Good. Morning good afternoon or, good evening depending on where you are in the world and, welcome to today's Enterprise Connect webinar, contact. Center AI fireside. Chat bond surd by Google and broadcast, by informa, I'm. Eric craft with enterprise connect, and I'll be your moderator today, we. Have just a few announcements before, we begin. This. Webinar is designed to be interactive the. Dock of widgets at, the bottom of your screen will allow you to learn about today's speakers, download. Resources, share. This webinar via social media outlets, and participate, in the Q&A session, that takes place at the end of our presentation. Wise. Will advance automatically, throughout the event and, you may also download a copy of the slides via the resources. Widget, or, the, end of our webinar will ask you to complete our survey which, is found on the right hand side of your screen, please. Take a minute to fill this out before leaving us today as, your feedback will provide us with valuable information on, how we can improve future, events. Lastly. If, you are experiencing, any technical problems. Please click, the help widget found at the bottom of your screen or, type. Your issue into the Q&A area and we will be glad to offer one-on-one, assistance. Now. On to the presentation, contact. Center AI fireside. Chat, discussing. Today's topic, is Sheila McGee Smith founder, of Magee Smith Analytics Matt. Jones product, manager at Google and. Celeste, grade Alliance. Relationship, manager, at Google, geela, Matt Celeste, over, to you. Thank. You Eric and welcome, everyone. So. If, you, are regular. Readers of no, jitter you, may have noticed. Within. The last week or so I wrote. A post. Called, are you ready Google. Contact, center AI is here. And we. Are very lucky today to, have two, of the people who have helped bring us to, this point in time, Celeste. And Matt and I. I have created. Some questions, that I, know that if you were here and you had an opportunity to be with them in front of a fireside you. Might want to ask the. First one Matt, is why. Is artificial. Intelligence, so. Hot, right. Now I, mean, we think about it for the last you. Know AI started, 50, years ago I mean singularity. Theory you. Know was something that came up in the 1950s. Some. Of the logical. Reasons why AI now is there's so much greater. Computing. Power there's. So much more data available to. Support. Machine, learning, certainly, cloud computing, I think has been such a big part of it but. What is is your perspective on why artificial. Intelligence is so hot right, now Matt. Thanks. Sheila I think it's a really. Good question to ask so, I think, you've hit on a, couple of the key points there there's aspects. Of on the data side aspects. On the hardware side as well as on the software side and so at. Google there, we've kind of noticed significant. Advances, in technology. Both on hardware, where we have developed what are called tensor. Flow processing. Units or TP use which. Are specific. Hardware to take advantage, of the vast amounts of data that's. That's not readily available for. Kind, of AI as well, as we've invested. In developing, algorithms. That can take advantage of that hardware so that we can really make this, vision, of AI that as you said has been, around for decades it, can make it into a reality, now. When, we think about context. Centers specifically. There, is a unique, opportunity to, take advantage of this, historically. Brands. Have tried to identify ways, to, reduce the cost to serve customers, while, simultaneously, trying. To improve, customer, experience. Now. Unfortunately, this. Has led to automation. Experiences. Such as IV arteries I think most of us can agree don't, feel super, customer, centric right, and so we're. All too familiar with the experience, of having, to listen to all the choices. They are multiple. Times in order to make a proper selection only. To then be forced to do it all over again and keep going deeper and deeper through, this hobby archery, and we. See it in as many ways our attempt.
To Modernize and, improve the customer service experience actually. Ended up in shifting, part of the burden from, the contact center to the customer and so. Artificial. Intelligence, is, relevant for contact, centers now because the, technology, is at a place where we can really remove. That burden from our customers, and incorporate. Technologies, that, actually deliver, on this goal of improved, customer service. As well, as increased operational. Efficiency, now. Conversational. AI specifically. Which is technologies. That relate to how conversational. Experiences, are allowing. Us to create systems, I can understand. What, users are contacting. Contacting. Us about let, them speak to us in a way that's natural, and simple, and kind of feels like a human-like conversation, and, so, we should be seeing use cases for instance one of our customers, Marks, & Spencer has. Has. That migrated, to a truly conversational. IVR where customers, call in and they just say naturally. How, they typically would what, their problem is and what they're looking for and, they've been able to scale that in a way that previously. Wasn't possible they're, handling over 6 million calls routed and that conversation, IVR. Can handle over 2,000. Customer, intents now. If you think about in the context of having a human rather that call that's a lot for a human to manage right to really accurately, route. To thousand different types of requests, and so, we're seeing that now is really the time for artificial, intelligence to, have an impact in the contact center space. Excellent. Yeah, it's interesting, to think about. Having. Been in the market as IVR. First, came in right, in the early 90s, originally. What, we wanted, from, IVRS. Was the ability to automate, some portion. Of calls right, maybe, it was 5% my, maybe it was 10% things. That could be very easily answered, and when, I think about conversational. AI I, think. About what, you know again we don't want to get rid of all, live, agents. Right what, we want to do is perhaps take that 5 or 10% that. The 1990s. IVR, was, able to automate. And perhaps, lift, that a little higher you know and to your point take. On some of that burden that the customers, have had to explain everything and let, artificial. Intelligence, maybe get us to 20, 25 percent, of, the kinds of interactions, and questions that customers have being. Automated, and then so much easier for customers, let's. Move on, exactly. What. Is Google. Contact, center AI what is what is that. Good. Question so, Google. Context, where AI is a, solution, that brings together a couple, of different conversational. AI building, blocks to. Create a product specifically. Tailored, to serve the needs of the contact center to. The first building block we think of as cloud speech so there are a couple of components to cloud speech itself the first is speech. To text transcription, which. Provides the ability to take audio, and identify. Specifically. What words the user is actually saying the. Other side of that is text speech which. Provides, the ability to take the text and then turn it into an audio signal and it's actually a voice response saying, those words the. Other building block we, think about is dialogue flow and dialogue, flow is effectively, the core, natural, language understanding piece. Here this. Is what allows us to go from this sentence. Or words to, actually, meaning, to understand, what is this user or, customer trying, to convey to us what is their intent. No. We've, taken these building blocks and then we've added some very contact, center specific, functionality, as well as, partnered, with lesson providers to, package this into, what we call the context, in our a I solution, now. That solution, is made up of three key, components so, the first is virtual. Agent which is effectively, voice, box chat BOTS you can think of conversational. IVR experiences, all of which are powered. By dialogue, flow the. Second is agent assist which provides real-time suggestions. To customer service agents based, on what the customer is saying and this, is where we think what you discussed before is you want to have the virtual agent can handle hopefully, 25, or more percent of the, simple, calls and then empower, the agents, to be specialists, so they can handle the, tough calls and the. More interesting use cases and then. The third component is insights and this is where we enable, contact center managers and quality assurance analysts.
