Microsoft Corporation (MSFT) Earnings Call Transcript & Summary

Hi, I'm Michelle Huenink, Director of Customer Experience and Analytics at Microsoft. And I'm here today with Melinda Ritchie, Lead Program Manager from my team. And today, we're going to be discussing how we harness the power of predictive analytics to deliver next-generation customer experience. Thank you for being here with us today. Before we begin, I want you to note that this session has been prerecorded, so we'll both be able to answer live Q&A during the session with the widget on your screen. So let's go ahead and get started. To help you anchor on what my team does, our team sits within customer support, and we're focused on the customer experience. So essentially, we look across the entire support landscape to illuminate opportunities to drive end-to-end customer experience journey improvements across the board. And we do this by infusing customer experience industry trends into our strategy as well as through tools, data and advanced analytics to improve the holistic customer journey. So today, customers demand more from Microsoft and their service experience. When we think broadly about how customer expectations are evolving, customers no longer compare Microsoft to their direct competitors. Instead, they're looking across all the brands that they consume regardless of product or industry. With every support interaction, the best of what a customer encounters, whether it be online, in-person or via self-service support, that becomes their next expectation. So consumers today are mobile and self-reliant, their preferences are becoming increasingly digital, and they seek more ways to interact with us. Much of the technology customers use every day is exceptionally well designed, and they expect nothing less from their support experience. They want it to be simple, intuitive and effective, and they want to be able to move seamlessly between technology, self-help and human interactions, reentering the support loop at any given time without missing a beat. They want us to know who they are, what products and services they're using and have a view of all their open issues across the business and what questions they've already asked or answered. They really want us to know them. In a cloud-based world, customers' expectations are not only getting increasingly higher, they're continuing to evolve. So customers expect a support experience that's immediate, effective and seamless. And we're going to cover how we leverage these techniques to move Microsoft support and engagement from a reactive model to proactive. They want us to anticipate their needs to make it personal, so we're going to talk about how we've done that because support truly has become a competitive advantage. And by putting the customer at the center of everything we do, aligning our organization for success and engaging our employees, we can unlock the potential of support. In order to transform the customer experience and really truly create best-in-class experiences for our customers, we have to leverage all of our customer listening systems. And we do this through a closed feedback loop. We're constantly leveraging insights and learnings from all these listening systems to drive improvements and enhancements to both our products and our support experiences. And we take those learnings and feedback to the business to drive change, which is key. So why predictive analytics? The business wanted us to get away from being reactive and really move to a more proactive approach. And we know this is a key customer experience industry strategy. As we look to transform escalated support from reactive to proactive, we leaned into the predictive analytics and machine learning space. And we're not alone, becoming more proactive through predictive analytics is a game changer for customer experience. Companies across all industries are focusing their efforts here. I'm going to hand it over to Melinda and let her talk about we've done -- what we've done in this space at Microsoft. So go ahead and take it away, Melinda.

