Conversation Intelligence 101

Joining us today, for one of my very favorite topics, which is conversation intelligence. Your customers are talking. Are you listening? My name is Marcy Reardon. I lead the global analytics practice for TTEC Digital. And as I said, I'm just thrilled to be here with you today. I hope you're having a great day and, hope you get some value out of, this discussion on conversation intelligence. And, you know, I'm so excited to speak with you about it today because this particular topic has really gained momentum. I would say in the last year or so, I've seen a lot of interest and also a lot of value coming from the area really picking up. And I think that's likely due to a few things. One is, kind of the confluence of a number of things. Gen AI really taking off, the abundance of data, which keeps getting greater and richer, and also a heightened interest that we see from businesses in being very data driven when it comes to looking for business improvements. So as far as the agenda and what I'm gonna be covering in the next, let's say, forty to forty five minutes or so, talking to you about how to drive business value from conversation intelligence, talk about some very specific use cases that we've seen out in the marketplace, how you can actually get pilots off the ground, and then, a behind the scenes exploration of of our specific conversation intelligence solution. Okay. So about conversation intelligence. What is it? And, you may be wondering how it differs from traditional speech analytics. So first of all, conversation intelligence. You know, what is it? In its simplest form, it's the analysis of the conversations with your customers. Now these these could be through voice channels or through digital channels. It really doesn't matter. But it's taking that unstructured data and analyzing it in a way that's going to be valuable for your business. Now along with that conversation data, which by the way, on its own can give you a lot of rich insights, But the difference, let's say, between speech analytics and conversation intelligence is we're then taking that conversation information, and we're integrating it with other data, information about your customers, your agents, your operational data. And we do that for a couple of reasons. In one case, to answer some very, very specific business questions. You may have some challenges, some pain points, some friction points that you're looking to explore and answer through that conversation intelligence, and perhaps to uncover some information about your business that is sort of lying dormant in the unsolicited feedback that customers are providing you every single time they interact. So we think about, you know, traditional voice of customer survey tools and things like that, And, I think we can all recognize that there's a bit of fatigue as far as, you know, directly soliciting input from your customers. We're not saying that you should stop doing that, but there is some fatigue out there from all of us as consumers. But think about that unsolicited feedback. When customers are calling, they either need something resolved, they might have a pain point, they might have a question. All of that is such rich data that really, really deserves being mined. And so that's really what conversation intelligence is all about. So let's go back to, you know, that difference between traditional speech analytics and speech analytics tools and what we mean by conversation intelligence. So what is speech analytics? You know? And how is it different? So we see speech analytics as a methodology that looks for very specific keywords, usually in call transcripts, but, you know, across all channels. Traditionally, it's in call transcripts. And looking for those keywords to categorize the main topics that people are calling about. So this approach tends to start and and somewhat be limited by what we already know is happening in the business. So let's say we know people are calling about payments or they're calling about an address change or all of those very typical reasons that customers may be calling into your contact center. So you may look for all of the words that you know would comprise the sentences around asking for those specific questions. You find them, create your topics, and then do some exploration as to maybe trends and volumes and that type of thing around the topics. So conversely, conversation intelligence is usually using machine learning techniques. That's a more unsupervised approach, a more AI driven approach, and therefore, it will uncover those expected topics and conversations, but it will also uncover some unexpected insights or maybe some more granular insights about your business. So, you know, I like what we have on the page here. It's worth reading it through about conversation intelligence. It really does empower your organization to use the data assets that you already have to identify and to solve key challenges before they turn into problems, before they turn into frustration, or employee churn, or operational efficiencies. Inefficiencies, I should say. So, let's let's talk a little bit more about those challenges that you may be struggling with in your in your contact center and why they're so important. First, customer frustration. So, you know, our customer service representatives, the people that are on the front line talking to our customers every day, they're really your brand advocates. Right? And so, you know, eighty percent of businesses are losing customers due to poor service. That service interaction is just so important. Therefore, there is so much to learn in mining those conversations, highlighting where the friction points are, and then determining what you can do as a business to resolve those. Again, rich, rich information sitting in your in your customer conversations. The next is employee churn. And we know there have been many, many studies that show that, you know, a certain level of tenure and a certain level of