375. AI That Understands People: Blending Tech and Behavioral Science

Welcome to episode 375 of the Brainy Business Understanding the Psychology of why People Buy. In today's episode, I'm excited to introduce you to Eve Boudreau, head of Technical sales at Google Cloud. Ready? Let's get started.

You are listening to the Brainy Business podcast where we dig into the psychology of why people buy and help you incorporate behavioral economics into your business, making it more brain friendly. Now, here's your host, Melina Palmer.

Hello. Hello everyone. My name is Melina Palmer and I want to welcome you to the Brainy Business Podcast. In today's conversation, I am joined by Eve Boudreau. As you will hear in the chat, I've known Eve for a really long time. He was a very early listener of the Brainy Business podcast as we talk about in our conversation, and also one of the first students in our certificate in Applied Behavioral Economics, which is all done virtually through the Human Behavior Lab at Texas A and M University. In addition to being head of technical sales at Google Cloud, he's also worked at shapa, which we discuss as well. We have been talking about having him on the show for quite a while, and this chat he suggested about generative AI and language models and the intersection with behavioral science and how to apply it all in business. Felt like the perfect opportunity for an episode. Really quickly, before we get into the conversation, I want to be sure you know that there are links in the show, notes for everything, including related past episodes, links to articles and books, and more. It's all within the app you're listening to and at the brainy business.com 375 now let's jump right in. Eve Boudreau, welcome to the Brainy Business Podcast.

Thanks. Thanks for having me.

Oh, yes, we, as I said in our little pre chat, which of course you know that we've known each other for a very, very long time now, many moons of time on this. But for everyone who doesn't yet know you, who's listening, you know, can you share a little bit about your background and the work that you do?

Sure, absolutely. I think I might have found the Brainy Business in the very early days of like the first maybe 10 or 20 episodes as I was just ramping up on behavioral economics. But yes, my name is pronounced Eve. Sometimes people pronounce it Eves. I keep trying to explain to them there's only one of me, but I'll, I'll answer to all of those, all the names. And I work at Google Cloud. I'm a head of technical sales here in the US Retail team and I've been here for about four years working on cloud technologies, but also on AI contact centers, working with software companies and several retailers here in the US and globally. I'm a technologist at heart, but I'm also a budding applied behavioral scientist who's been lucky enough to learn from you and others like Dr. Palma at the Texas A and M. Also from Dan Ariely and several other folks over at Irrational Labs. So very blessed to learn about the intersection points of technology and social science. I live in Atlanta with my wife Jose and our three kids. I'm Canadian, so that's why I have the funny name. But we've been here in the US and now we're dual citizens. We've been here for almost 25 years ago. But I think my life is kind of proof that a kid from a small mining town in Northern Ontario can end up in the Bay Area and work for a big tech company and be able to share some some knowledge with with the listeners today.

Nice and like. So you said think that you and Jeff may have some sort of. We'll see if there's a competition as to who found the brainy business first. But I know that you both were very early listeners and I appreciate you both very much. And that's Jeff Poole from the Human Beh, your lab as well, which has brought in a lot of that connection there. Do you know what brought you where you said, you know, you found the brainy business was like right at the time that you were starting to look at behavioral science. What kind of prompted that for you?

Funny enough, it's actually longtime friend of mine in the Bay Area who started the company Sheepa, Nadia Lavi, who is the CEO over there, co founder with Dan Ariely. And I remember the night that I was begging him in his driveway to actually let me participate in helping because the the Shape of mission. And I'm sure you'll reference the episode that, that Dan came and chatted with you on was very interesting to me because it has a personal connection for me in terms of, you know, health and, and weight and, and how to actually influence people in such a way to make it a bit more of a positive experience as opposed to the negative one there. But that was six or almost seven years ago. And from then I got a chance to meet Dan, read books that I had already read, frankly, and just got started into binging on every single book and research and podcast, which is how I ended up finding you and a lot of the folks that build this community.

Yeah. Nice. Well, thank you for that and for what you did to make that episode with Dan take place. So thanks for that back in the day. So excited that you're here now to be able to talk a little bit. You. We've been talking about doing an episode on something for quite some time, so I'm excited for this to come to fruition today. And you had proposed for us to have a conversation about AI and some of the stuff you're doing, you and, you know, the collective at Google. Do you want to share a little bit about what your, you know, thoughts were and how we got here?

