Revolutionize Your Business with Realtime Guidance

Get Your AI Strategy Sorted https://activatecx.arroyo360.com/ai

Join the conversation on contact center ai solutions using behavior optimization and uncover the impact of agent assist in the contact center world with Episode 29; "Expose Your Blind Spots." Explore how real-time guidance can transform agent-customer interactions, leading to improved outcomes and a more efficient workforce. Don't miss out on this insightful video - hit subscribe for more enlightening content!

Chapters
00:03  The Balto Origination Story
03:32  Changing the Conversation
07:28  Baselining the Organization
08:34  EX and the Big Picture
12:36  Follow the Money
14:20  The Chicken or the Egg
16:32  Know Thyself
21:06  You are My Density
25:31  The Nail in Your Forehead
28:14  Thanks

Chapters

• 0:03 The Balto Origination Story
• 3:32 Changing the Conversation
• 7:28 Baselining the Organization
• 8:34 EX and the Big Picture
• 12:36 Follow the Money
• 14:20 The Chicken or the Egg
• 16:32 Know Thyself
• 21:06 You are My Density
• 25:31 The Nail in Your Forehead
• 28:14 Thanks

Transcript

0:00

Hey Mark, welcome to the show. Thanks Frank. Great to be here. Awesome. Focusing on a niche in business is often Really thought of as a great strategy for growing a business. And in fact, there's a term that says there are riches in niches. And I like that term because a lot of times people don't really think about narrowing down their scope. And so what inspired you to develop something that is real time guidance? In a contact center, niche, niche, and then go all in on AI. Well, it started with the smallest possible version of the problem. And it's one that I experienced myself where I was on the phones. I was in sales and I'd go to my manager's office for coaching. My manager would pull up the call recording, give me a bunch of really good advice good coaching. And then I would get to the actual call with the customer. And forget to apply all the stuff I just learned in my coaching session. And that's just something that everybody has experienced. It's a human thing. So I first wanted to solve that little problem just for myself. How can I take the things that I'm learning from coaching and actually apply it when I'm actually talking the customer when I actually need it? And it started out actually creating. I created an Excel macro where I could take all my coaching notes and I put them in the map in the Excel sheet at the end of the coaching session. And then when I'm on a call, I could type in a keyword that I wanted to search for. So I could type in implementation and then I'd hit command shift T for talking points and it would pop up my notes for implementation. So that's kind of the first version that I just built for myself. And, from there when you really can nail one tiny piece of it you're kind of pulled gravitationally to expand further and further as customers ask for more capabilities. And you have new stakeholders that hear about your product and are excited. So it started out just building real time guidance for agents, analyzing everything they're saying live on a call. Everything the customer is saying live on a call and giving the agent prompts on their computer screen to help them be as effective as possible right there in the moment. That's the problem I experienced. And from there we got pulled to serving a whole bunch of different stakeholders in the contact center. Cause it's this big interconnected beast. It's not like a Figma where you can just serve like a design user and then not serve the other, like everyone, the contact centers and everyone's business for better and for worse. Yeah. So we got pulled to serve the supervisor because they said, Well, if you're gonna give my agents guidance, I need to know what you're guiding them on. And can you maybe give me some insights on how I could do a better job coaching? And then we got pulled to the quality team that said, Hey, you're telling all my supervisors all this coaching insight. Can we make sure it aligns with our company's scorecards? And then the executive said, Oh, nothing, the dead end. But if you keep pulling and pulling and pulling your end, you end up getting to like a really great result. It's actually how we found our first investor, our first investor we ever had in the company. I was doing a sales demo and the person said, Hey, I don't want to buy your software, but I think I might have somebody who would be interested in investing, I can introduce you. And then I met that person. They said, Oh, I'm not interested in investing. But I think I might have someone who's interested in investing and that person ended up being our first investor. Follow the thread. I call that the moving boxes theory, which is like sometimes you just can't sit there and stare at the problem. You have to get up and start moving boxes. And once you start moving boxes, what's behind there will always surprise you. From the standpoint of, once you started putting this out there into the contact center world, how did you see that start to change the contact center agents conversation? At first in very subtle ways, and then in much more concrete ways. One of the first things that was so interesting is noticing agents ask more questions and talk less. And that's because one of our first prompts we had was, you're talking 90 percent of the time, try turning over the customer and asking these three questions. That was one of the first things that the software did. And that was from, again, a personal realization, because I was like, man like talking too much is one of those things, there's just so habitual, so ingrained, you can get coached on a million times and still do it. So what if there was, and that's something that people want, they want like a gentle tap on the shoulder, like, hey, you're rambling