proficiency, contributes to a really great, customer experience. And, you know, the the happier and the more tenured and the more experienced your customer service reps are, the better job you're going to be able to do and the higher you're gonna be able to drive those scores like, you know, NPS and and CSAT. So when we analyze the conversations and we find out not only what's frustrating our customers, but what's difficult for our employees. And what do we need to fix either in the contact center or upstream in processes, to make their interactions with your end customers better? Lots of rich, rich data, sitting in the conversations that can help us, provide insights and ultimately resolve those types of issues. And then finally, all kinds of operational inefficiencies, you know, all of the key metrics that you're probably very used to tracking, whether that's call resolution, transfer rates. You know, we hear so many issues around the IVR and the IVA. You know, are there, very simple questions that could be resolved in more of a digital or self-service manner that are going to live agents? All of those things, all of that information, is is very, doable as far as, finding it, calling it out, and making those improvements when we're, when we're analyzing conversations. Okay. So how do we do it? How do we turn conversation data into the answers to all of these problems? And, you know, what what I have observed, I've worked with many, many clients who are, at various levels of maturity in this area. So if you're just beginning, don't feel bad. You are absolutely not alone. And there's there's a path forward in terms of that evolution from the data that you certainly have today to turning that data into the insights that are gonna make an impact for your business. So let's let's kind of go from left to right here and talk about that maturation, and how you can go from simply having the data all the way to, uncovering those transformational insights. So first, the data sources. And I touched a little bit about this. We have customer transactions, interactions. You may be recording the conversations, the voice calls, but not transcribing them. So the the transcription of those, conversations is definitely step number one. It's quite possible you may have the transcriptions and lots of them, you know, hundreds of thousands even, but they're sitting in a repository, and they're not being acted upon. So we go from, you know, call recordings and and chat information and all of that type of thing from your various channels over to the transcription. From the point of transcription, we then summarize. And then from the summarization, we move to the visualization and the calling of the insights. Going back to the data sources, so we have we have the conversations. That's what we say here as customer interactions from all channels. And then there's all of the other information that we can attach to those conversations around customer profile, customer behavior and history, agent information, what agent handled that call and how proficiently did they handle that call, and the channel metadata, like all of the call detail, and, you know, all of the various metrics around, the channel that the customer chose to interact with. So that's the data. And we know the data is there. We know you have it in your, in your organization. The next step is to do some of the basics that, is pretty common today in terms of utilizing call transcripts if you have them. And it typically is around, you know, the the QA area. So are my agents you know, and these are things that you have to do kind of from a compliance and regulatory perspective. Sometimes that involves only the manual listening of calls. Other times, there's a combination between the manual listening and maybe some speech analytics or maybe even more advanced techniques. But it's all around, are my agents, you know, going through the steps that they need to go to? Are they satisfying the regulatory requirements? And an assessment of overall kind of what are the agents' skills so that we can either understand, you know, how training needs to be changed or, you know, where there are particular issues that need to be addressed. That's kind of the foundational step number one. From there, we go into more advanced analytics that still tends to be centered around contact center. And most organizations, that I've worked with have a lot of untapped opportunity, in this particular area. And those are questions like, how do I get better at channel containment? How do I do more automation? You know, whether that's the use of conversational AI or other tools. You know, how do I automate some of those conversations? Which frankly, is not only, you know, better for the business from an efficiency perspective, but oftentimes much more preferred from the end customer. How do I deflect calls from my live agents, to more digital channels? How do I make what I'm learning about specific agents, you know, and where they may have strengths and weaknesses to be very, very targeted around the coaching, not just from a kind of, you know, one off manual QA, but overall, how can I cover my entire agent base and, do the best job at being targeted at helping them, be much better at their jobs? Where are there opportunities to make improvements in the knowledge base? You know, oftentimes there are gaps and there are areas that, you know, a large, representation of agents are struggling with something that that there's a gap. So it it it uncovers that. And then there's lots and lots of analysis around how we can do a better job at, you know, improving our customer experience measured by, let's say, NPS scores and CSAT scores. From there, there's even a next step in the, you know, in the benefits of doing this type of analysis, which goes way beyond the contact center, and starts to look at the contact center and these customer conversations as the megaphone for the rest of the business. If this is where the unsolicited feedback is sitting, How