Yeah, absolutely. So just, just to get this out of the way, I'll be representing my own opinions today. I can't, I can't say that I could ever speak on behalf of Google as a whole, but I can certainly share my experiences and my opinions as things are quite fast moving here. Ironically, with this whole kind of push from ChatGPT, Google and several other companies, frankly, who are doing really fantastic accelerated work in the AI and large language model space, I think it's forcing a lot of people to now look at how we get folks to adopt this technology responsibly, how we can look at the design and the adoption of the technology, what it means to displace certain ways that we do things. So, good old status quo bias. What's wrong with Google search? Turns out there's always improvements that we can make to any type of technology that we have, but what are the barriers there that actually are going to prevent people from doing these things? So I, I think I had been listening and reading quite a bit and having people from the technology space and people from the social science disciplines kind of like trying to learn from each other for the first time in a more discreet or maybe, you know, thoughtful way, what this technology means for humanity and how humanity can actually make sure that they're doing the adoption application and frankly, just learning about the technology in a responsible way that actually moves us forward and doesn't hold us back. So that's what I wanted to at least share with you today and also get your perspective on as a social scientist yourself and maybe not a deep technologist necessarily, because this will not be a deep technology podcast. You can go and read and listen to hours and hours of this stuff on tons of that. This is more of a, of a conversation around comparing, contrasting what it will take from a social science standpoint to get this technology in the hands of people in a responsible way.

Yeah, definitely. I really Liked. So you sent me an article which will talk about a post talking about centaurs and cyborgs at least as we get toward the end of how to be thinking about this, I'll let you talk about that more. But in the article, the author was talking about how there was some research they did with Boston Consulting Group and looking at when the consultants were able to use AI and when they weren't and comparing. And it was, I think 748 or something, different individuals that were part of this grouping across the globe, different consultants, and across the board really in every aspect. They found those that were able to use AI while they were doing their regular course of work outperformed those who didn't essentially every time, all of the ways and you know, having to do with the way you use it. So he talks about this like jagged edge of AI and being able to use it in the right way, but where you can increase some efficiencies and you can ask good questions and all this to be able to have it help you get. So they got more done. It was typically better work and just, you know, across the board, really effective. So do you want to talk a little bit more about that article and you know what that spurred for you?

Yeah, that's a great, that's a great summary, Molina. I think that what was interesting about the article was not only what was highlighted, but what was actually kind of tucked away in some of the, some of the results there. So how the research was conducted, of course, a lot of, a lot of surveys, which of course you can be a fan of or not be a fan of. I'm not sure how much objectivity those put into this, but I think the big takeaway for me from this article was a really super simple one. And it was that if you put this technology in the hands of people and you don't show them how to use it properly, it actually has a negative effect on the outcomes that you're trying to produce in the first place. So we've had a lot of, you know, customers and even people here internally that are very, very excited about this technology where they jump right into things. They don't necessarily fully understand how it works, what the power of it is, and then end up spending a lot of time doing more damage than good as they learn through the process. So a matter of, you know, giving them a manual and saying have at it really can be detrimental to companies. So we spent a lot of time with a lot of the global system integrators like Boston Consulting for example, to go and make sure that we provide, you know, free workshops in a lot of cases to customers who really need to understand what the technology is, how to use it properly and more importantly, what not to do with it. So we've spent quite a bit on responsible AI initiatives. We have an entire organization and group of people who actually make sure that we're doing the right thing by the technology aspect, even if it means taking more time to roll out the models that we make available to the public, whether that's via Bard or via our large model garden in Google Cloud. Making sure that people understand that you can do a lot of harm with this technology if you don't know what you're doing. And I think that's what this, this particular research, which I'm sure will be in the show notes, will kind of give people an idea of, to take this very thoughtfully and not rush into using the technology without understanding the capabilities.

Yeah, well, without getting too deep into all of the aspects here, can you share a little bit of some of the, you know, what you should use it for and what you shouldn't, how to use it, how not to use it. I know that everyone listening is there will be some that are doing a whole bunch of stuff with, you know, AI chat, GPT or whatever else they are choosing to use. And there are some that are kind of saying, I know it's there, but I don't want to because, or that's a tomorrow problem or not ready yet or whatever. And so if they're looking to step into using it and want to make sure they're doing it responsibly, do you have some insights of some clear, yes, do this and definitely don't do that.

I think there's two aspects to this. I would say that there's a consumer lens that you can look at it through like people like you and I who are just using it on a day to day basis, that's really where Bard, the Bard experiment kind of like steps in and where we expect, you know, search users or just traditional people like you and I who are just trying to understand what the technology can do and using the Bard experiment to do that. And then there's the business aspect, which is a bit different. The business aspect is the one that I'm a little bit more concerned about only because there is a lot of different ways to use data against these models that can be seen as a way to use it for good or use it for, you know, nefarious purposes, if you will, as you know, we're really big on privacy here in general and we're also big on people owning access to their own data in general, whether you're a person or a business or what have you. And so in certain situations you'll notice that when you're trying to use things like Bard or you're trying to use some of our models in Google Cloud, we actually will prevent you from doing certain things, like if you're trying to launch an attack on somebody or do some reverse engineering or if you're trying to, you know, trick the models and stuff like that. I mean, we do a lot of red team testing here. You can find some information about that from some of the teams actually that have done some good public speaking on it, written tons of white papers in order to actually kind of mimic right from a white, a white hat perspective in a responsible way, trying to trick these things into doing things that are not good. And I mean, I'm not here to debate for what people are going to say is good and not good or what's true and not true, but I think inherently we have a pretty good idea of something that's going to cause harm and how not to use those things to cause harm or to produce information that frankly, maybe you don't want to see. I'll go a bit into more detail here, going back and taking a blast from the past, I guess in both of our histories in customer service. But for the most part, I would say that I've not seen a lot of nefarious use of these things you do see on social media and YouTube and X. You know, folks kind of, you know, finding ways to make different models and different products do certain things. Those loopholes are quickly closed by any company. It's not just, not just Google. I see everybody trying to do this incredibly responsibly. And so a lot of those situations, I would say that would be negative ones or that are, would be, you know, kind of forcing you to pause in terms of, I can't believe like a large language model can do this. Those loopholes are often very closed very, very quickly, very specifically around making sure that we're using these technologies for good and not for, not for the opposite of good. I use the word evil. No, evil can't do any evil with these things.