Excellent. Thank you. All right. So I'm going to talk a little bit about what is predictive analytics. So basically, it's another tool in your business toolbox, one which gives you educated predictions on what to expect next. And this tool, through recent advancements and democratization of the technology, has become increasingly used -- useful and irreplaceable as a strategic lever and source of business insight. This is because predictive analytics can give you a view into the valuable information that already exists and you already have. You just need to root around on it and dig it up. Aside from being a tool to leverage, there's also well-established benefits of employing predictive analytics and machine learning across all areas of business operations. So just to kind of give some ideas and get the juices flowing about how this can be used by different industries. Here are some quick examples of how and why predictive analytics can be so transformative. So if you are producing items, you can improve efficiency in production. Companies can effectively forecast for inventory and required production rates while also using past data to estimate potential production failures. All of that's possible through this analysis. They can then use this to prevent those same errors from occurring the next time. You can gain advantage of our competitors. By tapping into customer data you have available already, it can present you with insightful information as to why customers choose you over your competitors. You can highlight unique selling points that you can then further promote to enhance leads. It reduces risk. So sectors such as finance and insurance use predictive analytics to help construct a valid depiction of a person or business they're screening, and it's based on all the available data they have to them. They can then form a more reliable interpretation of that person, the business and incidents, and all of that can be used to make sensible, effective decisions. Other examples are detection of fraud, better marketing campaigns, better ability to meet consumer expectations. All of that can be unlocked by being able to see around corners and predict customer needs or behavior. So going one step deeper now beyond how predictive analytics is used, let's talk more about what it actually is. Predictive analytics uses technologies, you've probably heard of like artificial intelligence and machine learning. It runs on tons and tons of historical and real-time data, and it helps us understand things like how customers will behave; what they think, in our case, about Microsoft; how they use our products and lots of other valuable information. Knowing this type of crystal-ball information means we, as a business, can plan and respond accordingly. It's great because it means we can use the information we already have to improve the experience of the next customer. And that's because we know how the previous customers thought, felt and behaved. It's like having a guidebook to reference. Because we have layers of information that can be analyzed for opportunities, gaps, answers, et cetera, we can approach crafting ideas and solutions differently, aided by those insights and personalized to our customers and other stakeholder needs. One important facet of predictive analytics is also to highlight the role it plays in informed decision-making. So it's a quantitative and sometimes qualitative way to embed data-driven strategy into the business. This is increasingly important due to the prevalence and availability of data currently. And with it still growing at such a rapid rate, it means that the need for efficient processes that can turn all of that data into points of insight and value are increasingly important. And predictive analytics answers that opportunity. It's how you can turn all of that data efficiently into business value. So now specific technologies we turn to in order to perform this efficient sorting and analysis of data are statistical algorithms and machine learning techniques. They're used to identify the likelihood of future outcomes based on historical data. The goal is to go beyond knowing what has happened in the past to be able to instead provide a best assessment of what will happen in the future. In data science, there's 4 types of analyses. Most business intelligence is currently in descriptive or diagnostics, telling you what has happened, what has been -- looking in the past and visualizing. When we get into predictive, we look at the idea of this past behavior, predicting future outcomes. So really, we're moving further down that value-add analytics line into the next level of, how can we see around corners, how does this data tell us what's next. And the next phase coming in predictive analytics is prescriptive, which is not just what will happen, but the complete analysis of why. And that's coming up soon in the industry. So on that notes, it's really about what you can learn from the data you have by analyzing it a new way. By employing different model types and tackling novel technology, new gaps and opportunities are constantly uncovered. This means -- or this helps the business evolve at a faster rate and through much more informed planning. This value and what can be scientifically predicted for you gets increasingly powerful, too, as you include more and more types of data. So for example, an algorithm predicting the quality of the wine produced from this year's vineyard harvest would likely look at things like the aspects of the grapes themselves, like the volume of the harvest or the size of the grape. But to get a really accurate model and how to unleash the true predictive power of analytics, you would also want to include seemingly unusual different types of data but related around things like how many days of sun were there that year? How many inches of rain? What has the historical quality been like? What type of canopy was used? What harvest decisions were made? All of that combined together can help make a truly accurate and powerful prediction. So now we're going to turn to how our team at Microsoft is applying these predictive analytics strategies and leveraging data in these ways to improve our customer experience. So for some background, as Michelle mentioned, we operate in a very narrow but important band of customer support cases, those that have dropped off the design support path and are in need of recovery. We broadly refer to these as recovery escalations. For our area of the business, we started up to make the following 3 goals around these recovery escalations: we want to predict when they're going to happen, we want to classify the issues we see in them, and we want to see if we can pull out the customer sentiment and the issues that the customer experienced. Escalation prediction would use past customer data around behavior and case information and predicts if certain types of support cases might be a need of intervention. Issue classification would look to automatically sort and group our case data into the type of problem the customer called about using text and telemetry analysis. And lastly, sentiment and breakpoint extraction would use past issue data and text analysis to predict the customer's concern and their temperature. Hopefully, you'll notice some important similarities in these projects. For each, we've taken historical data and used what we learned about past cases and customers to tell us how to react to current cases and customers in an automated, data-driven way. And we're also using historical data analysis to help augment or automate processes that are either inefficient or are recurring tasks. And that's in an effort to just help our people succeed and be happier. The ultimate goal of all these projects is to improve the holistic customer journey to really anticipate customer needs by proactively identifying trends, gaps, opportunities, outliers, all from customer support tool data. With these predictive models, we're constantly mining for actual insights so the business can make data-driven