Unlocking AI: How to level up your data maturity to maximize AI value

Good afternoon or good morning depending on where you are. I am Chris Bennett. I I help lead our Azure practice here at TTEC Digital, and we're going to be going through how we realize value within our organizations from from from AI and the underlying data layer that enables it. With me, Suresh, you wanna do a quick intro? Hey, everyone. I'm Suresh Ramanathan. I lead data and AI, Anything to do with engineering, data science, and AI implementations, our practice gets involved. Nice to meet you all. Just a quick background on on TTEC Digital. We are a systems integrator. So we sit in between the management consulting and staff augmentation firms, meaning that we bring some expertise in the platforms that we're working on, as well as the perspective on doing this type of work for a number of organizations. So that background knowledge and and some of the the breadth of experience our team has got will help us inform, you know, some of the the opinions we've got today and how we're positioning and looking at artificial intelligence within the organizations and the critical data component under that. What we do within that is delineated between, you know, various parts of Microsoft, dealing with the the CRM, ERP customer facing front end, wrapping the customer experience into that, and then also around the contact center. So the ability to bring data and AI based solutions to the customer facing agent base to make their jobs easier, to make them more effective, and wrapping all of that in an organizational management framework that enables our partners to go further faster than if they would have done it themselves. So part of that DNA is a focus on customer experience. So the bread and butter DNA of TTEC is improving customer experience for various outcomes. We have found that the ability to do that increases not only brand recognition, which is kind of table stakes, but increases client engagement, grows revenue, and opens up additional business opportunities for the organizations that we work with. And there are a number of pieces to that. Specifically within our team and what we're talking about today is on the technology side, within, the hyperscalers. Obviously, there is a lot more to customer experience workflow that we dive into around the actual process, consulting, etcetera. But again, those perspectives are being brought to how we're applying technology to solving these problems. Now the first thing I wanna dive into is AI maturity. And and essentially what that is is how ready is the organization to be able to consume AI workloads at scale. A number of Microsoft did a study on this, interviewed over a hundred different organizations and leadership, and had them rank themselves, and then had them then take that and refine it into the individual components of actually being able to consume AI, and things changed a bit. But the big focus of that was on data. Data availability, data quality, and breaking down the silos of data to be able to drive into a singular platform to then be able to apply AI on top of. Without having those capabilities, it becomes fragmented and AI does not go very far within the framework of the organization. Now the Microsoft study that was done was on top of the Cloud Adoption Framework. The cloud adoption framework is really around the ability to consume cloud services at scale within the enterprise. So having our center of excellence, having a defined strategy for consumption and deployment, having cost controls and governance structure in place so that, you know, we have a relatively consistent cost model for the cloud resources that we're deploying out, and we've got an ability to integrate that into our existing business processes. If those things do not exist, then adding in an AI solution on top of that is is, doomed to fail in some way. So the big thing is lay on top of that foundation, that's where we start layering in our AI strategy planning and going through essentially the same workflow in terms of process and defining all of those controls and methodologies so that we're able to actually start consuming AI workloads at scale. And that's the key here, is there are five logical stages to maturity. Exploring, know, doing proof of concepts, identifying use cases that AI could bring value to the organization is where a lot of organizations are at. So, more of them are are in some, form of implementing. That could be, you know, doing small scale proof of concepts, maybe rolling out something within production within a, you know, a specific opco or department, doing, smaller scale implementations, to be able to, you know, bring value to the business with AI. Where all all organizations wanna get to is is really realizing value. So not only in terms of your internal operations or being able to use AI in outward facing use cases, customer facing applications, and ultimately driving new services or capabilities that you that you can drive revenue from. That's the ultimate goal, and it's a very small pool of organizations that not only have the Cloud Adoption Framework down, but also all of the individual components between those two spaces so that we can actually consume at scale. Several drivers behind this. Not entirely surprising. You need to have a strategy aligned with the business to be able to get value. If we're just going and doing AI for the cases or for the sake of doing it or diving into technical use cases, then, again, it tends to