To Better understand. What's happening, in the contact center in real time. So. What, are the words that you've used a few times is, intent. And we. Talked about Marks, & Spencer, creating. A conversational. Interface. That can handle 2,000, intense. Can. You get the audience a couple examples of, what an, intent. Might be so. We can people. Can understand it. Yeah. Definitely. Um so. An intent, at the end of the day is just what is the user trying, to say to us what is trying to understand, what they want to do so for, example if you're, Marks & Spencer then, an intent might be I want to understand, what hours, are certain stores open right, or if. You're, a. Telecom. Provider it might be I want to understand, what are the international. Travel plan options it's, really at the end of the day what is this user calling, about and so, as, contact. Center manager you could understand, that your users have hundreds. Of different reasons that they might call in and that, means you could have hundreds, or even thousands, of different intents, and that's, what is at the core, of this natural language understanding technology is, the ability to effectively. Understand. What is this user reaching. Out to us about and let. Them talk, to us in a natural, and kind of human friendly way. Okay. Now I'm going to put on the spot a little bit. Are. There types of intense, that don't lend themselves to, conversational. AI that. Once. You when you when you see them you say okay that's really going to have to be a human, agent. So. That's a great question Sheila, and I think the way to think about this is what are those, types. Of intents that you want to have a human touch. Right so if, for, instance maybe you're working with a bank provider and someone, lost their credit card and they've identified that there have been some fraudulent charges that. Customer, wants to have a human in the loop to feel like they that. Their concerns, are being met as well as to feel a sense of confidence that their. Account is being managed properly there. Are also other instances, where maybe that's a truly truly complicated. And unique, instance. You. You, rarely want to build the. Intents to handle every, specific, type of intent, that. Might come out there you want to get the intents. That. Can get the most coverage right, and then at a certain point you, let you can start to lose a little bit of the, their ROI if you get the intent that will only show up once a year right, so, the, way I think about it is kind of to two, ways to think about it you. A want. To make sure if there should be a human in the loop that, the customer, knows that you care about them and you bring in your agent to handle their case specifically, on the other side is if it doesn't happen very often but it might not be worth the effort of building out a tip for it.
Excellent. That really helps, us. Understand, you. Know how this applies to our businesses. So. Let's move on. Why. Does Google see what. If Google sees its role in bringing a eye to. The contact center. Because. For, example you're not a contact, center provider you, know you're not saying you have a contact, center, software. Solution, but, what is your role what. Is Google's role here. So. Pierpoint we are not a contact, center provider but, we have been investing. Significantly. In conversational. Iia, technologies, for, the past few years and. I, think this kind of makes. Sense right if you think about the. Development that we've needed to build to make Google search and Google assistant, as effective. As it is currently we, had to develop and had, to invest in both the algorithms, side as well, as the scaling side to make those consumer, products be able to integrate, conversationally, I at scale, and do it well and, the. Reality is that conversation. Is it's, very relevant to con to the contact center, so we. See this as our opportunity, to improve. Consumer. Experiences, for all brands, in many, ways it's, an opportunity to democratize. Better, customers, experience, and, ultimately, we, hope that brands, spend, less time worrying, about how can they technically, bring these experiences. To customers, and moreso. Spend time thinking about how do I create an experience that's true to my brand and allows. Me to develop a deeper, relationship with our customer we. Like to think that we are the, we. Can create the best conversation. On AI technology but, our customers understand. Their voice and their brand and that's those are the things that they should be focusing on within, the context of. So. Are you going to walk through these elements, you touched, on virtual. Agent and agent assist and insights, but. Could you talk a little bit of more perhaps about, how. I as. A. Company. Might. Install. How I might, implement, Google. Contact, center AI do, I have to do all of these things can I do one, of these things can, happen how would that work. Great. Question and thanks, for that Sheila so as. I. Mentioned before they're kind of these three components to contact scenario there's the virtual agent which is the voice bot the chat bot agent, assist which, is how we actually help the human agent while the with the customer as well as insights which gives more information to, your customer service managers, and their. Quality assurance analyst now, when we think about where to start there are a couple places but the, easiest, place frequently is on the virtual agent side you know under, you'll have an understanding of what. Our customer is calling about and so, you could build out a conversational. Experience, to start, to offload, some of the burden of the mundane questions from your agents and, secondly, after you get the virtual agent started then we start to think about agent assist how can we empower. Those agents to be more knowledgeable and, have higher, kind of first-time fixed rates, things of that nature, but. Frequently customers we're seeing customers start with virtual agents and then move to agent assist and insights afterwards.