a little bit. So it started out with just kind of these very like subtle little behaviors that they did differently. And then we started to realize you can structure these behaviors, right? Instead of just asking more questions, you can say, what kind of questions? Can they do a better job? qualifying. Can they do a better job closing? Can we go around everybody in your contact center and go to your top agents, all the folks who are really excelling and say, Hey, can you all put together a list of your top 10 closing questions and then we'll put it in Balto and we'll be able to AB test them and see which one works the best. Like that's the sort of specificity you're able to get to, so you're able to see people like doing more closing attempts. You're able to see people, the, the agent said this, and then the customer said this, you can kind of try to put the two together. We're actually able to get experimental data, which is we prompted these users to try this technique and when they tried the technique, here's the results. It got so the ability to be able to prompt people creates this very unique data set. We're able to say when this thing, this technique is used. The result is a 34 percent win rate versus when it is not used. The results of the 28 percent win rate. That sort of fork doesn't really exist in traditional speech analytics data. Interesting. This is really all about behavior optimization, isn't it? It is, and the, why are you optimizing behaviors or ultimately optimizing behaviors? Because your organization has a hypothesis that a certain set of behaviors will lead to better outcomes, especially if done consistently. And you can ask any sales leader or customer service leader can you just tell me five things that are really important that like, if your folks just crushed these five things, then you would have better outcomes and everybody can name them. And then you can ask, okay, well, how often do they do those five things? The answer is never. I can't get them to do them. So why is there that disconnect? Well, behaviors are in service of, of outcomes and people tend to forget the coaching that they had we've published a lot on the Ebbinghaus forgetting curve and actually surveyed thousands of agents and said, how often do you forget to do what you're supposed to do on a call and agents say that, they forget their coaching 90 percent of the coaching after a month and they forget 70 percent of the coaching after a week. So, every month you're like retraining on the same stuff you were previously training on. We really focus on driving behaviors to drive outcomes and try to tackle that fundamental gap, which is just people's like brains and memories are limited and that's to no fault of anyone who's on the floor. Yeah, that's really true. And around that memory, you draw particular conclusions and you set up assumptions. Your organization is functioning in a particular manner. And then ultimately you have something that shows them evidence to the contrary. How often do you get that feedback from the customer that they have that aha moment that, what I was thinking about the organization is truly not what was happening. Yeah. Pretty often, pretty often. And to the extent where we'll actually show people the data in the beginning, what we'll do is when we actually turn on Our technology will have to the first two weeks will have it switched into real time intelligence mode, which is just gathering data, not giving the representatives any help. So we're getting a baseline. We're getting a baseline. We say, all right, you see how often all this stuff is happening. And then we flip the switch and they see all the behavior spike. And it's like, of course, this is an obvious proposition. If you prompt people with the thing they need to do, right. When they need to do it, and it's right in front of their eyes, they're going to do it more. So it's pretty remarkable to be able to, Create that sort of data baseline and then like watch the behaviors change once you actually introduce the real time guidance aspect. That's straight up evidence right there. But obviously it has a downstream impact of customer experience. So better agent experience, a better customer experience. How do you see that the Balto AI indirectly impacts the customer experience at large? The first thing I think about is a few years ago, call it like right before the pandemic, 2019, 2020. Early 2020, the CX world was a buzz with like empathy and customer delight. It was about a lot of empathy in your interactions and a lot of customer delight. And people were saying, how do we go above and beyond and create great interactions for our customers? Then the pandemic hit and call volumes surged and hiring was harder than ever. And all of these contact centers were understaffed and their wait times and hold times went up and their first interaction resolution went down. And the philosophy shifted or folks said, maybe the most important thing. It's just handling the customer interaction as fast and accurately as possible. And I think actually the consumer mentality has shifted toward that too. Like a little bit of empathy, a little bit of , thank you and nicety and let me try to do something extra for you. I think that's always appreciated, but I think when you get, on the line with somebody, And they just immediately say, got it. I see your problem. I'm here to fix it. I'm going to do a, B and C you're all set. You're good to go. I think that's what folks really, really want now. And truly one of the best customer experiences I've ever had was actually with one 800 contacts where I had accidentally they had my old address on file for sending me the contacts. And I didn't have my contact lenses and the representative in, it was under three minutes, it was in the twos. I remember got my address updated, got new