IVA is the New IVR
Chad McDaniel: Dan, if we could go to the next slide, and really want to welcome you to what we want to talk about the IVA, Intelligent Virtual Assistants. Some of you may have heard this concept or have not heard of the concept, but we’re going to clarify a lot of things around the IVA, and this being the new IVR. So, be ready for that. Be ready for a lot of good case studies and other insights that we’re going to be getting through today’s webinar. But I really would like to thank our partner, SmartAction who has some incredible insights in this topic, and I’ll introduce our speakers here shortly.
If we could go to the next slide. Just some quick housekeeping here. This presentation, everything you’ll see in the slides today will be made available after the webinar is over. Tomorrow, you’ll get an email link sent to you from WebEx, which will include the recording and also include the slides. I would encourage you if you like the session, please share this internally. Really, Execs In The Know is a sharing community, and we’re trying to help the education set in advance, the learning as we as all leaders in this space.
As soon as we conclude today’s webinar, there will be a short Q&A. And then also, after the webinar, you’re going to get a link from WebEx that will ask you to complete a survey to how you felt about the value of the content and the discussion. If you could just take a minute at the end of the session to complete that survey your feedback is very important to us. If we could go quickly to the next slide. For those that may not be familiar with Execs In The Know. Execs In The Know is a very large corporate and user sharing community we call leaders learning from leaders. Many ways for you to get involved in the sharing community. If you just go to our website execsintheknow.com, you can see all of our resources, our content, our publications, our research, our events. All of this is available for you. I would encourage you not to do this alone, get involved. There are many people wanting to participate in the sharing economy.
In the next slide if we could go to, if you happen or able to join us for our corporate listeners, our next leadership gathering. We’ve got a tremendous group coming out and great keynotes and sessions. A two day powwow in Hollywood, Florida, just north of Miami. If you’ve not been to one of the leadership events, I would absolutely encourage you to come out. They’re very impactful, very educational, and learning, and great networking along the way. On the next slide, if we could move on. Also, want to make aware of our private online community called Know It All. This is a persistent community of 24/7 where we’re sharing the most active and direct, most detailed type of conversations around this customer success journey, great conversations, great insights. I’d love to get you involved in the online community. You can send a note to Chad at execsintheknow.com, and we’ll get more information from you.
All right. Enough about Execs In The Know, I want to move into today’s webinar. And Brian and Dan, thank you so much. You guys are tremendous industry resources. You’ve been doing a number of great things, I think for the industry over the years. I’ve had the chance to work with you in different variations over the years. But I’ve always been impressed honestly, with the subject matter expertise, and the continued evolution that we’ve seen in automation and you guys’ thoughts around it. So, Brian, maybe we could start with you. Just a quick introduction about the company, your role, and then Dan, and let’s jump right into the topic.
Brian Morin: Sure. Thanks for that hand off, Chad. We certainly love working with Execs In The Know, and appreciate the time that we have to interact with your audience. I’m the head of marketing here at SmartAction, and with me is my partner in crime, Dan Fox. He’s the VP of Product Marketing and Strategy over here. And just to set the table for you, we operate conversational AI solutions for more than 100 brands for both in the large enterprise, either Fortune 500, and also in the mid-market. And so, it’s put us in this unique position, this vantage point where we’ve seen what works and we’ve also seen what doesn’t work.
And so, we’ve been part of this trend and really watching this whole shift from the early adopters to now what we’ve called the early majority. Now, I do want to mention outright that today is absolutely not meant to be a SmartAction commercial although we are a participating vendor in the space. Although you may have to forgive a shameless plug from us here and there. To add on that, we just actually had a customer of ours purchasing power, big e-commerce play. They won the Excellence and Self Service Award at the last Frost & Sullivan Contact Center Awards last month.
They replaced their IVR with AI powered virtual agents, which we see about 80% of our customers doing. We’ll also talk about the other 20% that actually augment, and we’ll mention that in a minute. But in their case, they were automating 25 to 30% of their call volume. They saw NPS rise 17%. But the big trigger on winning the award above and beyond the CX. They had to do with the fact of achieving ROI within the first three months. So that’s really the emerging story here in this new climate with conversational AI. It’s not just the maturity of the technology that your rivals and exceeds that of live agents when used, I underscore, when used in the right way, which we’ll talk in a minute.
What we’re seeing is organizations actually moving straight out of their opex into these solutions, and where they’re seeing their savings is just through the course of their natural churn of FTEs and not needing to replace that until they’ve right sized. And so, we see this now happening all the time with organizations where five years ago, people wonder, wow, how do we even get the ROI out of a voice automation package that’s then omnichannel. And now it’s a very different story.
