The ActivateCX Podcast
Join Frank Rogers on The ActivateCX™ Podcast, your resource for demystifying, clarifying, and providing guidance around AI, CXM, and the modern Cloud Contact Center.
In this Podcast series, Frank interviews Thought Leaders, Unpacks critical AI & CX technology, and addresses the leading Experience topics of the day.
#cx #customerexperience #ai #ex #cxm #contactcenter #salesstrategy
The ActivateCX Podcast
Unlocking the Secrets to AI Mastery
Get your AI Sorted https://activateCX.arroyo360.com/ai
----
Here Frank Rogers and Quinn Agen unpack the secrets of using AI contact center solutions to solve major customer experience needs! Our world is changing fast, hear from these two thought leaders to get a bead on where it is heading. Don't forget to subscribe for more tech updates!
Chapters
00:00 Does AI support Multi cultural-language needs?
02:37 How does AI Work with Omnichannel?
07:57 Scotty, We Need More Data!
11:36 Can AI Help Close Down Blind Spots?
14:30 Feedback Management gets an assist from AI
17:46 The Big Question...How do you get Started with AI?
20:49 Making the Case for Omilia AI vs Anyone else!
24:36 Closing
#podcast #contactcenter #ai #customerservice #salesstrategy
Hey Quinn, welcome to the show. Hi, Frank. Thanks for having me with the world getting smaller how do you think multilingual and multicultural user stories fit into the AI and is the AI something that is going to be a big part of how companies deal globally? So from the customer's perspective you in general definitely want to meet the customer where they are. And in the way that they want to speak, which is of course going to be their native tongue. the more languages that you can support in general, I mean, If it's a significant portion of your customer base, then it will justify the investment and having that language available. If you have one or two customers out of, a few million that speak a language, it might not justify the cost for making that language necessarily available to them. But, From an operations perspective, it's key that you have a platform where you don't actually have to manage like different applications per language. You can have a single application. And again, as the business owner, you can have Simply be responsible for managing how you want the prompts to be expressed in that local language. And besides that, you don't need to think of anything else, you don't need to redevelop the NLU. Everything is, it's developed once you want to add a language. You click Spanish or you click German, what have you? And you you write the prompts in that local language. So, it's not, it's not only important to be able to offer a lot of languages. It's important to be able to do so in a way that the business can manage it. And, that's one piece of it. And then the real time transcription is huge. Again, you think about some of these agent assist capabilities or you can't have latency, if you're a couple seconds behind, it's too late, the conversations are in real time. That transcription, all of that, processing power has to happen within a few milliseconds so that it pops up right on the screen to the associate while the conversation's happening so that, that the person can read and, give the answer. And then, I think the other piece there is trans translation, so maybe, for the small percent of customers that you may not have a language for can you, facilitate a conversation in their native tongue but translate it back to a language that your associate or the AI was, built to understand. When we think about the contact center world, before AI really popped, omni channel popped. So it was the precursor big change in the contact center. I kind of hate the word contact center because it has such an old connotation, but fundamentally the omni channel capabilities, whether people were working across, the standard channels, they were maybe on voice and email, potentially some chat, but they we're doing zero social listening. They were not, communicating via Facebook with customers via messenger, via anything to do with WhatsApp, with your AI and knowing the need to perhaps bridge from a conversational AI, somebody came in through the standard voice system and then bridge to another channel. What are your capabilities for facilitating that particular experience? Yeah, good question. So it's important to have a develop once deploy everywhere platform. What that means is you design your application once and you can deploy that application to really any channel, so you named a few there. It can be any digital channel, meaning text based or, chat based, whether that's a website, what's app, Facebook, any of, the different social media channels that are out there that could include email. And then of course you have voice, which is, which is real time. As we just discussed, you don't have a lot of room for latency. Everything needs to happen, basically in, in ahead of real time, if you will. And so you do. There are, let's say, whether it's voice or chat, of course, there are different things that you need to worry about and deal with perhaps needless to say, a voice conversation is harder, more difficult than a chat conversation. You don't have the, in chat for starters, you have the. You have the ability, you have the screen real estate, so you have visual cues you can present, a carousel of options, buttons it doesn't necessarily have to be a pure conversation or, back and forth dialogue based experience. You can have rich content, et cetera. And with a voice call, if you have a true omni channel platform, then voice will be. Just as good as chat or better. But to accomplish that, you really need to have specialized models for voice, and so if you take a technology that