to flail a little bit and not get fully off the ground. Then we need to have the technical foundation. And part of that is a robust data strategy that extends into the AI strategy once you have the data. So this is where we spend quite a bit of time, and Suresh's team has done a lot around how we aggregate all of that data into into a usable state. Part of that is is who has access, the modeling, and everything around the overall strategy for it. Organization and culture is important in that, you know, if if we've got delineation of duties and and fragmentation of authority within an organization, it tends to slow things down because everybody wants to own their piece of the pie, and and and that, you know, leads into the final piece, which is governance. If you've got a center of excellence around cloud, you are well positioned to be able to very quickly iterate on that to bring AI workloads into place. Security falls in this space. So the confidence of being able to bring data into AI, and I'm not talking about ChatGPT, I'm talking about privatized AI based solutions where your data is secure, that is a critical piece, especially depending on the industry you're in. And that we found is one of the big blockers to adoption is not having proper data governance and security, which gets back into the data platforms we're going to be reviewing further into the presentation. Data is the full foundation that we're after. You know, the ability to get everything into one place and to build a modern data platform is table stakes for being able to even play. If you're not doing this, then it's going to be something relatively commodity, and the and the breadth of of of the AI capability within the organization is gonna be much smaller than it would be, and therefore the value of going down the road is gonna be diminished. The other piece is we wanna actually be able to take action. So it's it's it's you know, there's a lot of technical capability that's there, but how are we going to tie the outcomes of that into something that the business can actually do something with that's going to drive revenue or going to drive some term of efficiency internally that that lowers costs and increases productivity. Everything starts with data, as as we mentioned. So there is our the way that we look at this is is bringing all of our of our our data into a unified platform. So, you know, a a cross functional single source of truth that we can use for analytical purposes, that we could use for enterprise applications, That that that foundation has multiple facets to it. What we found is organizations that invest in that are able to use AI solutions, are able to have forward looking predictive analytics, and be able to get actual true value out of their data. And then depending on the organization you're in, there may even be an opportunity to monetize some of the data and findings within this and and empower your end users through self-service capabilities. We find that that is a a driver in actually being able to get the business case to invest in, you know, building out the platform to begin with, is it's not just tied to a theoretical AI use case. It is tied to a number of outcomes, which AI happens to be one of them. Tactically how this looks, we have a number of accelerators that we've built out from a code base perspective to be able to get things going quickly, but we're going to build a lake house. We then use that lake house to bring in our data science and AI capabilities and reporting and downstream integration. Foundational pieces to all of the interesting things that Trash is gonna go into here in a bit on what we actually action on top of this, we've done this. We've done it at scale. We've done it in various different industries. The patterns are are relatively consistent between them. You may already have some of these components. You may already have the core. What matters is that we've got these these layers of capability that get laid down, the foundation being the ability to get data aggregated and landed in Azure to be able to start working with. Now what we what we found, and this, you know, follows the the the logical flow of the the past decade or so, is dealing with analytics is pretty easy. Getting a Power BI dashboard landed on some structure tables that drive a business outcome, that's been happening for a long time. Being able to take all of that data and all the trends over time and be able to make predictive insights for what's going to happen, what could happen, and then be able to action that from a business perspective, that's where we wanna get to. We want to understand our current dataset, be able to report on it and use it. And then really what we where we are we spend our time is being able to drive a number of of value streams out of the data itself. And depending on the organization, what what industry you're in, this could be a new services opportunities around, you know, predictive maintenance, for example, around, you know, field equipment. It could be AI based optimization of workflows and scheduling based on some of the analytical data around the, you know, employees or field service folks. Like, there there's there's a number of different use cases that kinda flow from this paradigm. And we have seen, you know, the the value of being able to build a three hundred and sixty degree view around our customers, our business, and and various parts of our our employee experience as well. And we do all of this through the customer data