4 practical AI use cases that will benefit any contact center

Good morning, good afternoon, and potentially good evening, and thank you for joining TTEC Digital's webinar for practical AI use cases that will benefit any contact center. In today's webinar, we'll explore high value use cases that you can quickly deploy through AWS intelligence to your existing contact center without a major upgrade. But first, let's take care of a couple housekeeping items. This webinar will be available on TTEC Digital's website on demand. We will have a q and a session at the end of the webinar, but please feel free to ask questions at any time via the questions tab on your screen. Please, keep your oh, alright. Let's take a look at the agenda. Before we jump into the main content, we'll do a quick introduction to TTEC Digital before talking about practical ways that you can leverage AI in your contact center. And today, I would like to introduce you to our presenters, Josh Sherbaum, vice president of professional services dedicated to AWS at TTEC Digital, and I'm Lisa Colbert, partner marketing manager for AWS at TTEC Digital, and I'll be your emcee for today's webinar. Taking a look at the power of partnership, TTEC Digital partners with leading technology providers to complete the capabilities portfolio. We are uniquely positioned to deliver the best in class CX services to our customers. Here at TTEC, we are a global systems integrator with established established practices in North America, EMEA, and APAC with over a million customer interactions supported by our deployed solutions, including our client, Amazon. We are an advanced AWS partner and a part of a bigger team with over forty years of experience in the contact center industry. TTEC Digital's AWS practice, formerly known as VoiceFoundry, was the first signed Amazon Connect partner, reseller, and systems integrator in the AWS ecosystem with over two hundred dedicated professionals focused on AWS's contact center solutions. We have more engagements on Amazon Connect than any other partner in in the market with expertise and Lex based automation and end to end customer experience for more, and more in the CX realm. Now with that, I'll hand it over to Josh to share more, information on practical application for AI in the contact center. Take it away, Josh. Good thing. Thanks, Lisa. Hi. I'm I'm Josh, as she mentioned, and I, lead engineering in the AWS practice, at TTEC Digital. And, what I'm gonna start talking about, initially is just, why we are talking about AWS or why, why, AWS would be a good, a choice for you, in your, in your situation. And, the first of which is that, AWS is a platform that is intended, for builders. And there are, various layers of, you know, building on top of each other these services. And so the accessibility and the depth and the breadth of the services, is, is remarkable and, has, yeah, has, really we we founded a lot of customers, obviously, to be in a successful platform, to use AWS. In addition to that, for the the situation we're talking about today, the way that AWS has, structured a lot of their AI services to be, purpose built for, individual missions or individual, capabilities, lets you go faster and, get value from them, from them quicker. And then in addition, there's, you know, economies of scale, with using AWS, etcetera. So, diving directly into our, our content today though, we're gonna talk about four practical applications, in your contact center today that don't require a big migration, don't require a a gigantic, investment or, you know, like a repositioning, a transformation of your business. So, like, the the chief problem that I wanna talk about today is that, customer expectations are continuously changing. And, many times during a transformation effort, we'll pick a target, we'll pick a goal, and we'll set off for that goal. But in the meantime, customers' expectations have shifted and have changed. And unless we build systems and unless we, figure out ways to, continually listen and learn and adapt our plan, we are always going to be, heading towards a goal that's that's outdated. And so what we really need is a sensing engine, a way to adapt, our plans and our vision to continually realign them to match customer expectations. And in a very, Amazonian, way, we have, kind of discussed a, or have a flywheel that that that we've, talked about with with customers to, walk them through how they can create this engine on their own. And and this is where we are gonna use, AI and AI services within our, customer experience systems to continually learn and adapt to changing customer expectation. So, the the just kind of a quick explanation of this flywheel is, we're gonna start by, measuring some performance. Right? Like, this is, typically, we'll, we'll just we'll start with a problem and we'll begin to to measure it. Right? And and after that measurement leads us to some insights. Oh, you know, we've got, customers that are calling in about this specific item. Let's say return of a very specific product. And so now we should, create a special return channel or there's a way that we can begin to use this data, the insight that we glean for it to make a clearer strategy. Right? Obviously, there's something to change about that product or a way that we support that product. So now with that clearer strategy, we can begin to provide a more personalized, experience. It's more aligned with the customer expectations because we we understand that they're more frequently than not that the percentages of times they're gonna be calling in about this particular topic or or product feature, etcetera, which then leads to a better outcome. And then those those better outcomes provide a plethora of digital exhaust for us to begin to find the next item of measured performance, to to begin to whittle away on and improve. And in this way, we continually adapt and learn and move towards, our our customers' expectations all the time. So we take a a staged approach to this in the AWS practice of TTEC Digital. And the first part is is really the the cycle, the first cycle that I talked about just just just a second ago, and and that is identifying, what information sources are available today that are, are able to inform your roadmap. What is the the the lever that you're wanting to move? What is the, the very specific business outcome that you're targeting? And then as part of that, pilot engagement, work a way to, you know, manipulate that value. So if it is increased deflection, then, like I said, there would be