can we then, you know, package that information and communicate it out to the rest of the business to fix some even bigger problems? And that's really the, you know, the ultimate dream or goal of where this type of analysis can go. And that's around things like proactive detection of product defects before they become a problem. You know? Wouldn't the contact center be the first place when, when there's issues with products, even issues with services that you're offering, you know, and the customer, they're having problem with those. They're gonna call. They want their problem resolved. So how quickly can you take all of those clues that are coming into the contact center and push that information out to the rest of the business? So product defect, you know, innovation in terms of better ways to use channels that customers may be preferring? Are there issues in your retail experience that people are contacting, customer service about? Where can I do a better job with communication? You know, the kind of the list goes on and on. It will vary by business, but there's definitely techniques to pull out all of that information, you know, and really act upon it. Okay. So how do we do this? In my experience, I really believe there are four kind of pieces of the puzzle that all have to come together, and it's pretty simplistic. First, I talked about the data. There's the foundational data that is your customer conversations, capture them, transcribe them, use them in a way that can be curated for analysis, and then integrate that with other data about your business. Next, you really need the right people. And like many other things, there tends to be in this area a gap between the promise of what a software tool can do and the actuality of being able to really leverage that software tool to extract the right value for your business. So what I mean by the people is, typically, with any of this, whether it's a speech analytics tool or the full conversation intelligence that I've been talking about, you really do need skilled analysts, and that either can be outsourced or insourced. You know, what we typically find is companies that are doing this very well, they either have they have invested in a staff, not a person or a few people, but they have a staff of people who know how to work with this data, who know how to interpret and pivot the data, and present it in a way that's meaningful to the business. It takes skill. We certainly do this at TTEC. There are, you know, there are plenty of, you know, sort of outsourced experts who also know how to work with this data and use the tools and derive the most value. But, my humble opinion is there's not an easy button where there's a canned set of reports and, you know, you hit the button and you're off and running and you have all of the magic answers that are gonna help solve the problems of your business. Hand in hand with the people is the visualization. So where I've seen companies, get stuck a lot is they've got incredibly rich data. And when they look at the data kind of on a record by record basis, it's very impressive, all of the information they have. But now when you're talking about hundreds of thousands or millions of customers and all of the different permutations and conversations that are being had, issues that they have, friction points, all of that together. It can become very overwhelming. And how do you summarize this in a way that's easy to consume, especially for kind of the non analyst or executive or non, you know, CX or or customer service or contact center type of professional, serve it up in a way that, is meaningful and easy to understand. So great visualization is a piece that I find is often overlooked, but very, very important in terms of really getting some traction in this area. Okay. So what is our, best practice approach? So this is what we have in place. We do this for, many, many customers, and we find that this is this is kind of the best way to move from what I was talking about, which is the raw data all the way to the actual insights. So it starts with your data, certainly the call transcripts. I've mentioned the other types of data that you can integrate together. That data then comes together, and we run, custom algorithms on that dataset. These custom algorithms do the following in that you see these kind of in those, purple boxes there. First and foremost, we run topic modeling. It is machine learning driven. It is the unsupervised approach that I mentioned. It gets extremely granular. And what we find is back to the visualization and the summarization, there's a very logical way to present very granular information so that is it is consumable. And what we do is we kind of have a three layer hierarchy. So we have your main categories of topics. Those are typically the ten to fifteen things that most people who are, you know, close to customers and working in the contact center, they they know. They could kind of rattle off. These are the main reasons people call. They may not know the volume distributions. They may not know the complexity of the calls or the sentiment of the calls, but they generally know what those calls are. So or those categories are. So that's the category level. From there, we go down to topic within category. And now you're getting a little bit more granular. You know, if a if a category is make a payment, there might now be questions about, online payment, didn't didn't receive my payment, you know, etcetera. You can think about all the sort of subcategories under that. And then from there, we even go one level deep deeper to subtopics. And that's really what we find is when we get to the subtopic level, first of all, these tend tend to be very kind of, you know, revealing interesting information, that most people in the business are not quite as aware of, but they also tend to be much more actionable when you get down to that level of detail. So, that in a nutshell is how we approach topic modeling. We've got, you know, kind of