So, Dan, if you just move on to the next slide. Just the only need to know about us on the screen is that we provide AI powered virtual agents as a service. It means that we bundle the AI technology together with end to end CX services, and do it as a single subscription charge. It’s omnichannel, most start with voice first and then they’ll scale the said application digitally chat and text. We see that with other vendors in the space, not just ourselves. And lastly, if you’re interested in any of our street cred you can look us up on Gartner Peer Insights where we happen to be the top rated solution there.
Dan, we’ll just go ahead and jump right into this. We’re talking about how IVA is the new IVR, and IVA meaning Intelligent Virtual Assistants, and why are we calling it the new IVR? Well, conversational virtual assistants, they first had their start sitting behind an IVR. When we were first doing this, nearly 100% of customers would put the virtual agent behind the IVR. Someone might press two for said claims, but instead of being transferred to a live agent, then that would be transferred to our network where it would be a service by a virtual agent.
So, more often than not though, what we are seeing is this trend or shift of our own customers moving away from that model, and instead the very first interaction is actually with a cloud based virtual agent who greets in natural language. How can I help you today? Captures intent, authenticates and routes to the appropriate rep. Or if it’s an intent that should be automated, and help the customer self-service, then it will just stay with the virtual agent.
In this example that you’re seeing on screen, this uses the virtual agent to go a step beyond what I mentioned with algorithms that can quickly scan all data points you have on the customer and actually predict why you might be calling. And so, more and more of our own customers and those that we see with others is that they’re implementing this, which really cuts down the friction or customer effort to get the job done. It also sets the tone on how you might interact with the system when you realize that it’s this intelligent knowing something about you. So as you can see, this approach just bypasses that kind of traditional IVR architecture, which we will show here just in a minute, a couple reference architectures, Dan.
Dan Fox: Yeah, sure. So, jumping in here, Gartner created a statement about that by 2020, 85% of a brand’s interactions with its customers is going to be conversational in nature, but not with a human. So I think this is incredible, and whether it proves out to be true or not, what we’re seeing in the industry is a giant shift over the past 10 years from people who call up customer care, and they really just want to talk to a person. And these are the days of bad IVRs and really ineffective chat bots. People are calling a company and they’re like, “How can I get through this automated system as quickly as possible so I can talk to a live agent who can fix my problems?”
What we’re seeing now is a total shift where enterprises who aren’t leveraging automation are really shorting their customers of what could be a really intricate, delightful customer experience by picking up the phone, calling a company, and any of those issues that are typically handled by live agents that don’t really add a lot of value for live agents to handle can just be handled completely in automation, which means you get your issue solved, it happens quickly, it’s data contextualized, and then it gets on your way. Rather than having to go through a bad IVR experience, holding for an agent, establishing a rapport with that agent, and then going back and forth. Before you know it every customer service call with a live agent, maybe five minutes, six minutes. What we’re talking about here are these conversational customer experiences, which really you could get resolved in a minute, two minutes. So, what we’re seeing is a huge shift into this arena of conversational AI being an actual preference on behalf of customers.
To set the stage what we wanted to do is play a call example from one of our customers TechStyle. TechStyle is the parent company for many brands who you may have heard of, JustFab, Fabletics, FabKids. They’re one of the largest subscription retailers, which means you can pay $50 a month for their service. And every month, you’ll get a box delivered to your door with a bunch of clothes that you can have for that month. There’s a lot of customer service calls that they get that they decided they didn’t want live agents to handle because, yes, they were complex. But do my agents really benefit from figuring out simple billing questions? If someone wants to skip a month, why can’t we let them do that and automation? If someone wants to know where their order is, can we just automate that? So, what we’ll play for you is an example of one of those calls that they chose to automate.
IVA: Hello, thank you for Fabletics. How can I help you today?
Sarah Peters: Billing.
IVA: I’d be happy to help you with your billing question. Is the name on the account, Sarah Peters?
Sarah Peters: Yes.
IVA: Are you calling about the $49 and 95 cent charge billed on or around July 19th?
Sarah Peters: Yes.
IVA: Okay, what’s your question regarding this charge?
Sarah Peters: I skipped the month. I should not have received this charge.
IVA: Looks like you were charged because you did not make a purchase or skip by the fifth of the month. This charge is added to your account as a member credit that you can use at any time. Would you like to keep the credit in your account?
Sarah Peters: No, I’d like to be refunded the money as I skipped the month on July 3rd.
IVA: Okay, no problem. Your refund has been successfully processed. Please note that it may take up to seven business days for your bank account to reflect the change. Is that everything for today?
