and starting the conversation with our Blue Cross Blue Shield customers, and I’d like to play the first five seconds of every conversation, but before I do that… he put on the screen, the intro that he wanted every agent to use, and it was benign. It was like, thank you for calling Blue Cross Blue Shield. My name is x. May I start by getting your verification number or something like that. It was really benign. He played all twenty of those intros. They were all different.

Now if you were a QA person in a call center, you would think, that’s not a big deal. The person said, hi instead of hello. The person said Art Coombs, instead of Art. The persons forgot Blue Shield. They just said, hey. Thank you for calling Blue Cross, whatever, but it was benign in my opinion, but he made a really important point. He said, how in the world can we expect you to get our compliance statements right when you can’t get the intro right? And do you realize how big of a problem that is for our company when it comes to class action lawsuits? It was at that point that I said, we have to figure out a way to combine human intelligence and the consistency and accuracy of technology and sort of bring them together. So let’s see if we can do this again. And it’s gonna come up. I promise. Alright. One more time. I can see it, but you can’t. And I can’t hear it. Alright.

On that note, I encourage each of you… there’s three cost… three of my colleagues sitting right here. Take a moment and go listen to that conversation. And just remember. It’s a young Filipino lady handling multiple conversations at the same time on the loop, and you sit back and you decide if you were the customer, what do you think you’re experiencing? Again, we’ve taken over five hundred million calls. Most of them the… that human customer believes they’re talking to a human agent, and to our… the best of our ability, about one percent chime in and say, hey. Something is… something else is happening here. We can also go and make it sound like AI. So if we if we go back and look at the entire industry, and you look at the influence… come on now. Now it’s not… nothing’s working for me. If you look at the entire AI call center influence has been actually very, very little. Until about right here and depending on the analyst and what data you guys gather, if you’re talking strictly conversational AI, by the year two thousand twenty two, that’s in just a few months, about twenty percent of conversations that used to be between humans will now be done totally by AI. By the year two thousand thirty… and, again, we’re talking call centers.

By the year two thousand thirty, we estimate that’s gonna be north of eighty percent. Eighty percent of all conversations are gonna be handled by AI. Now what drove one of the major driving points, and we’ve heard this over and over again already, is COVID. Those companies that were in two thousand sixteen, seventeen, eighteen, and nineteen, that were already already working on AI, were in a better position to handle that kind of distraction. Humana is a great case study. If you’ve never… if you haven’t… if you don’t know what they accomplished, absolutely mesmerizing what they did. Their call volume went out the roof. It went up exponentially, as you can imagine with COVID, and yet, they had to send all their agents home instantly, creating chaos. Now I don’t know about you. I’m not trying to be a doomsdayer, but if you think this is the only pandemic we’re going to see in the next five to ten years, think again. I believe there’s going to be another pandemic. I believe there… and I’m I’m not trying to create panic, but I believe that our environment is going to create more and more situations that are going to make it difficult for us to have human beings talking to human beings all the time. We have to figure out a way to have interactions with our customers on a massive scale no matter where we are, no matter what’s happening on this globe. And, again, we’ve heard it over and over again.

One of my favorite sayings, technology has consumed our lives. We cannot get away from it. Just this morning… and I… this is true. While Jeff was talking about the garage door, I got a notification that my garage door was opening, and I looked, and my garage door was opening. And I could decide whether I close it or shut or open it. Technology has consumed our lives. We are right now in the process where AI is consuming technology. I’ll say it again. Technology has consumed our world, consumed our lives. Right now, AI is consuming technology. With that, I’m gonna open it up for questions real quick, and if there are no questions, then I’m gonna assume that it was completely understandable, and I don’t need to go into any AI. Thank you.

If you have any feedback about this post, or anything else around Deepgram, we'd love to hear from you. Please let us know in our GitHub discussions .