three different levels of information, and it's, it's machine learning driven. Okay. From there, I wanna talk about sentiment and complexity, which are very important components. Now sentiment is is interesting. Sentiment scoring is pretty commoditized these days. There are a lot of different tools that produce sentiment scoring. There are very, very different ways to calculate sentiment. It can be an overall average throughout a conversation. You can look at the sentiment at the beginning of a conversation versus the end of a conversation. You can look at sentiment, you know, getting, stronger or declining. You know, all of those things are important. We have we generally start with a sentiment score, and we are happy to work with if you have, something in your business that is already being provided from another software tool, we are happy to read that in and at least evaluate, if it's doing a good job. But sentiment's important. Right? Because it's one thing to know why people are calling, but it's even more insightful to understand which topics are associated with customer pain points and friction or which ones are actually customer satisfiers. All of that's super important to understand. Next, we have complexity. Unlike sentiment, I don't see complexity scores out there as much in the marketplace. We've been doing complexity scoring for many, many years. We did some early complexity scoring with MIT actually, about a decade ago. We've continued to improve it, but we find it most important, I guess, for for two applications. The first is where are there the best opportunities to automate, you know, use AI, and have, those nonhuman interactions? So, for example, if we have if we're observing conversations, particularly within a certain topic that are high sentiment and low complexity, that's where you should get started in terms of automation. And then finally, emerging topics. This is really important and a differentiator from sort of a static speech analytics, which is looking for new things that customers are talking about. That becomes very important either, you know, is was there a change in the business that is a dissatisfier? Is there something happening with a product that needs to be addressed? Was there a process change that people are now calling about? So something on the website that is a glitch, you know, etcetera, etcetera. So looking for those new and emerging topics instead of just having a static set of topics that you're always evaluating, is a very important part of overall conversation intelligence. So we run those algorithms, and then we, we use those to as the foundation to develop our, insights dashboards. And in a bit, I'm actually gonna do a very quick demo of, the dashboards that we typically use in this area. But, you know, I have some features down here on the page that I do wanna call out. One is that where possible and, you know, we we're definitely, cognizant of, you know, info security concerns and etcetera, etcetera about using l LLMs and GenAI, where there is, you know, sort of a corporate acceptance of utilizing those tools. We tend to use GenAI for call summarization and also for the topic labeling. So depending on, you know, if you have experience in this area and there are tools that you've used in the past, sometimes the topic labels are not very intuitive because they're more sort of a a funny combination of the different words and you read the words and you don't know exactly what it means. So we actually use GenAI to have very, user friendly and and understanding understandable labels to our topics. We also use AI to serve up dynamic insights. And what that essentially means is based on the data that's on the page, of the dashboard, we will have typically have a narrative at the top of the dashboard that describes the highlights of what the data is showing. We also have a lot of drill downs and search functions and filters. And the last thing that I will say that I think is important about the dashboards is a level of customization. Every business is different. Every business has has, you know, their own way of looking at metrics and their various nuances. So we take an approach of, you know, let's take, industry best practices. So, you know, that's another just important call out is that industry by industry is quite different. So having that understanding of, you know, typically what's done in an industry and where there are best practices, we apply those and we apply our frameworks, but then we always have a level of customization, to a specific business. Okay. So, that's that's our approach. And kind of looking at my notes, there is there is one point that I didn't make that I do wanna highlight, which is, when we're deploying this, for our various customers, we're hearing a lot about frustrations with the natural language IVA. And the fact that, you know, perhaps the IVA was deployed and it hasn't been tuned for a while. It's not picking up the customer utterances in terms of what they're saying they're calling about and basically, like, routing it to the right place the first time. There's a recognition that, there's a lot of upside potential in improving the natural language of, improving the tuning of the IVA. And so we we have been running this conversation intelligence on both the utterances from the initial you know, from the customer, why are you calling? How can we help you? That short set of utterances. And comparing those utterances to the full conversation, and using that comparison to help us tune the IVA or even the IVR. Okay. So here are some I'm moving now into some kind of cases and examples of how we tend to visualize. And these are little snips from our dashboard, and like I promised, I'm gonna show you the dashboard in just a minute. But I wanna kind of bring some of this to life and show you how we use this. So this was actually, for an automotive manufacturer. And here, you know, we're looking at I I explained to you, we've got sort of that category, topic, and then