was designed for chat. And you say, Hey, I have this, chat bot here. I'm just going to, plug it into a transcription engine and I'm off to the races. It doesn't necessarily work like that. There's, there's voice is complex because of, accents, slang. Environmental noise, background noise and again, it happens in real time. So one thing that we can do is drive specialized models in that voice channel, depending on the context of the dialogue, and this is something that you couldn't do if you're using like a generic sort of vanilla API service because you get access to one model. So if you actually control the models, if they're yours, based on, what's happening in that dialogue in the voice channel, you can pull in targeted models that will give you the ability to achieve a higher accuracy rate. Which ultimately translates into better customer experience and more fulfillment. The other thing that we can do on the voice channel, again, in an omni channel setting is cross channel. orchestration, so you could come in on a chat channel for, whatever reason you would have to go, you could escalate to voice and pick up that conversation where you left off or vice versa, you could start a conversation on voice. Maybe you're applying for life insurance, for example, and you have to have some deliverables. You can end that conversation and pick it back up on the website. The other great thing we can do specifically on the voice channel is what we call multimodal experiences. So similarly to a web chat, where as we talked about, you can have rich content, while you're on the phone we can do sort of what was traditionally known as a visual IVR experience, so you're actually talking on the phone with the virtual assistant or with the human assistant, and they say, hey, can I send you this link? You want to add a new driver to your health insurance. I need a, a picture of the driver's license. Here's a link. You can take that picture and, I'll wait for you to upload it. So rather than that, customer journey of, Hey, I got to upload a document for, to get my insurance or you can do it right then and there. Or you could at least facilitate the option for the customer to do it right then and there. So, from what we've seen that, that sort of channel to channel deflection cross channel orchestration and multimodal experiences are really important to, again if we look at the business benefit, how are we maximizing the auto fulfillment or self service in a lot of these channels, AI solutions throw off a ton of data, obviously through everything that you just discussed, just then talking about going across various different channels and having engagement and recording the data from that. Everything from sentiment, we're understanding the journey much, much better. How do you provide access to that data? And do you have a very rich set of APIs that allow the customer to pull that data and put that into a data warehouse and drive unique visualizations? Yeah, you nailed it. That's exactly what we do. So we have an event driven data warehouse to your point, All of these interactions generate a lot of data. That includes the audio of what they said. If it's a voice conversation, it includes the transcription back and forth along with all of the, detailed session data like timestamp how long it took for the API to give us the response, from a speech to text perspective, the engine will actually give you. a variety of different transcriptions, and it will give the strongest one to the NLU engine. And that can also be taken into context. So you get all of this really rich data and we provide what we call conversational analytics. So those are visual dashboards, which give you all of the most relevant and pertinent information that you would want to know as a business owner. So, how many calls or chat sessions am I handling? What's the intent breakdown? What, what's the how frequent folks are asking for different types of requests. How successful I am at completing a task or automating a task. Are we 94, 95, 96 percent successful in completing payments and where are people falling out? And why that's, that's the big ones. It's great that we have a high percentage, but actually we want to know why and where we're failing so that we can do something about it. And that's hugely important is to be able to have. That really transparent insight and to be able to pick apart and say, okay, well, I see, folks are falling out of the journey here. Why is that? And typically it's because of a variety of things, maybe the API times out, maybe the speech to text missed something because it's an accent that the person has a heavy accent or there was background noise, or maybe they just asked for something that the system, didn't understand. I mean, you think about a couple of years ago overnight, folks were calling in asking, Hey are you guys closed for covid? What's the policy on this? So again, this sort of goes back to that, that need to have a platform that gives you that ease of use and agility. But ultimately, from a data perspective it's also important that you store the data. Going back to that previous topic around, compliance and regulation in a way that's safe. That data is encrypted. Again, if we're asking for PCI or PII information, maybe we don't save that data at all. Maybe it's immediately redacted. Maybe it's irreversibly masked. And then we also provide a set of APIs for our customers to get their hands on that data, this is a cloud service. And so we have an export API that gives customers the ability to take all of that data and either stream it or batch it into, their storage and then also, real time streaming and metrics for health monitoring and making sure that, at any given moment you have