Got It okay. So I think as we, move on to the next question, what. Has Google's, approach been, to. Building, contact, center AI solutions, I'm going. To bring in Celeste, grade here who has been working with, all of these partners to, answer. This question what. And. Thank you Sheila and I'm very excited to be on this fireside, chat because I have been working. These partners for years now and so it's exciting to see where. We have done and what we've done together so. Our ultimate. Strategy, as, we look at the space and a lot of what NAT talked about earlier as far as truly democratizing. AI is. How. Do we ensure that, each one of these contact, centers can take advantage, of truly. The the. Best types of customer experience possible, and, make. Sure that it's not a science, experiment and make sure that we're not doing. Pocs, for the next five years and, we realize, the. Best route here is. Really. Partnering. With it the the. Telephony. Providers, that all these contact centers providers, have and just, adding, in AI on, top of their existing solutions, this. There's a lot of things that make sure that you're. Leveraging, the solutions, that you've had for, years to come that has the overall contact. Center experience right. And B it, makes sure that you can really. Have this great experience, that's, already integrated with these providers, so. You can start taking advantage of, it you, know in three to six months, versus. Five years after you hired all, of your data scientists, so. You see all of these providers and as part, of it most. Of them have already been pre integrated, a bunch. Of them after our G announced they've also announced, their. GA and we, have, various. Different system integration, partners also to really help your. Organizations, in whether it's building, more conversational. IVR, whether. It's assisting, a lot of that change management, to, really ensure. The. Best experience for our contact center customers. And providers. So if I look at this list Celeste, I see, companies that are pure. Cloud. Plays, you, know like an ice and contacts, or you know Salesforce, or Tulio and then, I see others that have premises.
Based Contact, center solutions as well. As cloud, right, people like Genesis or Cisco, or Avaya is. There any difference, in how Google, works, with those customers, looks different you know trying to apply Google, contact center AI to. A premises. Solution, versus. To a cloud solution, what's different from your perspective, I. Think. That's a great question and I think it's one of the values. Of our partnerships, and working with these partners for years now and that. They've. They've, worked, on areas, in which in integration. With contact center AI is, essentially. A hybrid to. Our making it easy to actually, integrate with on-premise. So. It doesn't matter if there's maybe an on-premise. Solution. It's so. Long as it fits in that software, context, we figured out ways in which to integrate with context, Center AI so. It doesn't matter if you're on Prem or cloud, it, can still be very easy to enable. Yeah. If I look at some of these vendors. Several. Of them that have both premises and cloud. Solutions have. Created. Paths for their customers, to, begin to take advantage. Of cloud solutions. While. Staying on premises, and one. Of the top things that people want to be able to add to existing systems. Is artificial, intelligence so it's great that Google. Has that approach that says okay if your cloud yeah, that's going to be easy but, if your premises today we. Will help be part of that bridge letting. You take advantage of. Artificial. Intelligence which, for, the most part does come from the cloud but, on top of your premises solution so, thank you so much good stuff, thank. You. So. We're going to go back a little bit tacky we're going to go looking a little techie here with Matt what. Is an API an, application. Programming, interface. And how. Do you use one, and you, might even start with why is Sheila asking, me this question. So. I. Would, presume, that you're asking me this question Sheila because at the, end of the day context. Center API or context in our AI rather is a. Suite, of api's that we've built to, fit. This context, in our use case now. When, we talk about api's, they, are pretty much a way for, different, applications. To communicate with one another in the. Case of contact, center AI our, partners, have integrated, with our API s as a means of bringing our AI technology, into, their context in our platforms, so. Let's, kind of walk through an example to talk about how these actually work so let's. Take for instance a, customer. That uses the contacts in REI a virtual agent technology and an, end customer, calls into a contact, center now. The, contact, center as a service, partner will send, us that audio via an API call then. Within, the Google size our application. Will first, transcribe. That audio, to text using. Our speech to text technology and, then, afterwards, we'll send that transcription, to, dialogue flows natural, language understanding to. Determine, what is the right response for, this user the. Next step is to take that response, from dialogue, flow that's in text form and actually, convert it to an audio signal so, we take the text to speech and use, that to create a voice response, that, we've then finally. Send back to that context, Center as a service partner using. Once again in AP on call and then. Ultimately that, your contact, center as a service partner plays, back the audio to the end customer to, complete that interaction, now. While. We use API, to work with our partners one, of the benefits of these partnerships is that customers, don't need to concern themselves with, our ap is our. Goal, for contact center area at the end of the day is to, make it as low to no code as possible, so while, this concept of API is in, the background enabling. All of this we, don't want our customers have to worry about it. So. I noticed, on the last. Slide, that talked about the partners that you have that it was not just contact, center as the service companies but, also there, was a, relationship, management company, their CRM, so, there are times when, Google. Is working. With more. Than one other, application. You. Know like a contact, centers of service and a CRM, and does that get more, complicated. That's. A good question so there. Are times when we are working with like a CRM, provider fred sense because there's, relevant. Information. Stored in. Their. System, that can affect and then kind of improve, this customer, experience, so, in. Those instances there, is another party, that we are communicating with via api's, however. Once, again in the. Goal of making this kind of a load you know code solution, it's, not something that the customer has to work with has to worry about we, build, those integrations, with the partners to CRM providers, to, ensure that the kind of end to end solution. We provide is very turnkey. And so that if you have information, that's, relevant to the call within, the CRM system that, we can seamlessly pull, that into the contacts or application, and that your agent can have, access that information as a need or even in the context, of kind.