Yeah. So as we look at jumping in and starting to use the, you know, I found with chat GPT, you know, just playing around with stuff and I, I've tried some other, we've started doing a different system that does some of the show notes, you know, first pass of show notes goes through some AI and just starting to see ways that we can be optimizing and sometimes what comes back is terrible, like super, obviously wrong in many ways. And it was a lot of, you know, searching through to find the same, the right thing. It is amazing to me where it's like I uploaded an audio file and say like give me the summary of what was in this and it will make stuff up that go well that we didn't talk about that. So I don't know why that happened here. And so I think things like that can make it to where you get a little bit concerned about anything that comes back. I know I've personally worry about some of. When we think about, you know, where people are using AI to write papers in college or wherever else and relying on data that may or may not be accurate and they don't know enough to know what they don't know. There's real Dunning Kruger issue there. And so has been interesting though to see how professors, you know, some of them have been talking about ways to. You're not going to get rid of the technology, right? So that's where it's easy for any of us to say well people shouldn't use that. You mentioned status quo bias earlier, right? And say, well if competitors are doing this, this is just their thing and eventually it won't matter anymore. We'll try to put up the walls so people can't use this and we'll block the WI fi so they can't whatever. And it's totally ridiculous and knowing like in the student case and like I said this professor was talking about, you know, that they're going to use it, it's here, it exists now. And being able to teach people how to use it effectively is so powerful. Being able to say, you know, go home and do a first draft of the paper using AI and then bring it in and we're going to have a discussion about what is true and what seems wrong and how you might optimize based on what's come back. Right. The way that what you're doing in class is different than what you used to do, but it's leveraging the tools that exist now. I really liked in that paper that we've already talked about or that article he was talking about how the AI really helped the less stellar consultants. They had a huge lift. They went, you know, it was like 43% uptick in their performance from being able to use the AI where those top level consultants didn't quite have as much of a lift because they're already, like, at the top of their game. So this, you know, lifting up, and he gave the example of mining and those who were able to, you know, dig well with a shovel into a rock wall or whatever it was, were very vital and necessary until some specific type of tool was created that made it to where anybody can get through the rock wall now. And you don't need that, you know, specific type of strength. And so seeing how this can help kind of lift everyone up and then make a new baseline, I think is a really cool thing to, to keep in mind, you know, as we look forward.

No, I, I agree with you. In fact, Ben Royce, a Googler here who is also a lecturer at Columbia University, has been on the road, I would say, for the better part of a year doing these Gen AI live labs, discussions for our, our customers and, and our partners around the country. And he actually gives a lot of very practical examples. And I think this is where kind of a lot of this, I'll call it hype to a certain extent only because people don't know how applicable it is just yet. But until we get down to brass tacks, I'm glad you brought up the podcast thing, because I listen to a lot of podcasts just in general. So I'm always observing how the applications are built in general and how good the show notes are, how good the summaries are and what have you. And it always shocks me to see the variability in not only the platforms themselves, but then the tools that are used to do these things. Right? Because I think there's some baseline things that you need for your podcast, right? You need the show notes and you need to produce proper links. And an episode summary is always good because then you don't have to write out yourself. And then, of course, there's even companies now that now that you're doing video that'll actually create, like these shorts, right, for TikTok or for YouTube shorts, caption them all for you in order to kind of drive viewership. I think in a lot of cases, people forget that just because something's being done currently doesn't mean it needs to be done forever. Like, for example, if I was to ask you, do you know how many people actually read the show notes? You may know from some of your analytics on the website, for example, but people do show notes because the status quo is that you need good show notes as a podcast. Now, ironically, in those show notes, the best part of that, a large language model can actually use to surface your podcast and search results. However, not every podcast platform has the greatest search results because they actually use different models to actually surface those results. And there's good search engines and there's search engines and that are not, not as good. Just so lucky that we work for one that actually works pretty well. But you'll notice even in our own search, we have a lot of different products. They all have a search bar in it. They don't always surface the same results either. And that's because we use different models and different products that are fit for purpose. So in a podcast, I may use specific show notes or content of a podcast, or even stuff that's in the video in order to surface a result. Whereas in Google Search, I may use a whole different set of data to go service those results. All that say that models in general for AI are very, very specific to applications and businesses, and those kind of vary. So when you have these large language models that have billions and billions of parameters and you're trying to make them do something that's fit for purpose, it doesn't always do the right thing because it can do so many things. So over time, you'll see companies like ours and others as well shrink these models down for very specific application like conversation summaries, like show notes, for example, like show summary, show summaries as a whole, in order to make sure that we can provide a very good experience for that particular function within the application that provides value to either you as a podcaster or as me as a listener. But it's surprising to me that there is still so much variability in just the quality of the applications themselves. And then underlying that is of course, some of the machine learning and the data that's required to actually make that machine learning and AI really kind of shine for the end user.