contact shipped out and also gave me a very polite slap on the wrist. It's like, Hey, if you have a wrong address again, we're not going to give you the next one's free. And I was like, great customer experience. So our focus is how do we deliver the right information to agents as fast as humanly possible. And we're seeing that translate to improvements in customer experience and also to the contact center metrics. Where it's not like you're having to quote pay extra for customer experience, customer wins and the contact center wins. Competency is the best way to build a relationship and solving the problem is paramount. And so ultimately what you're talking about with your solution is all the aspects of how we create a relationship, people going down the path of being empathetic being friendly, really the people getting on the phone, asking for help, just want the problem to be solved. And, and in the most efficient manner possible, that really reflects some level of respect around the time that these people are having to come out of pocket to take care of this problem. It may not even be a problem. It could actually be something where they want to buy a ticket. They want to make a reservation. Regardless of the fact, they don't need to have a lot of smoke blowing their direction. They just need the problem solved. Think about like when you've interacted with someone who is super nice, but also kind of clueless, like that's, it's like an odd type of frustrating interaction where you're like, Oh, I really. I want to be frustrated this person, but it's, I can't, cause I feel bad if I'm frustrated because they're super nice and I, I feel bad that I can't like let them off the hook and, and they keep trying to solve my problem. And I'm like, no, thank you. I'm fine. Like, I don't think you're going to, we're going to get this figured out. There isn't a very unique type of customer stress that that creates a nice and clueless. I think the competence is like the most important thing. And if you can be competent and be warm and personable, then you can be an all star. Yeah, that's a big win. Probably when you first started, was it more focused on customer support or was it equally focused on the sales and, and revenue generation side of the business? Yeah, actually the other way around it we started focusing very much on the sales and revenue generation side. And the reason was very simply what we call a one step to ROI, which is if you deliver a recommendation. To an agent live while they're on a call and the agent says something better than they previously were. They were going to just let the customer go. And they said, actually, can I ask you a question about why you have that concern and they were able to save the call, that's a call. Save that's dollars in the bank. That's a sale made. It's one step to ROI in the service world. When you save agents time. You reduce handle time, the, you then have to recapture that time and either them doing other work and being able to generate value for the business in that way, or letting agents roll off the payroll and, over time capturing savings from that time savings. That's two steps to ROI. So we said, while we're trying to validate. And make sure that this product can really drive value for customers and that we're able to build a fanatical customer base. Let's start with the one step to ROI use case, and then move the two steps. And it just so happens that now the two steps to ROI is easier with things like automatic note takers that weren't really possible in 2017 when we started the company. Now you're able to just like shave off easily 30, 60 seconds of after call work for the agent because you just wrote all their notes for them. Yeah. It's a big deal. Where people are trying to break down what they're not seeing inside of the CRM, understanding the pipeline and its movement, and ultimately it comes down to rehashing some of the conversations that the salesperson is having with the Fundamentally, there's a question that says, did you ask them this? Or did you ask them that? Do you feel that your solution is getting to not only help the agent ask better questions, which is probably one of the toughest things to teach, but also at the same time, is that data being transferred into the CRM so that you can clearly see that the appropriate questions are being asked? That visibility happens on all sides. And I remember when I was on the phones myself, there was this one time where, it was a very short sale and it was about 15 minutes. But I was, it was a 15 minute grind. We were going through it and grinding and going through it and grinding. And it was like a picture perfect call. And it was beautiful. I remember like hanging up being like, Oh yeah, that was excellent. And then my manager turned to me and said, Hey, did you qualify to make sure on the right CRM? And I was like, no, I forgot. And he said, call him back. And I remember, cause we, we ended that call with like, thanks. You too. Oh no. Great. Really great talking to you. All right. Well, congratulations again. Goodbye. And then to have me call that person back 30 seconds later, big high, I forgot to ask you a question, but I have to ask you for you to be qualified. Like I remember the feeling, I remember that moment. So we first start by helping. In the agent, make sure that they remember that. And then there's analytics everywhere. The supervisors can, can pull up every interaction where that question was not asked. The the quality team can say whenever the question wasn't asked, score it down by 10 points and show me the list of calls where that happened. So we take that data and bring the visibility everywhere. So all the stakeholders know. What's happening and even the percentage of time so you could say this agent is asking the proper qualifying question 95 percent of