Of That virtual agent automation experience, you can bring in the information, as needed, so that you can voice it to the customer. Perfect. Thank you so. If. I think about what I've heard so far I might think well am. I going to need dedicated. Contact. Center developers, in order, to use Google. Contact center AI I. Don't, know if I have any if I have any I don't know if they know how to do this so. Do. I need dedicated, developers. But. Yeah. So our goal, when, we talk, about making, it a loaded no code solution, is to really ensure that customers don't, need developers, so. Most, of our contact, center contact. Area customers won't need developers, to get the solution up and running, our. Speech technology, works completely, out of the box both, on the speech-to-text and, the texas beach side and then. When we think about the virtual agent side of things our customers, that are having the most success. Bringing. Virtual, agents into their contact centers. Are investing, in conversational. Designers, rather, than developers, to design and, create, these these conversational. Experiences, and, many times those conversational. Designers are experienced. In high-performing, agents, because those are the people that are most, familiar with the array of questions that customers might have rather than developers, right, and. Then also for those customers that don't have the investment don't. Have. Conversational. Designers on staff we, also work. With other partners that can help out designing those conversational. Experiences, but. Then lastly thinking about the agent assist side there's. An integration there with knowledge bases to enable. Agents. To get the recommendations, and the suggestions, from our technology but, we actually are also working on knowledge, connectors so that that's also very plug-and-play, at. The end of the day we don't want developing. Development. To be a hindrance to kind of unlocking, this technology, for your contact center and improving that customer service experience so, we. Don't, see dedicated. Contact center developers as a need, to make this work. So. Question. How. How, do I decide which intense. That I want, to work on so. I don't, need a developer I might. Be able to do that with a business analyst in the contact, center. Does. Insight, help, me with understanding, what might be, the. Right kind of intense to work on. So. That's a good question so insights. That's one of the key components, of it really so we have a feature, that we call our topic, modeling where you can just give us a bunch of data whether it's chat transcripts, actual. Audio logs and from, that we can pull. Out one of the key things that customers are calling it about so then you can say okay well twenty percent, of the time they're calling the let us know that they want to change the address for a shipment or ten percent of their time they're, calling to figure out an account balance and. From that you can then identify what is the best use cases to start with building out these conversational, experiences, our goal and once again this is kind of an out of the box solution that that when you work with your contact center providers, you, can get, this insights, provided, to you so that you can understand, how should I start like, where should I spend my time with building out these experiences, what's, going to give me the most bang for my buck in terms of, kind of time, time. Given to to building this up. Good. Stuff. Good. Job handling these on-the-fly questions, I'm thinking of as you're speaking I. Like. A common. Hahaha. We've. The. Google contact center AI was first announced, in. The summer of 2018, and, I. Think we've talked about the fact that it is now generally available, all of the components, as. Of a couple of weeks ago so, what, else has been going on other than getting from you, know announcement, to general, availability. What. Has been happening at Google contact center AI for the last 18 months and and what does GA, really, mean in, Google. Parlance. So. For. The last year, and a half we've really been focusing, an Rd, and so we're reading from both sides of that both the research and the development so, from. A research perspective we. Realize that contact, centers are a relatively. New space for Google and so we have a lot to learn by the past year and a half we've. Done a significant. Amount of research actually visiting, contacts that are sitting. With agents, sitting with QA, and I was sitting with contact center managers to, better understand. How do you contact centers operate, and how can we build contacts, and REI to be most helpful for, these users and now, that we have a much better understanding of, the various keyhole keeps various, stakeholders. From, shift. Leads to QA analysts, agents themselves and so at the end of day this is what allows, us to make better decisions when, it comes to development, now.