platform that that we're that we're talking through. And and the and the the key piece there is it's a it is a worthwhile investment in terms of time and resources simply because we're able to do a multitude of things with it. Now we're gonna go into how are we actually, you know, leveraging AI within within the the the customer experience and and realizing value from it. So Suresh, how are we actually applying AI to the customer's various components? Sure. Thanks, Chris. As you rightly mentioned, right? AI driven CX solutions are playing a pivotal role in ensuring businesses deliver the right messages at the right time through the right channels, all while fostering deeper customer relationships. AI solutions are not only personalizing experiences, but they also help optimize the effectiveness of every customer touchpoint engagement, right? Enhancing both customer satisfaction and business performance. We have implemented a variety of AIML solutions that help our clients customize and enhance their loyalty programs, optimize their engagement strategies, or also leverage voice of the customer insights to improve the entire customer journey. Right? Let's start off with our, what we call the loyalty refresh program. We all know how important it is to build long term meaningful relationships with customers. Our solution focuses on identifying the key drivers of loyalty, right? Basically helping us understand what factors matter the most to make customers stay loyal with the brand or defect the brand. As part of this, we deploy multi dimensional segmentation models that groups or segments customers based on their behaviors, preferences, and interactions with the brand. It helps ensure we are targeting the right customers the right way, right? Additionally, we also leverage machine learning models that are inferenced in real time to predict the likelihood of a customer defecting the brand. Think of at risk or churn risk models, right? And also recommendation engines that suggest products to customers based on their interests, keeping them more engaged, and also NBA and BX solutions that recommend personalized actions or offers that are most likely to resonate with the customers. The next one is around engagement optimization. The solution that we deploy to our clients, it integrates all of the marketing data, the marketing response data, all of it in one place, providing insights needed to manage the lead management process better and also to employ data driven customer campaign strategies, right? By employing or by identifying potential drivers of engagement, the solution predicts which customers are most likely to respond to your campaigns, making the outreach efforts more efficient, right? The ML models that we typically deploy in this space is around lead prioritization. Think of rank ordering the lead base based on the conversion potential or the revenue potential of the customer. The other one is around channel preference models, right? That provides insights on where the customers want to engage and how they want to engage, right? Be it email, SMS, voice channels, social media, right? Being able to understand the likelihood of customers' channel propensity is very important in that aspect. We have a very interesting solution that we have deployed for very many clients around conversation intelligence. Most brands struggle to understand why customers are reaching out to their contact centers, especially when they have invested so heavily in digital channels for self-service, such as IVAs, customer facing voice bots and chat bots. Our solution helps understand exactly why customers are reaching out. The solution identifies the top contact reasons, analyzes the associated sentiment, assesses the complexity of the engagement of the touchpoint, and also it spots the emerging topics that are gaining traction in the contact center interactions. These insights are very invaluable to the business stakeholders. It allows them to proactively address gaps in self-service channels, be it the IVAs or even the human agents. Also it helps them redesign or revise the training content for the frontline staff. And also it helps to curate the relevant knowledge articles that are leveraged by the AI assistants or the human agents. The other solution is around Journey AI, which transforms customer interactions by it integrates all of the customer data in one place, giving agents, the bots or the human agents, the relevant insights needed to deliver optimal outcomes. It tracks each customer's life cycle stage and also uses a variety of ML models to predict the next best action or the next best experience based on past behavior, preferences, to personalize interactions. The key feature I would call out in this solution is the self learning loop, where our solution continuously optimizes the journeys to ensure customers are receiving the most relevant content and offers at every stage of the relationship with the brand, right? These solutions that I just spoke about, we have deployed a combination of these solutions with a variety of our clients, right? We have clients who leverage the loyalty refresh program, the engagement optimization, and the automated journey management, right? For example, a major automotive brand leverages our solution set to manage and optimize their overall loyalty program. And also we have a major health insurer client who uses these solutions to maximize health plan enrollments and also to improve their CX. And then we have a streaming client