decisions and drive change. Now zooming out to the 10,000-foot level, just to emphasize that all of this is to help the business become more proactive and accelerate improvement, but from an informed place. We use predictive analytics to illuminate opportunities and feed them forward, so decision-makers can make smart, strategic choices based on specific learnings from the data analysis. So as I mentioned, these were our predicted -- or planned predictive models that we set out to make, but there is a whole journey and a whole lot of learnings between starting out with our ideas and actually getting to a place where we had an operationalized model that was producing value and where we ultimately ended up. So I'm going to go into that journey, and hopefully, it will help illuminate some key tips. So when we started, after planning our predictive goals, we found that the reality of enacting the theoretical models into the operational processes of the business came with a lot more strategic infrastructure coordination and planning than we had realized. Immediately, we were faced with some pretty major, but not wholly uncommon hurdles, things that most businesses face and would have to tackle in order to deploy predictive models on an ongoing basis. So we looked at all of our different data sources, and they were -- we had a myriad -- so much data, but they were all written in different types of code, and they couldn't be combined because they weren't designed to be used together. So we had to overcome that barrier. The data we captured wasn't answering the questions being asked, so it really wasn't that useful in terms of providing the insights we wanted to provide. And then as probably everyone's been hearing all over the news recently for the past 1.5 years or 2, there is an ever-changing landscape around global data and privacy regulations, and this greatly affects the analytical space because the type of data that we can retain, how long we can retain it and how we can use it is constantly changing and becoming much more hard to navigate. So now, the through-line from all of these hurdles really came from the reality that we were coming into an already designed and established data landscape. So what I mean by that is that none of the systems were designed with the end in mind, meaning they weren't created to facilitate machine learning algorithms, how the data was being captured or stored. Additionally, the new world of big data meant a whole new level of coordination between data types, teams and tools and how we classify and collect data. At the outset, it was like trying to combine separate chapters of a cohesive storyline for a book, but all of the chapters are written in a different language and one chapter's on paper and one chapter is digital and one is spoken word. Somehow, they could all come together and tell you a story, but the delivery and the means of each piece were wildly different. Additionally, a lot of the data was being collected on background business policy and telemetry. And as I mentioned, we weren't collecting information that could directly speak to the questions the business was asking of us and the things we really wanted to be able to predict. In order to predict something, you have to have that event recorded in a clear way and have component data that can relate to that event in meaningful ways. So for example, to predict an escalation, a customer support case that's gotten worse and needs recovery, data in general on how many Windows subscriptions we sold isn't particularly helpful. And finally, an important and constantly evolving -- finally and importantly, the constantly evolving legal landscape meant we needed to do work in an ongoing deference to these laws and security regulations which is just an enormous feat, even just saying up-to-date with what the latest is. So now, I'm going to go into how we address these barriers. Since we found that the issues we faced are pretty universal in establishing robust predictive programs in the business, we also want to present these learnings and how we addressed our barriers as a sort of best practice set. It's likely that you would face one or all of these in operationalizing your predictive analysis mechanisms. We'll go into a couple of these in depth, but it's important to note that all 4 are as each is equivalent to the leg of a table in setting up your predictive foundations. So you need all 4 for the table to stand. Even if not a single one of them is a barrier, you really need to work in deference to the set. These 4 best practice areas are also the actions we took to get up and running in a robust way and how we conquered our hurdles. So first, we decided we needed to kind of redo the way that we collect data. We needed to create data or update our taxonomy so that we could capture the information required in order to best answer the questions the business was asking. So this process was long, and it was somewhat trial and error and iterations and learning through the analysis of the model output as compared to the insight that was needed. We also established strong and ongoing partnerships with important departments in our business. The legal team, and the BI team, our business intelligence team, are particularly important. This facilitated our ability to keep current with legal requirements and privacy regulations and that we had permissions and access to multiple layers and types of data that we could use in our predictions. We also notified leadership has a need to plan ahead and begin keeping the value and needs of machine learning and predictive statistical models in mind when planning strategy and choices for the future. This is to invest in our ability to continue and grow and evolve in these capabilities. Having designed future processes in mind with this infrastructure would unlock so many more insights and opportunities. Lastly, we found the importance of the human delivery and accountability piece when creating actual impact with our predictions. We found that with all the amazing modeling in the world, it would make no impact if we couldn't deliver and follow through on that information with the business and the decision-makers. We, therefore, developed a means to establish accountability through a closed feedback loop to make sure that what we surfaced was ingested by those who could use that information and use it strategically. So I want to dig into 2 of these in a little bit more detail. The first one I want to go into deeper is this idea of designing your taxonomy and making sure that the data you're collecting speaks to the end in mind. You're probably -- or you've probably heard of this phrase, "garbage in, garbage out" when talking about analytics, and this is talking about the need for related and clean data in making predictions. If half your rows are blank or the values are wrong or they were entered sloppily, your prediction will greatly suffer. And if what you collect isn't relevant to what you want to predict, the prediction won't be accurate. So these are 4 things we did to ensure that the data we were using to make our predictions was relevant, clean and usable. I mentioned before we started with the end in mind. We redesigned the taxonomy and included new details to make sure that we had the raw materials we needed to work with to make the predictions around the insights we were being asked to provide. We also embedded a really strong ongoing data-auditing process. This is super important because in that garbage-in-garbage-out standpoint, if the data that you're bringing in is incorrect or incomplete, it's very, very difficult to get any kind of a reliable prediction. So the first, #1 thing you always need to do is make sure that what you're feeding into your models is strong, relevant and right. We also entered into partnerships with the delivery team and asked them to help enter data when able. That means that we're able to grow our training sets exponentially, and we didn't have to be limited to what was already labeled. And then we updated our