your finger on the pulse. I think that's a really good point that you made is that ultimately sometimes a source of truth Cannot be adequate to respond to the myriad of questions Maybe it just was not part of your universe of potential questions So is there something in the AI at least the reporting that pushes back data to the customer that says hey Here's here are 50 questions that we were not able to provide an answer to and and thus a jumped containment Yep. Again, you nailed it. That's exactly what we call OCP enhancements does. So OCP enhancements is a offshoot of our analytics. It does unknown intent clustering. So it will look at all of the utterances that were not readily understood by the system. It will run some generative AI and machine learning on those utterances. And present essentially the findings to the business user of, Hey, these were things that I didn't understand. These are, the new intents that I think correspond to these utterances. If that's the case, the business user can approve those changes and, ultimately update the model. And, program, whatever responses the business would want to provide to those types of questions. The other thing that we do is we use a tool, a proprietary tool called Pathfinder. So Pathfinder is a generative AI tool that will actually look at customer conversations either with the virtual assistant or with the human assistant and actually sort of reverse engineer, if you will how those conversations either could have been automated, what intents could have been present again, in a way where, We're automating all of that analysis and feeding the findings to the business for them to review, approve, and then ultimately, deploy those changes into production. So what this gives us is it eliminates a lot of the, what used to be traditionally manual effort from a tuning and optimization approach to automating that, that, that sort of cycle there. And conversational AI, natural language understanding, these living, it sounds funny to say, but they are living, breathing applications. Right. The macroeconomic environment can change the business environment changes. You launch new products, you have different business roles. It's not like this isn't something that you deploy and forget. You, there's a continuous learning. There's a continuous need to drive the business benefits. So having the tools that you, that will allow you to do that in a way that's efficient. And you don't have to have a, a team of 20 people or 10 people working full time to do that. That's, that's a big piece of it. And that again is where a lot of those, those advancements in generative AI have, really helped. One of the things that is somewhat of a standard in the industry is feedback management. And there are companies out there that focus 100 percent on that particular task and they'll integrate into various different contact center solutions. Do you find that your technology and your AI is starting to kind of move into that role a little bit as well to drive Feedback, which could be in the, in the form of a survey. It could be a survey in a chat could be a survey with a voice bot. Is that something that you're taking on and using as part of how you're measuring success in the business? Absolutely. So as you said, there's a couple of different angles there. One piece is of course, surveys. Those surveys can be done by the conversational AI platform itself. Some customers opt to do that, in the form of an email or a text message after an interaction really depends on, on the customer or call back, or they do it, in the same call, ideally before the customer hangs up. And, traditionally. Those surveys, anyone that's, done any of those surveys or been asked to do them. It's typically, you rate the experience, and that's how you get to the NPS score. And that's all fine. You need to do that because that's, what we've done for the last, 30, 40, 50 years is scale based questionnaires, but if you had the technology to be able to ask an open question, hey, how did you feel about the experience today? What didn't you like? What did you like? That's unstructured feedback that, you would be able to get from the caller and traditionally You could always ask those questions, but you would need a human, to either ask the question or review the feedback and really understand it from a business perspective. So what we typically do is, yes, we'll do the, traditional NPS score questions, but we'll do one or two open ended questions at the end to give the customer the opportunity to provide their direct feedback, which again, normally, You would need a human to do that of course, costs cost more money than, than the AI. And we do, with those surveys, it's a life cycle, so those surveys, we review them. And typically. I would say that a lot of what we see in the surveys, we already knew about because we can see in the analytics where people are falling out, where journeys are breaking down. So we know, Hey, there's an opportunity for improvement here. Maybe it's not the technology, maybe it's the way that the conversation was designed. Maybe we're not giving the customer the opportunity to do a specific, task. We're, making them go to an agent for whatever reason. And so the surveys not only help us to Sort of direct us in the direction of, Hey, like, where do we want to focus our efforts? But they also support and are, tied to the analytics and all of that rich data that we look at that really, it gives us the ability to effectuate change that ultimately is going to resolve any, any negative experiences that, that could be had. Fundamentally, AI is about adoption and essentially consuming it. So it's really a journey and definitely not an event. You just don't subscribe to