Yeah, definitely. So I had an interview with Rory Sutherland that's going to be coming out just before yours, and we were talking about, he was saying where it was asking for your top advice of things that, that you tell people and where he's had success. And he has said, you know, so being really good at Google searching, he said, is a thing that has really helped him in, you know, amongst other things, that was one piece. But saying even the, the nuance of where you can use brackets and plus signs and quotes and things that no one ever really gets taught, the functionality that can happen that can get you to really, really good search when you understand how to use those prompts and things. And so in that he had Mentioned, you know, we talked a little bit about AI and saying how now they're the, like, prompt technicians or whatever that are coming out into space now.

Yeah, prompt engineers. Yep.

Right. That's right. Prompt engineers being able to. I knew technician was wrong, but it was just what came to my head. So, yeah, a prompt engineer. Someone who's able to tell you how to ask the question to get to where you're wanting to get to or who can take you on that task. But it's the same thing I've been talking about for a whole long time in the value of behavioral economics and behavioral science and like understanding the question you're trying to ask how to ask it and being thoughtful about it, because a slight change in the question can get you somewhere completely different. And I think, you know, everything with AI and you know, these language models, it's helping to show just how far reaching that really can be.

Yep. And I can. I can wax poetic on the. On. You know, Rory's been at this game for a long time. Right. So have a lot of the agencies out there. In fact, I would say that they're probably the birthplace of. Of behavioral economics before it really even existed. Ironically, not so much by observing data as part of what's going on. Right. Because I don't really consider survey data data only because it's some what somebody's telling you. Right. As opposed to what they're actually doing, which is one of the cardinal rules that we're violating, I guess, in behavioral economics in general. Right. Trust, but trust, but verify. Right. Or more simply put, right. What people say and what people do can be vastly different, as we've. As we've learned. So I think a lot of the agencies, like Ogilvy, for example, have really kind of doubled down on these practices to go make sure that they're actually comparing between both things so that they can make strong recommendations to. To their clients from that standpoint. Ironically, you and I both got our starts in contact centers. You answered the phone at an airline. Right. If I'm not mistaken. And I used to work for an Internet service provider, and I used to be. We didn't have any IVRs or systems that would help automate these things. The line, every time somebody had an issue with their computer, it went straight to me. So I still strongly recommend today that anybody, whether you're an executive or not, you should always spend some time in a contact center, because that's where really the magic happens, or frankly, the frustration of consumers really kind of ends up being there. And so some of these applied areas, and I'll ask you some questions here because I think you've, you've got better experience than I do from that standpoint because it was a very small company and I think we were three people there. So we didn't really have any of the best practices of like quality management with our agents and training. It's like, you know, please pick up the phone, greet the customer and make sure you fix their issue was basically the training that we got. So I'm pretty sure it's a little bit more involved than what you have. But over the last four years, I spent a lot of time with a very specific discipline of our portfolio, which has been our contact center AI portfolio. And I've learned so much in terms of how we can actually help companies and consumers. And I've also learned how a lot of companies are really not interested in helping companies and consumers. You can produce all sorts of interesting data that just kind of slaps them in the face and says, hey, this is what you should do. And then they don't do it. Which then the whole social science kind of aspect takes over saying, okay, so you know, you need to do this and you're not doing it. And there's the why, why, why, why. Or you can do, do a bunch of like, you know, question storming to try and get them to snap out of it. And there's always good, good reasons for it. But then that's where all the behavioral science concepts just kind of come flooding through, where you have to really try and understand, okay, I'm trying to influence a behavior, whether it's for a consumer or somebody who's serving that consumer. I'm trying to remove these barriers, right, either by introducing friction or removing the friction. And then I'm trying to show people what the benefit to them would be in order for them to, you know, increase their motivation to go fix these things. So, for example, one of the technologies that we actually have is a product called CCAI Insights. And the Insights platform is a very large data platform that you would import thousands, if not millions of actual conversations, whether they be recorded conversations from a contact center or they might be transcripts from digital interactions, whether they're chats or SMS messages, basically they're customer conversations with agents. It's really what's happening on the front line. So when you call in and somebody says this call might be recorded for quality purposes, which I'm heard people have heard many, many times, this is a very, very large, the expression that we use is corpus of data where everything about your business lives and you can import that into this insights platform that we provide to customers. And then from there you can actually do things like identify with a large language model or even with just a regular machine learning model the topics that are being discussed from those topics. Then you can actually drill in a bit deeper in understanding. Okay, if I wanted to automate some of these things for people to self serve, what would those topics be? Where would I get started? And then you can review where you would actually build a virtual agent based on that conversational data. There's another part of the platform which that would feed into. It's a very popular platform that a lot of technologists know about called dialogflow. Dialogflow basically is a virtual agent. So today if you actually call the Home Depot, you'll actually get dialogflow and trying to route your call. Well, again, that's another natural language processing technology that we've provided to customers in order to actually have people understand what you're calling about and then make sure you're being routed to the correct department. People don't realize this, but there's a lot of money lost when you're on the phone for too long and you're tying up an agent for, you know, 15, 20, 30 minutes when it could be 30 seconds, or more importantly, when you're routing them to a different department. A lot of people don't realize that salespeople in contact centers are significantly more expensive than people that are just there to check on your bill balance and stuff like that. So I can't remember which department did you work in when you were at the airline?