mentioning we would be helping you understand the, matching callers to targeted categories and then, you know, finding out which utterances to align to our chatbots, as an example. And so then the next stage, though, after we have our roadmap informed is building a data foundation. And, the the the bar to the barrier to entry for AI and training is lower and getting lower all the time. But without a strong foundation in data, it's still gonna have you're gonna have limited ability to to realize that value. And so the the next step and the next, stage that we that we see customers, typically through is building the foundation of beginning to collect and aggregate your data across your customer experience as well as your agent experience, identifying opportunities to increase the efficiency of your agents, identifying opportunities to to help the flow of your customers, through your your support enterprise. And and then the next stage, is we take that data, in stage two, and we, we enrich it by adding information from, Salesforce, for example, or Dynamics, possibly, but really using the information that we've gleaned about the customer's journey and and then augmenting that in a way that allows us to provide a more personalized experience to them, which is, you know, again, part of the the flywheel that we're trying to continually, evolve and push forward. And then in stage three, we can, with a strong foundation and an ability to make data driven decisions, we can be a little, more speculative in our projects and begin to, use, large, large language models to, offload, particular tasks. Or we can do, predictive forecasting on propensity to buy or, identifying, indicators that, for churn for customers. And, but it really is contingent upon building up to this, you know, a foundation and then augmenting it. And then, then you can kind of you're unleashed, if you will. So then tying it back to our core, tying this methodology back to our core topic today, there's four areas that we see as practical applications within a contact center. And the first of which is, using, you can streamline customer self-service processes and reduce operational costs by automating responses for customer service questions through, you know, AI powered chatbots, voice bots, virtual assistants. The next way would be to use, generative and other AI techniques to, support and enhance the capabilities of your human agents in tasks, around customer service, around problem solving, around decision making. It is increasing the the capabilities of your humans. And then as as part of that, related to that, you can analyze the content of calls and chats in real time and extract, you know, kind of valuable insight about what's happening holistically across your entire contact center, or perhaps, helping identify an agent that's that's having difficulty, that that needs assistance. But these, these analytics and the ability to assist an agent go hand in hand. And then the last, the last place is, you know, analyzing calls from your contact center and after the fact to extract, you know, really valuable customer experience insights, and improve, loyalty, etcetera. So, this real, as I said, this kind of the one that the customers start with frequently because it requires the least amount of, integration. And I'll, cover this one explicitly here in a second. But I wanna I also wanna shift just to to level set on the capabilities of a generative AI and particularly the ones that are relevant for the contact center. And in, you know, there's many more that I could have listed here, but these are the most relevant ones. And the first three I I really tend to group together, and that's summarization, extraction, and Q and A. And what that is is that is a, a large language model, which is a type of machine learning model that you can, provide a corpus of data or a volume of data that is then evaluated. And you can, for example, take a transcript of a call between an agent and a customer and summarize it. Hit the the keynotes. What were the action items? That's something that is, you know, very good for a large language model is capable of doing that. And additionally, another example for text extraction would be, finding things within your user manuals, for your customer support and and highlighting those key items based on a conversational interaction. So, asking a feature, oh, what, how many adapters does this particular server have? And it would, be able to extract that particular portion of the volume of data and share that with you. And then additionally, in a conversational style, you can begin to kind of query your data, query your insurance policy, ask it questions, get responses. This is, not necessarily thinking, but just taking an existing volume of data and highlighting what's already there. And on AWS, there are several building blocks, that I'm gonna refer to in the specifics of some of these diagrams. So I wanted to, kind of go through them at the beginning to make sure that, we were all on the same page. And the the three big categories of building blocks that we have is our AIML services. And those are the services that AWS has created that might be, geared specifically around, some portion of machine learning, whether it is document processing or computer vision or natural language understanding. The next building block that we're gonna use frequently is a way to, how we handle our media transport and how, voice streams get into AWS from, your on premises or your your existing contact center as a service solution. And then lastly, the the other building blocks, and I'll just I'll give a flavor of of several of them, around integration services. And these are the the services that that will be used to to tie it all together. So going one level deeper on the AI ML services at AWS, we'll we'll talk about three three different classes. The first of which are services that are a wide variety of out of the box ML models that are purpose built for common tasks, and this lowers your barrier to adoption. Examples of this include AWS Comprehend, which is natural language processing, AWS Rekognition, which is, computer vision. There's Forecast. There are, additionally then, there are platforms that you can train your own models, with AWS. And the most, notable one at this moment right now is Amazon Bedrock, which as of today is in, generally available, for all customers. And this is a platform where you can, quickly, fine tune or utilize large language models within your within your own flows. A, Amazon Bedrock has agreements