data logic, just meaning we modernized it. So we looked at what pieces of data we had and what we didn't need, what could be used to make predictions and what was being collected just because it always had been. So now, the second piece that I want to talk about that merits a little bit more detail is around how we can get the business to ingest the insights our efforts produced. We found that there's a need to translate the output for the business and key decision makers. This engagement piece became incredibly important, and that's just so the insights we derived were carried forward into tangible and measurable impact. To ensure this, we developed a closed feedback loop process with areas all over the business. And these are some of the ways we made sure that, that accountability mechanism was successful. So engagement and interpretation. So you need to make sure after you have your model and it's giving you output that, that doesn't go into a silo or a black hole. You need to go out and make sure the business knows that you're producing it and provide an interpretation or piece or a translator. A lot of times, it's overwhelming for someone in the business side to be handed a dashboard or a set of raw data and said, "here's your information." There's a really large piece that comes from taking that and then being able to translate that into impact and explain how and why what's produced here can flow into value-add for the business, current processes and current customers. We also created an ongoing cadence. So this was really about this accountability piece. We needed to make sure that we revisited. We didn't want it to be that we only showed up when we found something new. We wanted to reinforce the things we already found and then measure and revisit them to see how they change over time based on the actions that the decision-makers took based on our insights. And then last is just systematizing all of this. So making sure that after you get these reliable outputs, you've engaged with the business, you've given that translation piece that, that gets used in everyday processes. So it's not just an added interesting piece of information, it's an integral and invaluable part of how business gets done every day. So now, just a little bit of tips and to summarize. These are the 4 areas we learned were equally necessary for success and the best practices we used in building our specific models. These are the foundational and infrastructure needs you need to make sure and cover in parallel to model development, so your predictions can be operationalized and bundled up to the business in a systematic way. So data quality. And we talked about this a little bit already, but really concentrating on making sure that the majority of your time before you even build a model is making sure that what you've captured is 100% accurate and 100% complete, and that it's focused on the question you want to answer. Second, embedding infrastructure. So these are long-term investments around making sure that, that ecosystem of clean data, the ecosystem of creating models and embedding insights, the accountability piece with the business, all of that is created in an ongoing way and there's infrastructure to back that up and reinforce its value. That goes hand-in-hand with accountability. This is the human piece to the infrastructure and making sure -- around making sure that someone is driving that and that there's a constant momentum and constant attention to the process. And all of that is really fed by partnerships. Having this information and keeping it in a silo is absolutely the worst thing that could happen because all of these insights and all of the value really doesn't mean anything if it doesn't get out to the business to take action on it. So now, the important part. So this is a summary of where we ended up after all of that journey, all of those learnings, all of our preparation, exploration and stress. These are the predictive projects we now have in flight. So through these projects, we have been able to have significant impact on our customer recovery escalations area of the business. We've been able to improve both the holistic customer experience as well as the process of handling a case for our internal support partners. We arrived at these projects through a combination of acquiring and creating data sets from what tools and data repositories we already had available as well as what the business was identifying as specific gaps that needed to be addressed. So these represent the best means to leverage our current data sets and to strive to add value back to the business by alleviating bottlenecks, by predicting outcomes, quickly identifying trends or issues and other proactive mechanisms for infusing predictive processes into the business. So looking at the projects now, you can see some of the things that I've already mentioned around escalation prediction and being able to see around that corner for which case is showing a tendency of unwellness. So what we do is, for this is we look at historical data and we pick out the cases that we know went off the rails or left the design support path and what we would call recovery escalation. We use that to train a machine learning model to help us look for other cases that show similar signs of unhealth. Queue monitoring -- queue health monitoring and alerts. This is looking at thresholds and triggers around data that were coming in, in the moment compared with data from the past to see, are we seeing big changes that are unusual or spikes and trends that are unusual. Really, is everything operating as it should be normally or is there something inside all of this data that can flag us, like the canary in the coal mine that may be a big emerging issue is coming or another resource deployment is needed. So again, it's around anticipating what's next but for the business. Active search, operational trigger prediction, customer sentiment scoring and proactive solutions suggestions, all of these, too, are around really trying to anticipate. So for active search, it's once-a-month start searching, can we start to pull up relevant articles and predict the next thing they're going to take to save them time and energy? Operational trigger predictions, same thing. Can we find a way to tell the business what the next thing is coming? So as delivery teams are trying to move around their ecosystem, they have that guidebook or that crystal ball that we've talked about. So now, to wrap up a little bit and the most important part is the impact. So here's how we were able to know that these projects that we enacted had a positive change and we were able to measure that. We're still measuring for final numbers. Most of these are in early stages, but even so, the results are incredibly promising and already facilitating secondary improvements and process changes all over the business. Tons of energy and momentum. The biggest impact has come in 3 main buckets: so the number of cases we're preventing from escalating; the over amount of time and effort it takes for a case issue to be resolved; and the level of customer satisfaction with their experience. For escalations, we saw fewer cases come up to the recovery level when we started trying to predict them because people could intervene when they found out that there were indicators predicting the case could possibly be at risk. For handle time and agent effort, we've been able to free up resources for overall improved productivity because those people that had then been able -- been stuck in those other processes have been able to move into more value-added activities because repetitive or time-intensive processes, those have all have been removed. And lastly, for customer satisfaction, we found that customers react best when they feel known. So responding to their input and being able to predict some key concerns has paid off in much fewer negative mentions and low-scoring surveys. That's just a little bit about our journey. So to wrap it up, I'm going to kick it back over to Michelle.