Multiple. So I started in reservations, you know, the, the booking tickets and such, and was there for a little less than a year. Maybe right around a year when I moved over to what? When, when I moved over, I was hired into a department that at that time was called web support. And so that was actually a much smaller department. There were 10 people or less that were in this particular department. And when people were having trouble booking tickets and things online and they had to call like the code didn't work or something didn't load properly, this small team of people was helping with that. And within the first few months of me being in that department, they started a process of converting and combining three departments together. So it was the mileage loyalty program, the web support, and then customer care, or you know, customer relations, the really angry people. We all got brought together into one department. And also, you know, you went from 24 hour call center, from the one to being, you know, when I was in reservations was 24 hours. You then switched to a thing where it was 6am to 11pm I think was what we had the hours for that web support and whatnot to then bringing in things with the chat and having to wear after 5 or 6pm that it starts being outsourced and helping to train those teams that are, you know, taking the type of work that you were doing before. So that. And then when I worked in financial services, you know, credit unions and banks have call centers too. And so helping to be thoughtful to and interact with those as well. And actually we had an interesting project while I was there that was cross training and trying to have it so that in, you know, the particular branch that you could have staff that could both work the front as, you know, a teller and then would also work the call center so you could ebb and flow between the two for efficiency's sake. And you definitely learn some things where you need dedicated people on either side and some things that can cross over. So yeah, I actually have a decent amount of background in, in all of that space and chat versus phone calls and all that.

Yeah, it's amazing because 15 years ago people said, hey, we're going to get rid of all these phone calls by bringing everything online. It's a fantastic idea, except when the technology doesn't work and ends up generating more phone calls. Right? Angry phone calls of people who've tried to self serve. In fact, I'm not sure if you've ever had Matthew Dixon on the podcast or not, but he's done a ton of research with Rick Delisi and Ted McKenna over the years and they published an article years ago, which is quite a popular one, called Kick Ass Customer Service that basically through their research was able to show that over 80% of phone calls coming into contact centers now are actually from failed digital transactions, which ironically, with a large language model, if you're analyzing millions of conversations, you can actually do a really good job at inferring how many of those phone calls could have been prevented by fixing particular issues in a digital experience.

So I haven't had the person you mentioned on the show, but I have had John Sills and we talked about his book the Human Experience and some of the value and power of getting things right the first time and going back and fixing stuff and how expensive that is when you where it's, you know, almost always better and, you know, more efficient and effective and customers are happier if you invest in a really good service and system, you know, on the front end. So things work instead of having to invest in fixing things when they go wrong and how expensive that is. So I will, I'm going to link to his episode in the show notes for, for, for everyone. And now that I've interjected in the middle of what you were saying, continue.

Oh, I was going to say the other person that you had on, which was fascinating, who's now at Walmart, is John List. So a lot of the concepts that he talked about and the experimentation that he did at Uber and Lyft are things that are now becoming, I'm not going to say pervasive within large organizations, but I fundamentally believe that the folks that do start looking and leveraging AI technologies in the contact centers and really listening and doing something about what's happening are really going to be the ones that are going to differentiate themselves. And it's still peculiar to me that you can provide a lot of information to companies and they, they will not make those changes. And so now we're trying to get to the point of we have some customers that are also using our conversation summary. So I'm sure you had to write up your, your call notes at the end of the, at the end of your calls. You don't have to do that anymore. We have a model that does that for you as part of our Agent Assist product. And ultimately it will do it for you to a very, very high degree of accuracy, like over 85%. Ironically, when you ask customer service agents why they don't click on the button to copy the conversation summary over to the actual CRM, you get a lot of interesting answers as to why they aren't adopting the technology. So there's a, there's a lot of reactance, I would say, to adoption of AI technologies and there's a lot of motivated reasoning that happens that really, I think is going to just now start being studied because nobody likes the, the slap in the face of. I understand that this data is telling you A, B, C, X, Y, Z, and it's not what you want to hear, but it's the truth. It's what's actually happening in the data. So you can spend your time in motivated reasoning land trying to tell yourself why that's not the case, or you can accept it for what it is and do something about it. So when I talk about the intersection of AI and data and social science, these are the types of things that are fascinating to me personally that I want to make sure that we can encourage more people Like John List, for example, to really push those agendas within the companies by also sharing information. Right. The deepest, darkest secrets are in the contact centers. Right. Nobody likes to air their dirty laundry. There's a reason for that. However, if you're able to actually share some of that information in a confidential manner within your own company or even within your industry, you can make some significant changes for good. Frankly, I hope one day I won't have to tell everybody about the poor customer experiences that my mom has when she's calling to try and get help for something or even, you know, the own. The. The phone calls that I make and then I record for test purposes for my own poor customer service experiences. We can do better. And I think AI technology with some good kind of, I would say massaging from behavioral scientists like yourself can actually help people adopt this in a way that they have a little bit less apprehension about doing something about what they know to be true in their contact centers.