with many of the most popular proprietary models as well as open source models, anthropic stability AI, Jurassic, Jurassic models from, AI21. And, this really is, an additional layer of convenience and automation on top of, SageMaker, which is the the lowest layer or the lowest level of, AIML service that you might interact with on AWS. And here is where you go even deeper and you expand beyond, you know, large language models and, perhaps you're doing a more traditional, regression or classification, model training. But this is, this is the platform to do ML ops in and all of your ML based tasks. And so the next building block, now that we have, services capable of performing, AI capabilities, right, depending upon, you know, what you're trying to do. The next thing you have to do is you need to connect your media to them. And the Amazon Chime SDK is a, is the purpose, the service that is built for that. And in this, there are there are several ways that it can be employed. One way, uses a SIPREC, and you can, just cast your, your calls, into Amazon Chime SDK for call recordings. From there, there's also options to do, real time and post contact analytics out of the box. But, getting those media transports, into into AWS at real time then gives you the opportunities to do, agent assist and, real time analytics services, using some other services. And so we'll we'll we'll hit one more layer in the building blocks, and then I'll show you how all these things kind of, go together. So the last layer that I wanna, share with you is is very simple, and you you might have you might already be familiar with it. And that is, the pattern of using an API, service called API Gateway, which, lets you just, set up simple HTTP based services and tie it to a variety of things on the back end. And this will be used, as a way to, handle the non media signaling that you'll need to do in your environment, as you handle call control or, want to signal to your agents or, send notifications to customers, etcetera. This is just another one of those, building blocks that is, put together. And Lambda is the other piece of that. And Lambda is a serverless compute environment, very lightweight, you know, function as a service, if you will. And from that standpoint, you are able to, get all of, from that standpoint, you're able to get, integration to anything else that's, that you might need. And so excuse me. And so putting it all together, you have at your disposal without migrating your your contact center a way to get media, to AWS, to process it for real time, total sentiment analysis. Or, perhaps you want to, do a transcription and an immediate summarization, or, any number of things that you now have available and you just, kind of plug these blocks in together. And so I am gonna give you, three real examples that we have of engagements that we've done for customers. And the first of which this is the last one that I mentioned in my opening, which is around post call analytics. And as I said, this is the the the simplest one to get started with because it takes, perhaps something that you already have available, which is your call recordings. And, you upload them to a storage bucket, in AWS. And from there, Lambda picks up those files and begins to process them, doing transcription using Amazon Transcribe, using, performing sentiment analysis on the transcription, maybe just of the of the words, not necessarily of the tone, using, necessarily of the tone, using, Amazon Comprehend. And all those, those values that are computed by these services then are are stored in a in a database. And in that database, customers then, you know, build a web front end or we, have a, like a dashboard that goes on top of it that lets you find, what are, the most common used words in the last three days of, of our transcripts? Or has any customer ever asked about this? And, there's, index searching of it is available. And, it is a very good way that, you can use the existing off the shelf services. It doesn't require you to train your own model. You really just throw your own data at it. You get the response and you use that response. And so, again, this is why customers, typically start with post call analytics to identify, you know, one particular this is the way that they inform their roadmap. We find, you know, what what is common, what is missing, what is, not currently available, a gap in the service through, you know, either sentiment analysis or searching through the transcriptions or classifications of the data that's in the transcriptions. And, then that informs the the road map for the next series of, changes to make to the customer experience. And then the, yep, the next, situation that I heard the next, kind of use case that we're gonna get specific on is how we automate, more customer interactions with a different patterns that this can take. But generally, customers will still call into your, contact centers of service or your existing platform, but you will have, connections of the the voice media through the Chime voice connector, which will stream that media to, Lambda. And that Lambda function will provide a variety of things. It can do real time translation, But one of the other things is it can then query a knowledge index. And so it will, or look for, use a large language model and a conversational style, along with Lex to, to, capture, deflect more of your support queries or allow customers to conversationally ask questions about their account, in a way that, couldn't be automated with typical slots and intents style, chatbot automation. And so within this one second. Thank you. And so within this within the situation, then we have a way to begin identifying with the first set of solutions where we have opportunity to automate customer interaction. And then we layer on, this, this solution to provide a virtualized, a self-service agent. And, so so now we've we've identified what's important to our customers. We've we've we've helped automate the things that we could automate for them and then allowing our cost our our human agents to focus on the most important things. And then the third situation we have, Ben, is is is how we enable agents, by exposing them to some of the same sources of information, but also, much of the automation as well through real time contact analytics. And in this solution that we build for customers, we, once again, they're using the Chime voice connector to do real time, processing of the the call as it's as it's happening. That can be a transcription, that can be, a post