Thank you, Melinda. Really cool stuff. So to summarize, leaning into predictive analytics is really helping us change our support landscape from reactive to proactive. This is a big deal, and our customers are feeling it. By focusing on that holistic customer journey, predictive analytics and AI empower us to move closer to our goal of truly being a customer-obsessed company. It's a critical, powerful tool to transform the customer experience. It really is imperative that we're able to not only evolve the customer experience to improve their end-to-end journey by making it a personalized experience but also how do we take it a step further and actually anticipate their needs, see around those corners through leveraging that historical data. Becoming more proactive by leveraging predictive analytics and AI truly is a game changer and a key customer experience strategy that everyone across all industries should truly be leaning into. All of this is to help the business become more proactive and accelerate that holistic customer journey and customer experience improvements. It empowers the business and key decision-makers to make smart, data-driven, strategic choices that ultimately transform the customer experience. It's all really cool stuff. So thank you for letting us be here today and for letting us share. And thank you so much to the CX Summit Digital for allowing Microsoft to be here. And now, we have a couple of pre-submitted questions from the audience. So I thought, Melinda, I thought I would ask them of you and you can go ahead and answer. You ready?

Yes.

All right. first question, how do you get your business leaders to see the value and make the investments needed in these efforts?

I think it comes down to 2 things, and they're very interrelated. And the first is communication and then bringing people along the journey, really getting buy-in from the top and approaching people with collaboration and enthusiasm in mind, and then really just constantly communicating out where you're at and what your goals are. I found a lot of the times, our projects gained momentum just when I'm having conversations with people and it sparks the right interest, and then we can partner up, and things move forward from there. So that's kind of -- helps gain momentum for one, but two, the more you talk about it, the more it kind of becomes buzz. And buzz, a lot of times, turns into action and can be really powerful in motivating leadership to get buy-in.