Yeah, well, I'm definitely a couple other. So I'll link to John List's episode in the show notes as well. And also I'm gonna link to. It's amazing. I was realizing as you're talking, how often we end up talking about contact centers and call centers on the show kind of, and where people are doing some work there, I think partially because there's a lot more data there than there are in a lot of other parts of an organization. And so. And you get, yeah, the, the tracking of. Of different things. But so Jonah Berger, in magic words and when he was on the show, he talks about different language within call centers and effectiveness and simple changes from like, abstract to concrete language, how that can help and make a big difference. So I'll link to that. And then also with Richard Chataway when he was on talking about some examples in his book the Behavior Business and how these kind of counterintuitive ideas going from, you know, where people were not remembering their passcode. Right. Like how often we've been told, you have to. For the bank, you know, what's your passcode? And in the case of this particular financial institution, they were saying something along the lines of, and if you don't get it right, there's nothing I can do to help you. And it's going to be seven days, your account will be locked. And like, so you better, like, make sure you know it sort of.

Yeah, yeah.

And they're trying to get people off the phone quickly because they're looking to reduce call time. And what they found, you know, is that Stress where someone's was trying to get information because they're applying for a mortgage or they're doing something, if their account gets locked, it's going to be a real detriment. So their brain would get totally flooded with the wrong chemicals, they would get flustered and then they would not be able to share their passcode. They couldn't figure it out. And so then you have all the steps, like you said, of fixing the problem. The person you have to transfer to, who's now going to try to unlock and fix it, and then the next person and the next person and back and forth and that person's angry and wants to talk to a supervisor and, and, and Right, all this stuff that happens.

Yep.

And so it was actually much more expensive and took a lot longer. And what. And you have frustrated customers. Nobody likes that. But then when they had the person at the very beginning say, and you know, take your time, don't all, you know, all is well, I'm here, you got plenty of time. People were more likely to get the first time, you know, the passcode correct and it eliminated all this extra stuff. So telling people to take your time actually allowed them to have shorter call times, which is seems, you know, counterintuitive. But when we understand and think about the behavioral aspects of it, it makes total sense. Which is, you know, my favorite thing about behavioral economics in general.

Yeah, yeah, yeah. And without making this like a product pitch for us, I think you already see this through most Google products. It's not in your face in terms of how we use AI technologies, but these contact center products that we have, we have four pillars of them are all infused with AI, machine learning and generative AI in order to make the jobs easier for the agents, but also to make the actual experience for the customers just a better and shorter one. I can assure you nobody wants to contact your business that they don't have to. They really, really don't. And this notion that you have to delight customers when they're calling you, you need to provide them with answers to get them on and off the phone or in and out of chat. And there's just so much that we're learning in terms of what we would think are just common sense things that just don't get implemented. And there's a lot of like agent behaviors that, you know, the herding aspect of it than not wanting to speak up or, you know, feeling like they're going to be, you know, made fun of or just the status quo bias that comes into it. The habits that they actually create, you know, the time discounting with, I'll do it. You know, I'll report this issue tomorrow. All these are things that computers are incredibly good at doing consistently at scale that can free up people to be people with consumers. That's what the agents do best. It's not writing conversational summaries at the end of a phone call and identifying which products were discussed and how many coupons you sent, you know, because there was like Indian peel moths in your. In your dog food. True story, by the way. But we're just at the cusp now of getting, I would say, a lot of our avant garde customers understanding the amount of information that they have in these call recordings and in these chat transcripts and how they can actually use that to really, really give themselves a big, big edge in their business. And so I'm really, really bullish on when social science meets AI, we can do some pretty magical things.

Yeah, definitely. It sort of tied in and potentially, again, not to keep tight. So I. The interview with Rory, like I said, that's coming out just before this one. I was talking with him earlier today, so it's also in my brain here on this. But one of the things that I wanted to mention, because they're talking about forcing people into what you think they should have to do and what the experience, like, what I would like in this experience, sort of deal, like you're saying, nobody wants to call your business. I don't want to have to deal with that. And I was reminded of there was a meme or whatever, a thing that was on my Instagram. I think, because I'm an old person, I go Instagram and not TikTok to where I'm sure this thing originated. But it had said, you know, I went to click on a, you know, reservation or a thing for, you know, learn more button, whatever, for a restaurant, and it popped up a. Like a FaceTime video chat with them. And I cannot stress enough how much I don't want to do that.