the time, whereas the other agents are only doing it 50 percent of the time and you're able to even have a stack ranking of who's doing that stuff most often For Balto you're in this category of agent assist. And there's a lot of players in it. Some of them have been around for a while and they're evolving. How do you differentiate yourself? If you were to say, Hey, these five things are something that we do particularly well, or we do, and nobody else does. On the agent assist side, there's two big things that we do better than everybody else. The first is investing in the agent experience. A lot of people, when they were creating these technologies kind of took the perspective of, well, it's going to be a big brother tool and agents are going to use it because managers are going to make them and. Then they're setting themselves up for these like internal culture battles where, agents are staging a coup and saying, we don't want to use tool. It makes us sound robotic and managers say, well, you have no choice. And like, nobody wants that. So what we realized is that we need to go above and beyond to win. The hearts and minds of the agents by letting them personalize the guidance and make it their own. So nobody else has that capability where not only can the agent personalize like the look and feel and they can change the color scheme and the font size, some folks who don't have as good eyesight say they want it bigger. Some folks who want to maximize screen real estate say they want it smaller. They can personalize the location and then they can actually change a bunch of recommendations to fit into their own words. So if they say, Hey, I get the gist of what this recommendation is going for, but that's not how I talk. That's not, I don't want to, that's not how I behave. I like to kind of do it different. They can actually edit it and change it and the system will recognize those two is the same and kind of aggregate that to their analytics and the managers are getting a full view of here's all the stuff your folks are changing. And if everybody happens to be changing one item That's a pretty good signal that like, maybe that's not a good part of our script and all the agents are quietly telling us, Hey, you need to change this part of your script. So agent experience is the big one. And then our note taker is also like way more configurable than others, but I can talk about that later. So, so obviously if somebody changes the script and goes bonkers in terms of positive outcomes, that's something that gets reported back as well as somebody who changes a script. And it's, an absolute dud in terms of getting the response that you would expect to get from that particular script point. Totally. And managers have the ability to throttle it and say, Oh, that change isn't that's not a good change. I'm sorry. I'm going to revert it. And then also the changes automatically socialized. So, but once somebody has a change and manager is not saying that's, sorry, I got to get, I got to stop that. And the change ends up producing a lot of positive outcomes. Other folks in the contact center will start seeing that on their screen and they'll start seeing, Hey, this is something that worked that has better outcomes than the standard version. So it's almost like this like crowdsourcing or this viral effect, which by the way is how a natural sales or customer service floor functions. If you're in person, how do you learn? You listen around all the people you go, Oh, that, that sounded good. I like that. I'm going to do some more of that. That's like, how will you do it? So let's take that and like replicate that sort of human virality aspect in the digital world. That's something that we've really put a lot of time and attention toward. There's a lot of tension, obviously, in anything that has performance attributes around it. And then you throw on top of it, a listening platform, that's gauging and monitoring everything that you're doing. I would imagine that coming out of some of the implementations that there's a lot of good humor that comes from this as well, because human beings are, just by the very nature, very funny and even awkward conversations, and battles that are had, are even at their very root, very funny. Do you find that being somewhat of an outcome that when you're working with clients, that there's opportunities to look at it and, and laugh a little. 100%. In fact, we actually one day at Balto, I think it was last year. Had an internal contest of who can source the funniest stuff that agents are like doing and saying through through Balto. And there were, there were some, some really good ones. And I don't think I can share. I don't think I can, I'm good. Right. They're almost unshareable, but yeah, that's great. The best ones are part compliance violation, part like one, two closing punch. No, it's not, but you get what I'm saying. So much of what you do puts you in a position of having to be an innovator and a visionary. What do you see as some of the evolving aspects of the contact center industry that you're excited to address? I, gave a talk about this last week. I think there's three big phases we're about to see. One is a deeper, more advanced prompt engineering. Where now these LLMs, large language models are able to be used by anybody. And for those who aren't familiar, like chat, GPT is example of a large language model. You type in a query or a question or ask her for something and it can spit something back out and prompt engineering. Allows you to write really thoughtful, really complex prompts that get really incredible outcomes. So if you think about it, when you're using chat GPT, you send one prompt at a time and it gives you back an answer. What if you could tell chat, chat GPT to do 10 instructions first, do this, then do this, then do this. And if you can