Sarah Peters: Yeah, thank you. I appreciate that.
IVA: Thank you for your call and have a great day.
Sarah Peters: Bye, bye.
Dan Fox: I love playing that call because it really goes to show at the end of the call, she gets what she needs and she says, “I appreciate that,” to an automated system and how many people say that to a robot. And then on top of that, she says bye, bye. It just really goes to show the customer satisfaction that TechStyle is getting out of this, which for them is actually higher than their live agents. So their live agents have a CSET of 90%, and their virtual agent has a CSET of 91%.
The reason I think this is true is because that customer she had an issue that she felt she needed to talk to a person. She got greeted by the virtual agent. She explained her issue in complete natural language to the point where she’s like, “Yeah, I thought I was… I got billed, but I skipped a month by the certain time, and the system was able to understand that.” So, that full call was a minute and 30 seconds. It probably would have been a six minute call with an agent not counting whatever IVR is in the process. So when she’s saying, “I appreciate that.” I think she’s actually thanking TechStyle for creating an experience where she can really get what she needs in an efficient manner.
Chad McDaniel: Chad here, Dan let me jump in real quick before we get to this slide, maybe go back, and sorry. I know we have a lot more to cover here. But Gartner predicted you mentioned by 2020, which is now that 85% conversational AI versus human which to me is very aggressive. I think we’re leaning that way, but clearly in my opinion not 85%. But my question quickly as we look at the shifting of conversational AI, the virtual agents to live agents. Where is the line from your opinion, you and Brian’s opinion? We’re going to get more into the technology structure and all this stuff, but where is that line that you’re seeing?
Dan Fox: So, the line used to be predicted by the technology capabilities. I think we’ll talk about this throughout this presentation, but what we’re seeing is when you dealt with an IVR system, it was like, “Okay, well, that billing issue, I don’t know how I’d handle that with speech recognition and natural language processing. I don’t have the right data integrations. So why would I automate that conversation and ruin the customer experience? Now, within the last three years, the technology has gotten so good. The tools that we’re able to use to create these conversations, and the design expertise that we’ve garnered over those years has really gotten to a point where we’re not limited by a lot of capabilities of technology. We’re actually able to look at a conversation and say, “Should I pass this to automation or should I leave this with my live agents?” Because you have to disambiguate between what’s a complex issue, and what’s a human only issue.
A good example is that billing inquiry, right? So, for TechStyle I’m not even sure if that customer was telling the truth, right? Did she actually skip her month by the third day within that month? I’m not sure. But TechStyle made a strategic decision that said, “Hey, if this woman wanted to skip the month, just let her skip the month because it’s more cost effective for us to handle that in a virtual agent than go back and forth, and risk our customer experience to have that call be handled by a live agent.” The interesting question and that line I think is unique for every organization is figuring out what are my human only issues? What do I never want to automate? Because that’s really where I can make impact with a human and what are those issues that are complex, but technology is now capable of handling.
Chad McDaniel: Perfect, thank you and great example on that call. Brian, quick question to you for clarification for our audience. Conversational AI is not a product, it’s an ongoing effort evolution or how would you best describe that before we move on?
Brian Morin: Sure. As a matter of fact, Chad, we had that preliminary discussion, and so when we did we actually put a slide here in that deck to discover that exact topic. It’s just-
Chad McDaniel: Okay. We’ll get to it.
Brian Morin: We’ll get into it, but the net is that I think that we’ve all lived in a world where IVR is it’s you design, build, and done. And in its very best day was its first day. But when you enter into the world of conversational AI, this is a world in which you have ongoing operation daily. You’re care and feeding where you walk through this process of perpetual improvement because at the end of the day you’re having conversations and having conversations is a very different piece. And this is partly what has to do with the evolving role that contact center leaders have now in their contact center. Because now it’s a matter of not just being a manager of human capital, but also being a manager of AI and how do you define those symbiotic relationships that we’ll explain more?
Chad McDaniel: Okay, perfect. We’ll move on. Sorry for my interruptions.
Dan Fox: All right. So, when we’re looking at that TechStyle example, what we want to do is set the backdrop for when we’re talking about conversational AI, really, what are we addressing here? For us, it means six things and you’ll see that in the in the circle here. The first which is the most essential as far as I see it, making sure that you’re connected to the same data as those live agents. That’s where I think a lot of the problems have surfaced. That’s why I think we’re still seeing so many virtual agents and chat bots that still don’t have a great customer experience because they’re not deeply integrated. Because for that, and that goes into predicting why customers are calling.