I think it's a really great point. We've seen a lot of that with the buzz and the momentum that's been created kind of going around and talking about it and bringing people on that journey and doing kind of a road show on the importance and the impact it can drive. So really great point. All right. The next question, what specific type of data is necessary to properly predict support needs?

So I mean you can make predictions with almost any type of data. So the 2 main types of data are qualitative, quantitative, meaning is it text for machine learning. The 2 things are mainly is it text analysis or is it a quantitative analysis. So as long as -- if you have -- as long as you have, for supervised machine learning, more of the quantitative side, if you know what you're predicting and you've recorded that, that's really the only thing that's necessary. So doesn't mean, though, if it doesn't exist that you're out of luck. A lot of our different projects, when we started out, we had a whole series of data and we knew that the thing we wanted to predict. We could see it in other systems, but it wasn't recorded in the data set that we were trying to predict from. So we would add an additional data point where we actually went in and created a brand-new piece of the data, and that was our prediction. So really, like for supervised machine learning, you need the thing you're predicting and then all of the supporting pieces that can help you make that prediction. And really, that's all the related data. And for unsupervised learning or for text analysis, which you can do both with quantitative and qualitative, really, that's looking for trends and repeating and patterns, where the machine is looking for an underlying fabric to the data that when we look at it from the 10,000-foot level and without that like rapid-fire iteration, we're not able to immediately see. So I don't want to say there's no such thing as bad data or no -- like data -- some data won't work, but like I haven't run across a set of data that has been unusable unless it was sporadic and random and really unclean and unable to be salvaged or if it was missing something and then we just found a way to add it. So I think it's really -- as long as you're scrappy, you can use almost any data.

I like that, scrappy. All right. Next question, what's the best way to consolidate data from multiple sources into one analytical data set to aid a predictive model?

So like, I can't not nod enough at this. This is an important question because it's hard because it can -- there's like 3 different ways that we have done. And this is the thing I was talking about with different data sets because you need to get them all together to be able to make a powerful prediction. So we've gone all the way from -- in some of our older model or some of our older data sets that are a little bit less agile and like can't really affect those as much, where we manually combine things. We just created a whole second data set of the historical data pulled about a year and updated it. That's kind of the most effort-intensive way. It's also maybe not the safest way, but like, you really know that what you're entering in has been got checked. But you can also do the same thing through, if you have an engineering team who can write a macro or a little script sometimes that can do the same thing the person was doing, pulling in the data from one state to another but automated. Those can be really powerful and quick, but it does require that resource. And then kind of a third in-between way is different tools like Power BI and Tableau. They have -- as long as you have a unique identifier that combines -- that you can match the 2 sets together, they can help you do inner and outer joints to bring those together. So that has to be set up in a way that, that's possible. It's not a magic tool, but it is a way that without kind of doing that manual copy-paste, you can start to combine things on the back end and make some initial pretty powerful predictions.

Excellent. Thank you. And final question I have here is how can AI help -- allow companies to be more proactive in their customer experience strategy?

Absolutely. So this kind of goes back to the data you capture. There's almost no customer experience gap that I haven't found a fun way to use predictive analytics to solve. So it's really about marrying the gap you have to the opportunity you have in the data. So some of the projects we went through, we've talked about the things that we've been -- the projects that we've come up with. To a small extent, we looked at the data and said, what can we do with this data. But really where most of it started was, what does the business need? And then we said, all right, here's what the business needs, if we are going to solve for that in a predictive way. So let's use escalation prediction as an example. What the business needed was to prevent escalations. They needed to have fewer customers who are getting that angry or needed that much recovery. And so we looked at it and said, okay, what is it that the risk is about. What do we need to know in order to say, is this case possibly an escalation or how would we prevent that. Well, we want to prevent Microsoft from losing money, customers from being angry, cases from dropping off the support path. And so you can start to look at the problem from these different angles, and say, why is that problem happening. And then you can look at your data and say, is there information in this data set that ties to each of these things. So for example, escalation prediction, 3 very simple metrics that we used immediately was how many times have they called in? Are they using words that display anger? Are they saying words like, "I will sue you?" Are they saying words like, "I can't believe you." And we would scan for those. So it's really about marrying the opportunity with the gap and finding what business needs, and then kind of being creative and working back to how you can use that tool to solve that gap.

Awesome. Thank you, Melinda. Well, that's it for us. I just want to thank all of you for being here today and a big shout out to CX Summit Digital for having us. Thank you very much.