subtopic. So you can see the category here was questions about the app. The subcategory sorry. The topic within the category was activation, setup, and reset. And then what we're looking for here is opportunities for improvement. So let me just walk you through, you know, in here what we're doing is we're looking at volume underneath call handles calls handled. And then we're looking at a distribution of complexity and distribution of sentiment. And then what are we actually finding from this data? So one of the big questions was, okay. We know that people are calling into customer service all the time about the app, and there are all kinds of problems with the app. And what we don't really know is how much of that do we have to take back to the app team to make fixes to the app versus how much of it is customer education. And this is an interesting one because it's an automotive company, So they have an app, and, obviously, you know, they want their customers to be able to utilize the app to have a better ownership and driving experience. Right? And so if you look at some of the ones we highlighted, you know, let's look at the second one down here, trouble connecting the app to the vehicle. And when we looked at this, it was one of the lowest, you know, percentages of high complexity. So this was a very easy one to explain when customers called in as opposed to activating the app, which actually had one of the highest levels of complexity. But let's just focus on on the second one here, connecting the app to the vehicle. So very easy to resolve in terms of the complexity, and the call wound up being extremely high sentiment. So I'm envisioning a call that says, oh, you just have to do this. Customer says, thank you so much. I didn't know. Voila. It's fixed, and and a happy customer gets off the phone. So this was you know, one of the things that they talked about was, well, shouldn't we go back to the dealers and tell every dealer when they take delivery of a vehicle, make sure you don't let them leave the lot without showing them how to connect the app to their vehicle. And what is that going to do? It's gonna take, volume out of the contact center, and it's gonna, it's gonna create much happier customers. So what we did was we worked with the business, and they did some very initial and extremely conservative, financial estimates of if we made a few of these fixes, what would the near term in year cost savings be? And that's, you know, some of the numbers that you see there. The next idea is about identifying product defects, which we talked about. This is a simple table that kind of looks at you've got product in the columns. You've got all of those topic categories here. And the point here is, you know, those quality issues vary, very, very much by the type of product. So, you know, we're we're obviously, you know, sort of genericizing the information here to, protect confidentiality, and we we did kind of change numbers and such. But just know that, you know so for a particular product or in this case vehicle, all of the calls, essentially, about, you know, multiple visits to the dealership were coming from one very specific, vehicle. All of the calls around brake issues were coming from one specific vehicle. So this is really where listening to the customer, understanding the product issues, bringing it back, you know, to engineering, etcetera, you know, is very, very valuable information that's coming out of the contact center. And, again, we had some, you know, financial impacts of that. And then finally, you know, I do wanna share, this is a very important framework that, that we tend to use where I talked about the complexity and the sentiment score. So what you're seeing here is, in the bubbles are the call call reason categories, and then the size of the bubble represents the volume. And then we array that by the, complexity associated with that topic as well as the sentiment associated with that topic. So, you know, when we think about call deflection opportunities, we tend we want to focus here on low complexity, high sentiment calls, and really work on things like, you know, chatbots and self-service, to address these very straightforward questions so that they don't have to go to the live agents. Like, case follow-up here, across industries, you know, case follow-up, you know, whether that is automotive, insurance, etcetera, etcetera. Very there are very, very simple ways to get to give status updates that, you know, really shouldn't shouldn't require, you know, the time or expense of a live agent. As opposed to in the automotive industry, you've got vehicle recall. Well, that's a very serious issue. It's high complexity. It's low sentiment. People are frustrated. People are scared. I'm driving this vehicle with a recall. What am I gonna do with it? So, you know, that's where a skilled agent is very, very important. And we can do all kinds of drill downs around this topic, you know, to find more more information about that. Okay. So, you know, just to kind of summarize even from this one case, oftentimes, you know, the the logical question here is, well, how do I know spending time on this and certainly investing the money is giving us the ROI? And what we will often do with customers is we will do a quick proof of concept, and, you know, I'm talking this is doable in about six weeks to say, let's run the data. Let's look at some of these initial findings like the ones that that I shared with you, and let's pull together what the quick hits are and what the conservative financial estimate is so you can determine if, if it's worth the investment. And clearly here, you know, we identified over a million dollars worth of opportunity that was very low hanging fruit and that that, could be realized within the first year. Okay. And now I will give you a very brief overview, kind of a live demo of what our conversation intelligence visualizations and dashboards looks like. This is the opening page. And for consistency, we