Wow. How bold. How bold. Wow.

And so it's potentially. It was a mission. We'll give them the benefit of the doubt and say it was a mislabeled sort of a link or something, but, like, force assuming, like, people would love to, like, chat with our agents and have the. The visual experience of someone that's there when you call the bank or whatever. It's like, no, I'm doing this thing at 2am and I hate that I even have to talk to you at all. And I don't want to have to be like, dressed up because someone is there for this whole situation. Like, nobody wants that thing. But on the inside of an organization, it can be easy to forget and think, oh, someone definitely they'll want this, like, amazing experience, but maybe we just need to keep it so they don't have to. To call that the service is better taking, like, going those like, two, three, ten steps back and forward and looking at things in a different way so that we can, you know, solve those problems a little bit differently.

You did a fantastic podcast, by the way, with and I forget the, the lady's name. I know she was French and she was with Content Square.

Ah, Lucy Buisson.

Lucy. That one was very, very good as well. Because there's no need to guess at what people are doing on a website or whether having issues and frustrations. In fact, there's a couple of other companies in that space. Full story here in Atlanta, actually, and Quantum Metric, who's one of our customers. And it's amazing that this is a whole other data set that can be used by AI in order to infer that somebody's having a particular issue where you can actually reach out to them. While the frustration is occurring to your point earlier about people getting so flustered and frustrated and then just kind of getting onto the phone and just, you know, blasting the agent. Poor agents, this is a way to kind of like, you know, meet them in the moment where if they're actually having issues booking a flight, that you can actually use those signals from that friction that's happening on the website and then interject with a live agent to come in and offer help or actually just surface an answer. Right? Just give them a quick little nudge of looks like you're using for looking for your points balance on delta.com here's the link for it. Because our web designers haven't figured out how Eve can find all of his trip credits and flight credits easily without 10 clicks. Shout out to anybody at Delta who can help make it less than 10 clicks to get to that, by the way. But this data and the conversational piece, it's just, I think the avenues are limitless and I think what we're trying to do specifically with things like Bard, are to understand how we're going to change and adapt to the search experience. Because search experiences are typically three and four words and we magically will just point you in the right direction. We do expect a little bit more when it comes to a prompt, we do expect a little bit more information because you're trying to do potentially something that's a bit more complex. But then we also want to understand how we're getting things right. Right. That's why you have the thumbs up and thumbs down buttons there. And you can double check the results to make sure they're grounded in some reality of Google search as well by clicking on that little Google icon. And we're trying to be very transparent when we publish all the different releases. So we just updated Bard with our newest Gemini models late in 2023. And you can provide feedback right in the interface. You can copy and paste things. So we're also learning how people are interacting and trying to use that actual user data itself of the points and the clicks and frankly just the feedback that they give us in order to improve these products so that we can make them actually usable and save people time. Because this is another self service mechanism that hopefully will stop people from having to contact the business to get information or sit on hold for an hour like I did with ADT to cancel my service, which was not a good experience.

Yes, I, I was definitely, you know, reminiscing on the call center when I worked there. And so in that small team on web support there was this perfect moment. So I worked 2pm to 11pm when we were in that and it was because I was going to college, you know, so it was actually a perfect shift for me. So I did, I did class from 8am to 1pm and then I worked from 2 to 11 Monday through Friday every single day and was working on, you know, then having to be doing research and reading and papers and, and all of that in my undergrad. But it was nice, you know, there's some lulls in the contact center where you can be doing some reading and stuff. But we had a point where so There were just two of us that worked until 11pm and there was a rush where people would be calling in that, like six o', clock, seven, you know, whatever. And if one of us got hung up on the phone where somebody's taking a long time, then the other person is having to answer everything in the queue. And if somebody's holding you hostage in that way that you can't get off the phone, you get stuck and it'll make it to where then the wait very quickly becomes an hour long or whatever as you're trying to get through the queue of people. And there in multiple behavioral things. On this one, there was one other person in the department that was finishing his shifts that would go until, you know, seven or eight or something. And where we ha. We look at incentives and structuring. And I will never forget this person that he would have where it was saying we had to do nine calls per hour or 11 calls an hour or whatever the thing was. But it was like a. The minimum, right? That's what they want you to do. But he would hit the number and would put himself on call work and not answer another call until the new hour started. So not going on an average, just would hit the number and be done. And if he hit that, you know, 15 minutes in, because they were quick answers, then he'd just be out for the rest of the time. And me and the other guy that was working, like you're the people you are punishing is us, right? Like you're hurting us because we're totally stuck here. And I get the point you're trying to make. And like this is what I got paid to do and I'm not going to do more. It was like where people say quiet quitting is like a new thing. Like it's not. And I have proof from this person. But you would also have then customers that would be on the phone and would say, I waited hold for 45 minutes and so you're going to stay with me until I'm done because I don't want to call back and be stuck again. And you say, well, the, the reason that I we're stuck is because like you're perpetuating the problem, not fixing it. Right. But when we get in the moment, it feels like you're owed something. And the way we think about things, it's, it's just a really interesting thing. You can't logic someone in that moment to get them off the phone. You just have to say, of course, I fully understand. I'm here and happy to help in any way that I can. I'm here for you. But it's, it's just a really interesting thing where that human side always matters. Even when we're looking at the technology, be it in phone systems and chat, AI, whatever else. It just. All the way we produce, where we put effort and intention and the way we think about problems, all of that understanding those aspects of human behavior, of course, are so important, which I'm, you know, listeners, they, you're here, you get it. You know, that's why you're here listening to the show, right?