who leverages the engagement optimization solution and also the journey AI solution to optimize or proactively address product issues and also to effectively manage customer support tickets coming into the contact center. So Chris, any questions that I can help address? Yeah, I think that the big thing coming out of that is we need the customer data platform. The extensibility is the ability to bring all of the external SaaS applications, disparate systems into one place, and then to be able to layer, the AI and ML solutions on top of that data. I think the only the only question there is, you know, what typical use cases have we covered? Really, what's the art of the possible when we land that? Because it seems that, we're solving for some issues for sure, But really what we're doing is we're laying down a foundation to be able to do more beyond what we initially enter at as well. Yeah, the key is the foundational data layer, right? There is no choice. You need to have a customer data platform that brings in unifies the customer data from various sources and applications, right? Integrates them, connects them together to build the customer three sixty. Think of it as a composable architecture where you have a foundational data layer and top of that compose on the foundational data layer, you are able to deploy various ML solutions, right? Depending on a client's use case, they would be able to pick and choose, Oh, I wanna, I have journey AI. I wanna be able to do some speech analytics, or I wanna mine for insights from voice of the customer interactions, or I wanna predict likelihood of a customer to make a purchase in a short term or defect the brand in a short term, right? Being able to deploy these deep learning machine learning models that reference the foundational data layer is a very important aspect of it. So decoupling the foundational data layer and being able to build ML solutions or AI solutions that helps the brand achieve certain business outcomes is the key. As an example, if you want to walk through how that looks from an integration standpoint, I think is the big piece that folks are gonna wanna understand. Because we know what a data platform could do, but the extensibility and the things that we can process within it, I think is the key to figuring out how it fits into an individual organization. No, absolutely. Right here is what I would call is a functional architecture of what a CDP should do, right? At a very high level, a customer data platform should do few things, right? One is it should be able to seamlessly integrate with source applications, right? Be it your CRM, CCaaS, WFM, your marketing automation systems or your traditional transactional warehouses that capture all of customer transactions. Being able to seamlessly integrate and land the data in a raw layer, raw data layer, and then quickly being able to take the raw data and curate it for analytics and reporting purposes. That's the next step, which is where the curated store comes in, which is like, either you can leverage Databricks, which is a very powerful ML and also a data curation layer or fabrics or Azure fabric in Microsoft could be leveraged for the curation layer. And also in the curation layer, it is very important to be able to use some type of a low latency data store that allows for external or frontline applications to communicate with your platform in a very low latency fashion. That is where we leverage Cosmos database quite a bit that Azure has to offer. And then the next one is around the machine learning layer. The analytics are the ML layer, which you can leverage Databricks MLflow, which is a very matured platform, or you can use what Azure has to offer around Azure ML and also Fabric, right? Then the last one is around the consumption layer, which is how are business stakeholders going to consume the data or take advantage of the data that you've assembled in the platform. It can be a BI layer leveraging Power BI. It can be an analytics sandbox powered by Databricks or Fabric, and definitely around having an API layer that other applications can consume. The gone are the days of performing or running predictive models once a week, once a month. You want models to be inferenced in real time. And all of the ML models need to be encapsulated into an API where they are callable in real time. So the API layer is very important from a consumption standpoint. And obviously you need to always think about data compliance, data security and data privacy. They are very important than anything else that you do and deploy in these solutions, right? Because being able to manage customer data securely, effectively is very important. And we leverage all the bells and whistles that Azure has to offer around ensuring that the application keys, the data is always encrypted in storage and at rest and being able to prevent DDOT attacks, providing only the right access controls to the right people at the right time, the role based access controls that Azure and Databricks both have effective solutions to deliver and also being able to mask PII data in the lower environments using dynamic data masking and techniques of that is very important. Excellent. Well, really appreciate your expertise and time here Suresh and really appreciate everyone's time walking through all of this with us. If there's any questions or would like to have a deeper conversation please don't hesitate to reach out to us. Absolutely, thanks for coordinating the webinar Chris, appreciate it.