kind of string those or chain those together, you're able to almost have it perform a whole bunch of operations for you. So step one is going to be really advanced prompt engineering. An example is, like you can set it up. So at the end of every interaction, it'll ask a question about. Did this thing happen on the call? Yes or no. Or what was the customer's reason for calling? So now you have at the end of every call a prompt that says what was the customer's reason for calling? And if you're doing that on every call you can then start to trend them over time So you're now able to get like contextual call reasons then lom is pulling out not just looking for billing issue Looking for like, can't get service. Like you're not just looking for like keywords and phrases, which are like hilariously wrong, you're using LLM to be able to figure out what the heck happened and turn it into data. If phase one is prompt engineering, phase two is intelligent screens. And phase three, I think we'll have like a natural language agent that actually people are interacting with just like a human agent at human capability. That will do a lot of the work that doesn't require human touch. Interesting prompt engineering is a fascinating topic because if you think about AI and what's happened with chat GPT, when I talk to people about AI, I ask them like," do you actually engage with chat GPT or some form of AI authoring tool"? Because I know that if they are, that they're having to write a prompt and a prompt is fundamentally, just an elaborate question. And if you think about just overall success in life, it's really not important that you have all the answers. What's really important is that you're able to identify the need for a question and be able to articulate it. And the fact that that's probably not a skill that is as common as you think. That verbosity of writing that question because you just went through a breakdown with key words that you would use maybe if you were using a data analytics tool and the data analytics tool is having you work through some form of cubing of the data and breaking it down into various different tiers. Look first by this. Look second by that. And it's very restrictive and the large language model gives you the ability to play up on this verbosity. And if I might even say like interpolation of the information, right. To tease out blind spots, things you really should be looking for, you're not. Do you see that prompt engineering is going to lead us to that type of betterment? It already is, and it's excellent at it. When you think about setting up traditional speech analytics, you say, what was the call reason? The way that you would do that is you would put in a bunch of possible call reasons and say, track all these with LLMs. You can say, give me my top 10 list of call reasons and then track those. So And I, usually the, the 10 you would put in are different than the 10 that the LLM finds are actually the truth. So really, really incredible that you don't have to start from a place of making a guess. You can actually start from a place of probing the data and then from there, working off that data and seeing what says, I don't know if you ever saw the Video from Franklin Covey, which is pretty well known. It's been out for maybe about 10 plus years now, maybe even longer. But it is a video to bring about attention to blind spots that we all have blind spots in our life. Yeah. And it's the boyfriend and the girlfriend and the girlfriend's complaining about these just radical headaches that she's. Experiencing. And then finally they pan out and she's got a massive nail just stuck right in his forehead and he's been trying to get to the point of helping her to uncover this. But I think blind spots are just natural part of being a human being and running a contact center is very, very challenging. And then under any condition, managing a large group of people is another level of challenge. So you've got the pressure of a contact center, their outcomes, their metrics, their KPIs and then just human beings involved in this. This is why I think that fundamentally, when you look at AI and you look at all the things that are being applied in the contact center, voice bots and chat bots and things that we're trying to do around deflection and containment, that this pocket of of AI and its application has Something that is so readily applicable, like you have a group of people, you're already working on things. This is not new. You're just optimizing what is in place right now. Do you, find that you have to struggle with objections from clients as to why they would want this or are they natively seeing right off the bat how this can help them? They're, they're definitely seeing right off the bat how it can help them. The bigger challenge and hurdle is the human component and you were talking about the pressures of the contact center. And, , to just say, yes, we can see the need. Let's do it. The thing that we really counsel our customers on is, like, this is a technology that will be used by humans for humans. It's going to be implemented by humans. It's going to be used by humans. When humans use it, it will impact the customer, other humans. So I think a lot of times people get technology, especially AI, and they say, Oh, I'm going to implement AI and just leave it and treat it like it's a bot doing a job. And the answer is it's a bot doing a job that is impacting a whole bunch of people. And the people are the unpredictable actors here. So you need to kind of continue to iterate it and refine it and make sure that like you're reacting to the various reactions that people have because ultimately technology is just one part of a people centric system. For sure. Brilliant conversation, Mark. Thanks for being on the show. I appreciate you! Frank. You too. This was awesome.