We've, Spent a lot of time developing new. Features, and new functionality, specific. For the context in our over, the past year and a half as well so on the speech-to-text side, we've, been, developing some of the most advanced, deep learning neural. Network algorithms, and now we support, over 120. Different languages, we. Have our speech, to text so Cabul area is 10x, that of the Oxford Dictionary. We're also working on specific. Telephony. Models too in that can handle the challenges of the telephony context, when trying to do this transcription, we. We've had American, English for a while we just launched, British, English American. Spanish and Russian and we have many more to come now. The, other side of speech is the Texas speech and so, we want to ensure that in this context, in a context we are supporting. A natural, and conversational. Sounding, voice and so one. Of the areas we've been able to take advantage of here is google's, deepmind technology. That really provides, us those nearly. Human sounding, voices and. Now we have over 90, different, wavenet, voices, which is really the cutting edge in terms of natural sounding human human-like voices as well as, over 20 different languages, of the can support well. The, other side we think about is this is for, dialogue flow and so that's we're talking about before which is that natural language understanding and the real kind of conversational. Core that, powers, natural, and rich experiences. And so we've. Now. Support, over 30 different languages, for, dialogue, flow with more coming we, can support over 2000, or up to 2000, intents, to really meet the needs of large enterprises, enterprises. Ni VRS and, the proof is in the pudding that we're seeing over a million developers, now using. Dialogue flow for, their natural language, understanding, technology, and. Then once, again we spoke a bit about agent assist before but we've been doing the research to understand, what the asian experience is like and so now we've been testing some of these features that were developing for agents with live agents to understand, how can we improve, that and. Then lastly, within the context, of context. An area we've been working on these integrations, with our partners now we want to make that API, integrations, that we have with them much, easier to use identify, ways to make, this a better user experience, as. Well as ensuring that context, center center, at AI is, a place, where it can scale reliable. And you. Mentioned kind of what do we mean when we say GA what does that mean within Google and we. Recently went from beta to GA for context and area and the biggest difference there really is the, ability to reliably, scale and in handle production, workloads. I mentioned. Before that we have a significant. Amount of experience in the consumer side with a Google assistant people, search YouTube, kind, of scaling. Kind. Of machine learning basic, artificial, intelligence based products and so, being. Able to do this. Product at Google scale, is really where that what we need to achieve, to, get to the level that we can call ga and so a big part of that is the technical aspects like ensuring, that we can scale, up our servers, and handle tens of thousands, of concurrent, calls at one time and, then another part is a little more operational, right like ensuring. That we actually have the support team and the processes, in place to support, enterprise, customers, and, provide them the experience that they expect. So. That's interesting. I've, done some work with companies. As they. Develop, their own, applications. And one, of the issues they have is how. Do we scale this right, we've done this proof of concept and, it's you, know it's fine for the small number of calls or intense but, how do we make sure that we can take this into production as nice. To know that you know even before you go general. Availability google, has has taken those steps in, you. Know for their customers, I want. To take you back and ask you a question, about something that you said there Matt you, talked about one, of the things that Google has been doing for the last 18, months is, building. Telephony. Models, and you. Said, you. Know Russian, and a couple of others for the speech. Understanding. But. What is telephony, models, mean because you, can understand, the languages, but I think, there's something more there that the audience might be interested in.
Yes. So what do we say to Latino models we're talking about that speech, to text transcription, model that's, made, specifically. For the telephony context, the, reality, is when audio, is coming through as the telephony. Reading. In telephony pipes though, the quality is a little bit lower and the. Actual, the technical, aspects of it are different you have lower frequencies and so we. Need to ensure that we, can handle, the audio as it comes through a phone signal and, you can transcribe it correctly, and, there there is there's, both that aspect as well as the reality that in. The telephony context, you can have a pretty, long utterance, as opposed, to if. You're doing just the traditional have, IVR, context those other answers are much shorter, so you could think of, for. Instance if a customer, calls in and they're, talking to an agent there might be a lot of small talk there right they might be talking about their day they, might be talking about whatever, happens to be on their mind as well, as eventually, getting to the issue that they called in about and so, there are some unique challenges in, that context, to ensure that you're transcribing, it appropriately, and so, our goal. Is to ensure that we are bringing a quality, service no matter the, channel, no matter the context, and so to do that we're actually building. Out specific. Speech-to-text. Telephony. Models so we can ensure that we are providing, a quality transcription, as it, kind of is the, first step to doing some of the natural language understanding on, the virtual agent side as well, as providing, great. Suggestions, to your agents for the agency side. Excellent. I think that that helps us all understand, you. Know the other thing I think about is you know cell phones and that kind of thing that you're purposely, working, on lower quality. Voice. Dreams that you might be getting because I think you know I was, always I was surprised to hear from one of your partners actually Avaya. That. Over 50%, of. Telephone. Lines in homes today in the United States are, mobile. Right, so, it's not an insignificant, issue, that, contact, center calls are coming in on mobile phones so being, able to handle that that. Quality of speech that quality of recording, so. That you can then do the rest of the work that Google contact, center AI wants to be able to do well, makes. Sense but now probably, the best one, is ID in your deck this is it this is this is the money slide. You're. Going to tell us about three of us tumors. That. Have actually deployed, Google. Contacts in our AI pre, general, availability and, I, think some of these even spoke at your, event in London last week. Yeah. Well so we actually had Marks and Spencer on stage. Getting with us last, week and so one, of the examples. That I touched on briefly before is Marks & Spencer is a retailer, in the UK. And they are now, successfully. Routing over 6 million plus calls. And so, one. Of the benefits there is they can have a high number of intent so using all close to 2,000 intents and being, accurately, routed to the right agent, as opposed you're having a human. Sit at that part and just try to route all day right you could think of if you're a, customer, service agent it's much more appealing. To work on interesting, calls as opposed to just routing, taking. The first call and trying to route it to another, agent all day now and from, a, kind.