are actually showing the same data that I, shared with you in the presentation. And I'll kind of walk you through our opening page so you can see how we present the data, and give you a feel for, you know, the ease of usability as well as some of the areas where there's quite a bit of flexibility. So first, on the upper left, this is our, automated and dynamic insights. So what we have here is in narrative format, a summary of many of the data points, the most relevant data points that you would see on the page. And if you were to choose to change some of the filters, like perhaps the date range or your product mix, anything else that you wanted to customize as far as, filter capability. Once those are changed, this, narrative will automatically update. And you can see here it captures things like, what was the most frequent topic of calls in the date range that you're looking at, which call reasons had the lowest sentiment or had the highest complexity or even some specifics about particular products. And then we have some trending. Now in this particular page, we only had a week's worth of data. So, the trending, looks looks pretty flat. But we do think that, following the trends is quite important on some of these key metrics like volume, complexity, and sentiment. And then finally, this is the depiction of the reasons in the categorization that I had described. Meaning, first, we have the top categories, and then we can do, drill downs from there. So within the categories, we have the topics. And so under the category of app, you can see that the first topic under that is activation setup and reset. And, you know, there are other topics that then go into the subtopics and etcetera. And then what we like to visualize is, kind of top to bottom, what is the volume, where you see higher and lower complexity calls, and where you see higher and lower, sentiment calls. And when I say calls, I really mean those call reasons. The next main dashboard that we have is the complexity and sentiment. This is the framework that I had described where we're looking at the categories and we're depicting them, across both the complexity and the sentiment, and the size of the bubbles there represent the volume of each of those categories. Now sometimes you may find that these categories are too broad and therefore not all that actionable. Where where there's a category that would be actionable is, for example, all the way in that upper left, you see that blue bubble that says case follow-up. Case follow-up or case status, that's something we hear across most industries where that should be a pretty straightforward way to, you know, provide some self-service or, you know, IVR contained information. But there may be some others like this, purple one here that I'm hovering over says vehicle hardware. That's really just too broad of a category. So in that case, you can do the drill through, and maybe we wanna show vehicle hardware by topic. And so then it will now array those various topics within vehicle hardware, and you can see some, you know, very logical things like related to an oil change. This is a, you know, very easy to address question as opposed to some more complex questions like, about the transmission or about the engine. And then lastly, just gonna show you one more page, which is the keyword query tool. So I had talked a lot at the beginning about the difference between, speech analytics and conversation intelligence. This kind of brings back some of the things that people like about speech analytics in that sometimes in your business, you're gonna be looking for something very specific that you want more information on. And our query tool, allows that functionality. In fact, we have, further developed this query tool to be more, you know, natural language driven. Right now, the one that you're the version that you're looking at here is is very much related to the specific word. The point is we've, we've enhanced our search capability within this tool. But even here, you know, one of the things, that in for this automotive client that they were asking about was, they know that certain trucks don't have folding side mirrors, and they were wondering if that is becoming, an issue that people are calling and asking about. So we had typed mirror into the text search, and you can see the number of records that came back, the vehicles that were associated with that word, and then all of the very specific information for every single one of those sixty nine calls that hit that, that keyword. And we could even drill that down further. Maybe we wanna say side mirror and see what happens there. And there were, in fact, three calls that had the keyword side mirror. We can see which vehicles were associated with that. It it turned out in the real dataset that none of them were associated with the truck in question. And then, you know, it's really interesting to look at all of the detail. If you're if you're really exploring a specific topic, all of the detail call by call, and this is where that LLM driven call summary comes in. You can see here, you know, basically, within two to five sentences is a really good recap of what was happening in the conversation. We've done a lot of work comparing the actual, you know, hundred to two hundred line transcripts to the, the concise, LLM driven summary, and they're very, very accurate and, very, very insightful. So here, you just see customer call to inquire about blind spot information system. Agent informed that that the vehicle did not come equipped with that feature. That was the gist of that call. So that's it that I'm gonna show you for the demo. If you feel you are, you know, ready to get started and interested in any more information, information specific to your business, I would really, really love to have a conversation. So, you know, think about it. Your customers, they're creating all of these insights every single day, every single time they interact with your brand. And are you really hearing what