call of desummarization, provide a sentiment analysis, scoring, augment the data with, additional, customer, CRM, data, what services that they may already consume or what products the this this person has purchased. But then that, that information is stored in a an event, caching database or a storage engine. And the agent's experience then when the agent is working in their workspace, they're, signaled through this event storage mechanism as well. And, they're when they get connected to the customer, they have what they need available to them. All of this could be the real time transcription. But, additionally, they have their own conversational agent, that has been built into their workspace that they can query and ask questions. The agent, frequently is able to also, listen in or is provided the audio stream to process and is able to, make recommendations or provide automations, on their own. And so in this way, the things that are left for the humans to do are improved by, having more information available to them about what your customers need and, what their expectations might be. And, you also then have an empowered agent that is, has access to a wider range of your your enterprise's knowledge through, the conversational agent. And so what we, want to just kind of draw to your attention is the way that, if you apply AI knowledgeably and you begin to use it as a way to just set your path, right, make little improvements, the studies have shown that you can, use this this approach to increase your revenue and customer satisfaction. And this is the way that it, moves your, helps you continue to adapt and align to customer, customer expectations as they change. And, kind of some next steps that we have at TTEC Digital that we can offer to help you take, are, there is a there is an e book that describes many of these same scenarios and goes in-depth about the applications, of them and the potential benefits. Additionally, perhaps you have a question about you're not even sure where you're at, where you're, you're ready to take your first step, but you're not sure where that next step should be. There's a readiness assessment that that's available. It's a it's a it's an online survey. There's, you get some, results and and recommendations from there. Additionally, there's, an opportunity to to schedule an AI workshop with TTEC Digital. And in that, we we, really help you align what your your need is with the the current state of the technology and really, cut through a lot of, marketing, hype and really get to how is this practically applied in a way that lets you, change your business. And and and lastly, if you'd like, we could schedule a demo of any of these solutions, that we've, that I've discussed today, for you. Thanks, Josh. It was a lot of great information. We'll now open it up for any questions that you might have. Please drop your questions into the question section on your screen. Alright, Josh. How do you see AI growing in the contact center? I think there's a couple different ways. I see, that there's gonna be initiatives to create very large single models within your enterprise. And that might extend beyond your contact center, but it might not. And so kind of one of the things that I really recommend is that you're aligned with other AI ML initiatives within within your your corporation. But I am not sold that that there are gonna be, a one model to rule, your enterprise internally, that I think that that that the pattern that I'm seeing in the industry, whether it is from AWS or from, Microsoft or, the other providers of Google, OpenAI, etcetera, open open source providers, Hugging Face, etcetera, is that there's going to be many small purpose built models that are deployed within your application or within your environment. And that sometimes though, you will need to fine tune them. Sometimes you will need, you will use a pattern that doesn't require fine tuning. But I think at the end, the final result to me is that every company is going to have to develop some sort of practice to perform their own MLOps. I think partnerships are very necessary and it can accelerate you, but this is this capability that you're going to have to build, yourself, if you want to make the most of, of AI internal. Right. Great. Here's another question, Josh. How amongst our customers, you know, how do you see AI making the biggest impact? Well, I think it's a bit of are we talking AI? Are we talking machine learning? And I know that your question was was nonspecific about it intentionally, and I and I think that's, you know, AI is, you know, the the idea that this is, there's going to be autonomous, agency of of software in your environment, I think, is quite a long ways away so far. So, if we're talking about how we apply existing ML services and even large language models, I I think that, companies will probably dip their toe in and find a couple of successful use cases. I think that a lot of this is, the the problem with with with AI is that it it doesn't do anything out of the box, but it does everything that you tell it to. And that kind of creates some some difficulty for, you know, the the the the challenges. How do you align it correctly to the problem correctly? And so I think that, you know, as these services get easier to consume, companies are just gonna have to get good at trying, experimenting, seeing if it works, and then moving on to the next thing if it doesn't. And and again, I can't I just goes back to, I think the impact on the companies are those that are able to apply it correctly, will be able to accelerate their change. Right? Like the pace of change within their environment. Machine learning and AI is fundamentally an automation technology, right? Like it is automating things. And so, as it expands beyond what it's capable of automating, those companies that have kind of mastered the deployment of models and applying them to their business cases are gonna move faster and they're gonna accelerate. And, you know, I I think there's gonna, you know, be some some distance between, companies in in specific industries. Okay. Great. Doesn't look like we have any more questions. For our attendees, this webinar will be available on demand on the TTEC digital website. You'll receive an email after the webinar that includes a copy of the presentation as well as some of the additional information on next steps. I wanna thank you, Josh, for providing our attendees the top four practical ways to incorporate AI into their contact center, and a special thank you to our guests.