Of Cost savings, and kind of customer, experience perspective, they're seeing an. Improvement of over 10 seconds, in average handling time all. Right so the proof is really in the pudding there another. Example, it's Policy bazaar where they are seeing actual, top-line. Growth where, they've had I've been able to close sales over. 13,000. Sales actually and then over 2 million, dollars. In annual monthly, revenue and so this is an interesting case where it's not so much how, do we make the business more efficient, but how do we actually improve, our top-line and provide. Better recommendations. And a better experience to our customers that actually increases. That kind of brand loyalty which. Results. In gonna you can get manifests rather and. How much they're spending with us and then, lastly its Woolworths which is a, Australian. Retailer. That's doing something very similar, to Marks & Spencer where, they are using a conversation, where they are to, provide an improved routing experience, for. Their customers. And. So, it with Marks & Spencer, and Woolworths I know what. Is policy, Bazar do. The. Policy, Bazar is a company that helps, you identify what's, the right, insurance. Policy, effectively, right and so you could think, about like insurance can be a pretty complicated, space. And so what they've been able to use. The. Contact. Center AI technology, to do is to provide better, recommendations. Over the phone to, their, customers, and so they're, providing, a new channel to get recommendations, which in may a somewhat. Complicated, space. Kind. Of easier, to, handle, and kind of understand, because you can whittle it down to a couple of options that are good for the customer and then deliver, those options to the customer and a conversational. And kind of natural, feeling, context. Excellent. So you know it looks like there's business, to consumer all. Happening, here but I'm sure there are probably some business to business. Applications. As well but, maybe that will save those for the next webinar because, now we're going to get Eric involved, again, and he's. Going to run a poll, question for us. Okay. Thanks thanks very much you and here's our question do. You see applications. In your business for the kinds of projects, that were deployed at Marks & Spencer Policy, Bazaar and Woolworths, and. Options. Are yes we've already deployed, virtual agents we're, prototyping virtual. Agents, we, don't use, them today but we're considering them and no, we don't have any plans and maybe. You. Know maybe just, just real, quickly Matt if you could just reduce, as. A, reminder, for people how. You might characterize in a couple of words each of those three. Projects. Marks. & Spencer was about sort. Of call handling efficiencies. I fair. So. I would, say Marks and Spencer and Wohlers, were about call handling efficiency for, retailer, policy. Bazar is actually about providing recommendations. To. Customers. In a conversational, context. And. Those. It how did how did policy, Bazaar sort. Of validate. The. Savings, that that they that. They had. So. In policy, bazaars case it wasn't savings, there's actually sales and so they could see that how. What, sales are actually coming from a conversation that's. Involving. This conversational. Experience, and so they could directly say that X percent of sales were, a function of the conversational. Experience. Got. It okay, well here's here's all our results, with that I'll just hand it back over to Sheila. You want to talk with Matt and Celeste about the results or just move on. But thanks everyone for helping i'm. So. Not, surprised, to, see that over 50% of, you. Haven't. Yet deployed, virtual, agents but, are considering it and good. On you you're, here learning, what google contact center i can possibly do, to be part of that future. 22%. Saying they have no immediate, plans and that's. Okay too because we did ask the question about you, know virtual, agents, there and. Some. Companies are saying you, know what i like, the thought of using google contacts, and rei to assist, my agents, you. Know the way Marks and Spencer did by allowing call, disposition. It. Helps, the agents save 10 seconds, on average and alcohol but. We're, not really ready to do, agents. And. Almost, 10 percent say they have already deployed. Virtual agents which you, know a year. Ago number. One I don't think the number would have been as high and. Number. Two I'm not surprised. At that and, just, today I was doing something about something I ordered and I got, a virtual assistant from a company and then I went.