Yeah, the co. The cobra effect is real, right? I mean these, these perverse incentives and vanity metrics are everywhere. And it's something that it would be Great. If we could break through those at some point in, in the. Just do it for Eve's mom. Let's just do it for Eve's mom. Eve's mom, your mom. Anybody deserves a better, a better experience. I mean, nobody wants to be wasting their time on a phone for that long. Especially the agent certainly doesn't either. Nobody wants to be there. So how do we prevent that from happening and how do we improve the self service aspects? I think is something that when this technology and social science meets, I think we're going to do some pretty magical things. I keep using the word magical. I'm not quite sure if it's a Disney day or what's happening, but.

All about it. Maybe it's Rory. It's the alchemy of it all. Could be.

It could very well be. It could very well be.

Well, Eve, thank you again so much for joining me on the show. It's been a lot of fun to chat with you about all the things today. For everyone who wants to continue a conversation and chat and learn from you a little bit more, you know, what's their best route to do?

So to connect, you can find me on LinkedIn. That's probably the easiest way. Or you can just email me. It's. You can just email eve@google.com.

Perfect. Well, we will put the link to your LinkedIn in the show notes. We don't do emails in the show notes because I do remember that spam. Yep. It's a real thing. And for spelling purposes as we've already identified, we'll make sure that that's linked for you there so you can check out and connect with Eve. So thanks again for the conversation. It was a lot of fun. It was magical, as they say.

It was magical. It was magical. Thank you, Melina, and thank you for everything you do for the, for the behavioral science community. It's very well appreciated.

Oh, thank you. Thank you again to Eve Boudreau for joining me on the show today. What got your brain buzzing in today's conversation? For me, it's sort of a side item, but I think it's so interesting to realize how often people talk about research in call centers and how often we've discussed it on the show. Eve and I chatted about this a bit after we stopped recording. And the best guess is that this is because it's the data that's most easily available to many businesses. Call centers have a different level of metric than most other areas in a company. Even in an environment where nothing is being tracked throughout the organization. The one exception is likely the call center. Plus, you typically have access to recordings and notes and logs. There's a wealth of information there that makes it a great place to start. Whether you're using AI and machine learning like Eve and I talked about today, and like Jonah Berger did in his research that is showcased in Magic Words and our conversation when he was on the show, or if you're trying this all a little more manually, there are lots of insights there, and I think it's also really important for everyone to remember, myself included, that AI is here and some version of it will be here to stay. That article we talked about with the research at Boston Consulting Group is linked for you in the show notes, and I think it does a great job to help any of the laggards to come off the sidelines and realize it's time to start embracing AI. Whether you use chat, GPT, Google, Bard, something else you've already found, or things that are yet to be invented, there are plenty of opportunities to help you begin to incorporate this into your work and gain back time and optimize the work you're putting out. As mentioned in the conversation, we're using AI here at the Brainy Business to help with Shownotes. There was a lot of research to find the one that worked on for us that was close enough to my voice and a good enough summary to allow someone else to take this on and allow me not to have to touch them anymore. Living the dream over here. I have also used it with clients. One example is with an ability to quickly calculate the sample sizes we would need to prove statistical significance on an email Drip campaign optimization project we've been doing. I have the questions queued up and it helps save tons of time that I can devote to other parts of the project and quickly, quickly answer questions like is this difference enough to know that we've hit the target, or do we need to increase the sample to be sure I can get that answer back in seconds instead of spending considerably more of my time or someone else's on the team to run those calculations, which is awesome. What about you? How are you using AI in your work so far? What do you wish AI or other technology could do for you and your business and how it all ties in with human behavior? Share it with me on social media. You'll find me as the brainy biz pretty much everywhere and as Melina Palmer on LinkedIn. And of course you should tag Eve as well. There are links to connect with both of us in those show notes as well as related books and past episodes, a full time stamp summary of the episode, and more. It's all waiting for you in the app you're listening to and@the brainybusiness.com 3. And thank you again to Eve Boudreau for joining me on the show today. It was a delight to chat with and learn from you. Join me Tuesday for another Brainy episode of the Brainy Business Podcast. It's going to be a lot of fun. You don't want to miss it. Until then, thanks again for listening and learning with me, and remember to be thoughtful.

Thank you for listening to the Brainy Business Podcast. Molina offers virtual strategy sessions, workshops and other services to help businesses be more brain friendly. For more free resources, visit thebrainybusiness.com.