AI and the job is done. It starts off a whole new track for you working with clients. Is your mindset really just galvanizing around one or two, maybe three use cases where you're able to fundamentally take a high volume and maybe a low complexity so that you're kind of damping down the risk right from the very start. Do you see that as the primary starting point for how you help clients move into this? Yes. So some customers have the budget and the appetite to go all in. And some customers don't have the budget and maybe not even the appetite, but the resource bandwidth, cause it's, we need a project manager, we need a API resource, so it's not only about always only about the budget. Typically what, what, the. Sort of the approach that we've seen really successful is the crawl, walk, run, so we would start from, hey, what is it that you offer to your customers today from an automation perspective? Let's start there making those experiences better. And then we can look at, Hey, these are the things that we always wanted to do. We could never do them with our old vendor because the technology was limiting X, Y, Z, and we can, position those as day two, day three. So you have sort of an evolution there of increasing the number of self service experiences or journeys that you can make available because the other factor here is. We could automate anything again. That doesn't necessarily mean that you want to, but if you do want to, you likely are going to need some APIs that we're going to be able that, we'll need to hit to exchange information and actually do, real transactions and, drive dynamic experiences with the customers. Maybe those APIs don't exist. Maybe they're old APIs that need to be redone. So a lot of times what we see is it's not only the. When looking at a project as a whole. It's not just, Hey, Omilia has to develop this application. It's Hey, Omilia has to develop the application, but as the customer, we have to develop all of these APIs, or at least, develop wrappers so that they can hit the cloud. So a lot of times also looking at. Not just the front end technology, which would be, the conversational AI platform, but also what are the enabling technologies that we have as an organization to drive or to enable, let's say, the conversational AI to have those conversations, and so that's again, another consideration of, okay, what can we do today? What do we want to do tomorrow? And that's how ultimately, you you, plan your, your conversational AI roadmap. So Quinn, you get onto an elevator with somebody who is clearly your ideal customer and it's 20 floors and you've got, two minutes to essentially share with that person in a non sales way, why they should be thinking about Omilia, what would you say to them? Six things better than human understanding and truly unstructured dialogues, so we talked about that. That's like 97 percent semantic accuracy, less than 3 percent word error rate, over 90 percent task completion, context switching, being able to go back to previous topics, ultimately having a technology that can. Can deliver a better than human experience. The next is best in class for voice, we talked about the fact that, yes, you can do a digital chat bot, but if you want to have a true omni channel solution you need to have something that is extraordinary for voice. And in order to have something that's extraordinary for voice, you need to have specialized models, prebuilt models and generative AI lead to 70 percent less development. So we didn't really get into it too much today, but this concept of mini apps where we have sort of conversational microservices that are prebuilt for specific industries and tasks that gives you all of this out of the box capability that you can tune and customize to your business rather than, taking a couple of different API services and essentially starting your development from scratch. The fourth is a white glove personalization and intelligent authentication, so that is this ability to really drive a better authentication and ID of verification experience for your customers leveraging either biometrics or NLU, and then also. Fighting fraud. That's a huge piece. There's been a huge growth in in contact center fraud over the last few years. It's a lot easier for fraudsters to attack contact centers and also in a way that disguises their actual identity. So having technology built into the platform that can expose, those attacks prevent against deep fakes. And then at the same time deliver that white glove personalized experience based on your customer's biometrics is very unique differentiator there and adds a lot of value to the overall customer experience. The fifth is new frontier automation use cases. So that is this ability to drive enhanced fulfillment unlocking experiences and customer journeys from a self service perspective that were not possible before, again, for, a myriad of reasons. And last, but as we've said. Certainly not least is the data security, government grade security, data protection, making sure that your data is in one place that it's being redacted and protected in a way that is compliant and acceptable to your regulators. And, you can ultimately report to your friends in fraud and security and cyber controls that, this is a solution that meets all of the regulatory requirements, certification requirements for these highly regulated industries. So those are the six in a nutshell. I don't know if that's the five minute pitch, but No, I think you could easily just tell them that, listen, if you want to be competitive and stay in business, you can just talk to us pretty soon. Yeah, exactly. Quinn. Thanks for being on the show. No, thank you, Frank. This was enjoyable. Thanks for having me.