To The shipping company and darn I got another virtual, assistant, in chat so, it's, definitely happening. Do. You have any comments, Matt on the prototyping. And trialing. We've, got 16%. Of people saying that you talked, a little bit about how people transition. From that to. Deploying. Yeah. So as, we mentioned before a, little, bit at least but we think see topic modeling as an opportunity, to identify what. Are some of the key use cases you can use then, once you start to prototype in trial you get a little bit more experience, with how do I build out these, conversational. Experiences. Right you start to develop a bit of that conversational. Designer, competency. Within, all, your, contact center and so what, we recommend is once you've kind of gotten past this prototyping. Phase you can start to once, you have confidence in maybe one key intent where a couple of can test you can deploy it in the contact center and then, using topic modeling to understand. The other key. Things that your customers are calling out calling about you can start to kind of add on additional intents. There as well, as understanding, are there certain phrases, that are getting, confused that the virtual aids getting confused about and improve. Kind. Of the accuracy. Of your virtual agent at the same time so, I like, to think that you start with a prototype, once you have confidence in the prototype you expand. The amount of traffic it's getting and then you can start to add in additional, intents while, you're using topic, modeling to understand, what are the, topics that people are calling about. Yeah. It's interesting what you say I remember, once that. This notion that you, don't have to guess. For. Those first couple of applications, of app. Artificial. Intelligence, all you have to do is ask agents, and supervisors, they, will tell you what the most often. Questions. Are you, know what one of the things that we end up spending time on that our could, easily be done by a, virtual assistant but. Then when you want to get more sophisticated and. It's. The tools like. Insights, that are going to help you take it further, good. Stuff. Okay. So we've. Talked about what you've been doing for the last 18, months Matt, you. Would I guess there's another couple of guys better, work on us. That. Google has as you, look out over the next year, or two from contact, center AI. Yes. So we're planning on expanding across all of those different areas that, I've touched on today so one. Of our immediate, key focuses, is expanding. The languages, that we can serve with our telephony. Speech models and so this. Is one of the key components. To enabling, context. In our AI for our customers, globally, currently. We support American. And UK English, American, Spanish and Russian and our, plan is to rapidly, build. Out additional, support. For other languages in the next year or two now. 51, with England, virtual agent side we're focused, on enabling. Enterprises. To handle the scale and complexity, of a contact center why look flow already supports, five, times the scale of our nearest competitor. In terms of number, of intents, and is being used in highly, complex use cases now, we want to double down on that strength and accelerate, enterprise. Adoption and so this, predominantly, will mean features, that better serve the enterprise you can think of things, such as better tools to build BOTS tools. To understand, where BOTS get confused, and the ability to handle more intense there's, some of the things that we're thinking about and working on currently now. On. The agent assist side we're just getting started so there will be a ton to learn in the next year our goal, in general, is to help agents be more effective, and efficient. By reducing, some of the more mundane tasks, they do and, allow them to focus on providing the human touch in customer, service we're. Currently working on a number of feature improvements, to our knowledge, article, and FAQ suggestions, to make them more relevant workflow. Suggestions, it easier fulfillment, to help agents, more easily and quickly solve, issues for customers we're. Looking at improved check tools like smart reply for, suggested. Responses and, smart, compose for our auto completion and, we have a number of other features that are a little bit earlier and they are in these days which, will crystallize, as we start to learn more right I think this, is one of the areas where we have a lot to learn as we have more adoption, and to understand. What's, really moving the needle for customers, and then. Lastly we'll also be go deeper into insights for contact center area topic. Modeling was kind of our first product, within this space which enables the automation of understanding, what, topics are being discussed in the call but. We're also working on more advanced call analysis, tools to help contact, center managers QA, analysts, identify what our specific areas, of interest fulness, within, these calls how, can we help them perform, more.
Efficient, That quality assurance process as well, as quickly identify, opportunities. To. Actually. Coach your agents identify what, are things that the agents are struggling with and you can help them out proactively. Excellent. Looks like you've got a busy couple of years ahead of you and that, we'll be seeing the, benefits event as we go on so. I am going to turn it back to Eric, who's. Going to. Handle. The QA. All. Right sounds great thanks very much Sheila and also Carson thanks Matt and Celeste was terrific really interesting counter for. QA just. Reminding everyone you've. Got the, text box at the right hand of the presentation, window where you can type in your question, or click, the Q&A icon at the bottom of your screen and. Let's. Start with. A. Question. On. Here's. One how this, let's, just kind of maybe start with. With. The Google team and then maybe. Get any perspective. You have Sheila well how will the, verification. And identification, process, work, with. Virtual, AI, agents. Basic. Search so for. Virtual. Agents when come, to AI agents, we have this, feature, called fulfillment. Which allows us to. Connect. With your back-end systems, and process. Any type of virtual or, verification, or, authorization process, that you might need another. Thing that we're actually starting to work with voice. Biometric, firms as well to start to provide another, capacity, for that verification or, authorization process, what we have a number of customers that are using it today. And we're seeing it this is very. Helpful because for instance if it needs to get to the human, age it's it's, ideal. If the customers, already been authorized, or verified, and you know who they are so we've, kind of developed, this with that use case in mind so we can easily support it. The. Other key. Lady any thoughts on I was just going to say that you know since I didn't open my big mouth I really don't have anything to add on that one. That. I think is kind. Of thing it's not a lot of people's minds and it is for the Google, team what. Does Google do with my data. That's a great question so, your. Data is your, grade of Google and no one else's so we are very, very serious about how we handle your data, no. Other, customers, will have access to your data it, is and share for training models at the. End the day it's imperative, from our perspective, that any. Data you provide to us is only use for. Models. Within your company and for your customers, as. Well as if you send us any data we also do, what we call DLP. Which just removes any forms, of personally, identifiable information. So. We ensure that we are not storing anything. That. Might be of concern to you and as well as anything. That you store with us will, only be accessible to you and no other customers. So. I. Think, what you talked about there that DLP. Process, is is what some of us may know is anonymizing. The. Data. So. What's the difference you know what a sort of follow-up. Question, here Matt what's, the difference between. Securing. Our data if, I'm one, of your enterprise customers, and, using. The data to make better models, so. If if. I'm, your customer Maggie, Smith analytics, is your customer, you're. Securing my data your anonymizing, it but do you do. You use it to create models, that might be used by other people. Other. Companies, very. Question fair question Sheila so we. Both, secure, the data so anonymize, it as you said but we also ensure that we do not use your data for anyone elses models so, any, models. That you, train. Or you build using, your data will only be used for your, use case and for your company it's, one of the the key components, as we see it is that we don't want there to be any data sharing amongst, customers, data is a very, valuable.