they have to say? We you should be. Right? And let's start with delivering those frictionless customer experiences that drive loyalty and contact center efficiency and, and great, great experiences. So, that is the end of my talk, And I think we did, have a few questions that came in from the chat. So let me just, see if I can get to those. One moment, please. Okay. So, the first question I'm seeing is, the question is as follows. What does the next iteration of conversation intelligence look like? Okay. Well, we have already been thinking about that, and, I definitely have some thoughts. The first is more use of GenAI. Probably, not a surprise there. So specifically with GenAI, semantic search is becoming very hot. And what I mean by that is the ability to ask a question, you know, in natural language and get answers that are more conversational. So, you know, examples like, this week, what which topic, increased in volume and which topic decreased in volume? Or, what are the what are the conversations that have frustrated customers the most over the last sixty days? You know, things like that where you can literally ask a business question and get the answer. Now we're finding there are already, tools that are out there and doing that. They're certainly not perfect, but I do believe that, you know, the next iterations, you know, will will just keep getting better and better just like GenAI keeps getting better and better. So I would say definitely be on the lookout for that. Also, what I'm finding is there are a lot of attempts to get to not only sentiment, but to kind of the range of emotion and, you know, scoring conversations that are much more diverse in terms of, in terms of, tone and emotion and things like that. I was just in a conversation literally yesterday, with a a client of ours that sells, financial products. And what what he was explaining to me is, you know, it really makes a difference when our end customers talk to our financial advisors, and our financial advisors, communicate the information with a certain level of confidence. And when they do that and they're confident in what they're saying, the customer then, you know, has a certain level of trust, and, you know, and the confidence back that the information is is accurate. And so whether it's confidence or any other type of tone of voice, those are things that are quite interesting to monitor, for the agents, and bring that back, you know, for those targeted coaching opportunities that I talked about earlier. So that's what I'm seeing as the as the two, advances in the area. Okay. I think there might have been another question. Alright. So our next question is, what are some critical steps for evaluating the partner and the technology needed to implement conversation intelligence? Okay. Definitely have some, some experience and opinions there. So okay. First of all, if you remember the stair step slide that I had presented, I'm sure, you know, anyone on this call in the business that you're in, you have a certain tech ecosystem, that the business is, invested in. And that may be, you know, your your cloud provider, like your hyperscaler, could be your your CCaaS system, your CRM system. You may have invested already in a speech analytics tool. So first and foremost, you know, understand the current landscape and ask the question, are you getting everything out of your current investments? And so that that's really, like that's the first place to start is just do an evaluation of your current tech ecosystem and start by using what you have today. Now you will know where you are on that maturity curve, and that is going to dictate kind of where you go next. You may just be starting with they need a transcription service. And my opinion on the transcription services is I've not seen any particular transcription services service that is more accurate than another. So, again, you know, look at what works the best with your current ecosystem. And if you're not transcribing, absolutely should be investing in that ASAP. But then think about that, kind of jigsaw puzzle slide that I showed, and one of the most important components is the people. So you may have the technology, but you may not have the people who know how to leverage that technology. And you absolutely need that skill set, whether it's in house or outsourced, to know how to work with the tools and know how to extract the meaningful insights from the tools. So the people piece is really important too. Another is budget. And, you know, there is ROI in these tools. I'm positive of it. I've, you know, been involved in those, you know, business cases many, many times, but you're gonna have a certain budget and look at, kind of what the budget dictates. What also goes into that is is yours the type of business that really need, you know, your sort of, like, high volume, high frequency? You need more of a real time type of update to really make those business improvements, or is it okay to do this type of analysis maybe once a quarter? For some businesses, once a quarter, especially if it's relatively static, is fine. So that's something else to think is, you know, what's my what's my need for frequency of update? So, you know, we can certainly guide you. We are totally tech agnostic at TTEC. We've worked with all technologies, in this space. So we can really help sort of, you know, guide you through what what may seem actually like a fairly crowded, and somewhat fragmented marketplace here. We can we can provide opinion. We can, you know, we can really help, get the most value of all of it in a way that's streamlined, in a way that helps you get the most of your current tech investment and help you get the ROI. So, with that, I am seeing that we are almost out of time. And, you know, I I think that we've gotten through, all of the material. So thank you, thank you, thank you, for for listening and for joining us. I hope to have, an update on this topic where we can talk about it again. So thank you very much.