Asset To your company and it should only be used for your benefit and so we. Ensure, that there's, no data. Leakage no data sharing if you give us data on for. Your, use case it will only be used for your use case it will not be used for any other customers, models. Excellent. Good to know thank. You. Okay. Thanks very much. Matt, and. Let's. Go to another question. We. Sheila asked, earlier about whether, whether, an. Enterprise need to hire developers, and Matt you indicated, that that, you really don't but, but AI itself, is kind of a whole new. Sort. Of field of expertise, and discipline, and. Is. The idea here that I can just they. Essentially, outsource, everything. Dealing with AI to Google. And and, my enterprise folks, can worry about what, they've always done or do, I need to hire people who know something about AI and, and who. Are, those people if they even exist so. I'm, going to start with this one because I have. Been working with companies that are, you know thinking, about these issues right, and I. Think the good news is that the. Graduates, coming out of. Business. School's engineering. Schools. Are. Taking. These kinds of classes they, are beginning, to understand, and in coming out with skills, that, might. Not be in your contact, center today but. Will be useful, not. Because you need them to build the AI but. You need them to understand. How, it will work in your business, right, this, you know Google is doing. Really, good work to, understand, the contact center market working with all the contact center vendors but, there's still vertical. Aspects. Of. Artificial. Intelligence meaning, you know the various industries, and so. I think. You do want to think about hiring people, who. May have coursework. Or former. Experience. In the, area, of artificial, intelligence without. Necessarily needing. Them to you know be coding, all day long that's. The sort of Sheila's point of view what curious. What yours is met I. Would. Agree, and of double down on what you said there Sheila I think. One of the goals that we had in mind is you don't need AI developers. Now you don't need machine, learning engineers, to make this work but, you should have folks, who understand. How a AI works at a high level right, that they understand that the, as, the. Quality, of the model is dependent, on the data that you provide that, there's, going to be a training process sometimes, to ensure that you can create, a new model but, these are more high-level, AI concepts, that they can that can be applied broadly as well, as your employees that have a deep understanding of, your key use case so it's, really not AI technical. Knowledge that, I would say you want to hire but just kind of more of the AI general, knowledge and then the deep domain, knowledge of your business so they can understand, where that intersection, is and how to build kind.
Of The contact, center experience appropriately. Okay. Great and, we've got time for one more question I mind we got a ton of questions here which is awesome and, we're. Not going to get to them all but we'll. Have all these questions. Collected, and we'll. Have them in the in the report that our system will generate for us so the. Folks, you've heard from today will be able to get back to one on one with. The information on. Your contact, that, you gave us when you register, and. They'll be able to help you with with, answers after. The event so. Let me close with this last question, and. It's a small. Question if, I was going to start an AI project, how. Would I begin. So. I'm going to start with this and then obviously Matt will, join in hopefully I. Think. If you go back to one of our earlier slides and you, looked at all, of the vendors that Google, contact center AI has, worked with contact, center vendors you. Will probably see, your contact, center solution vendor, there that. Means that they've already been working for, you, know 12 to 18 months with Google, and. They're down a learning, curve and so. You can start with. The people that you work with every day, in. Terms of how do i integrate, that to what, I already have, today so. I think that's the, best place to start is right with your existing, contacts, vendor what. Do you think Matt. Yeah. I would agree the goal. Here is to not, introduce. Too much additional complexity, right so your, contact, center. Software. Provider has been working with us they have an understanding of specifically. How can you use their, software, kind of their contact and application, within the context. Of contact, center AI so they, will be the subject matter expert to get you kind, of off the ground day, one and. Then using the other components, that we talked about right so using topic, modeling to understand, what are some of the key use cases that we can use as well as as Sheila mentioned you have a number of subject. Matter experts in your, context in or that's your agents your contact center managers, and. Then you can identify what, are some good use cases to start with and work with your contacts in a provider to actually kind of get that off the ground and running. All. Right terrific I'm gonna go to that's a great note, to end on as. You can see I've pushed out a slide, with some resources, a, bunch, of blog. Posts and some other information. That. Google's provided, for us so please. Do, take advantage of the, opportunity, to learn more about this topic it's a really rich topic and as you can see it's one that's going, to be just, increasingly. More important, over time we'll be covering it extensively, at Enterprise Connect and. You'll be reading about it everywhere, including no, jitter Sheila covers it extensively. And and Google. Obviously is a, huge piece of this, so. I want to thank Google for sponsoring, as well as thank everybody in, the audience for, listening, in today we. Really appreciate your, attention and your participation, now. Within the next 24, hours you're going to get a personalized. Follow up email from us with, details. And a link to today's presentation on, demand and and you, know since we had such a great response. I'm, sure there's more people out there who would love to hear this. Session, and listen to the replay so please. Do forward that along to any, any. Of your colleagues that that you think might also. And guess get something out of listening to this. The. Webinar is copyright, by informa. How, the presentation, materials are owned by or copyrighted, by Enterprise. Connect in Google and, the individual, speakers, are solely responsible for, their content, and their opinions, on. Behalf of our speakers Sheila Matt and Celeste I'm. Eric Raph thanks, for your time and have a great day.
2019-12-13