So, in that TechStyle example, we greeted customer by name. So we said, “Hey, is this Sara Peters?” Rather than asking her for an account number, and we did that by recognizing her phone number, and then dipping into TechStyle’s customer relations management to say, “Okay, that phone number is associated with this customer. Greet her by her name, Sarah.” And then we also knew that it was a billing question. We knew what bill she was talking about. We knew what month she was talking about. And that was able to drive that conversation. If we had to collect that information, it would not have been a minute and 30 second conversation, it would have been closer to a four minute, five minute conversation.
On top of that, navigating complex multi turn conversations. So don’t just deliver a bunch of information, have rambling long prompts, which I’m sure everyone’s familiar with. Keep it short, keep it simple, ask questions, and make it interactive just like you would with one of your live agents. If for any reason that conversation failed or Sarah didn’t want to work with that automated system, with the virtual agent, get that right over to an agent. Our job isn’t to trap customers in automation. I think like in the IVR industry, we talk about containment a lot, and I hate that word. You’re talking about containing customers, getting them away from your live agents. You shouldn’t be afraid of having calls go to your live agents. You should use self-service as much as possible. And make sure if someone doesn’t want self-service to get them over to an agent.
Omnichannel, so that experience that you heard is when you call TechStyle, if you chat with them, you’re going to get the same experience. So making sure that you have a framework to deliver these virtual agents in an omnichannel framework, so you’re not deploying siloed customer care experiences. Of course, recognition and cognition being the last thing, which is really being able to understand the granular intent of why someone’s calling, and then being able to extract any relevant information. So, understanding the true intent of what people say and then creating action to take it to the next step.
Chad McDaniel: All right.
Brian Morin: Yeah. Thanks, Dan. I do want to point out that as we go along here, please just chime in your question into the Q&A box and Chad will tackle some of those either as we go along, or when we move to a full Q&A here at the bottom of the hour. This piece is actually one of the most important visuals. It’s frustrating as a touchstone experience can be sitting and listening to a lengthy phone tree menu, of course, which we all hate. It isn’t the worst part. The worst part has to be the fact that it offers you almost zero opportunity to improve efficiency and lower costs.
Since an IVR or simple chat bot can only handle the simplest forms of self service, it means your customer service reps are handling almost everything, including all those simple interactions that are highly repetitive and transactional in nature. And that is what is creating that very mundane employee experience. But we were first working with Electrolux, you might identify this. Their live agents are handling 35 or more product registration or warranty calls a day. And the work was so mind numbing that the agents, while they’re sitting there talking to somebody on the phone, they’re actually turning a page in a magazine while they do it. They almost felt like robots on the call doing the job.
Whenever, you step out into the world of conversational AI on the right hand side of the screen. On that right hand side of the screen, it’s not just a matter of automating the simple linear transactional calls. It’s as Dan alluded to, you can actually begin automating some of those more complex interactions as well. And so, this is where that symbiotic relationship that I alluded to earlier between virtual agents and live agents, and how this really impacts you and your future role now and into the next five years. This is where things really begin to emerge in a very different way.
We automate, and I’ll explain in a couple of ways. We automate conversations between body shops and State Farm to improve work. And there are 17 back and forth turns in that conversation. It involves things like capturing long alphanumeric policy numbers as fast as someone can say them. I mean, that’s hard for a human to do. Of course, impossible for an IVR. But as you can imagine, and Chad, this gets to even a little bit of what you asked Dan. When you have 17 back and forth turns at a conversation that presents a lot of different forks in the road for that conversation to go. And the question is, is the virtual agent going to handle every single exception that can occur in that call? And the answer is no. The fact is it’s never designed to handle every exception.
What we do and in others like us is that we design swim lanes for the virtual agent that follows the happy path. You can consider that the widest path that most customers take in a given interaction where you can guarantee a great CX with a virtual agent. Those other exceptions, frankly, it just isn’t worth the effort to automate every possible exception that could occur or in some cases, those exceptions require perhaps access to data that isn’t available.
And so, here’s where that symbiotic relationship emerges because any call that goes outside that happy path, that wide swim lane designed for the virtual agent, then all those calls will get transferred to a live agent, along with data gathered up to that point, so they can just finish the call. And so, what that means is that you can be sure AI isn’t automating everything. It’s just automating the widest path that most callers take, and then transfer exceptions to live agents. So now humans are handling what only humans should handle. Those instances where judgment or critical thinking is required.
Just a couple other just to give you other examples of how that looks for Choice Hotels. They’re the umbrella brand for almost every budget hotel out there. And we don’t actually book the reservation. But we do ask you the five or so questions that every agent would have to ask. Things like name, your phone number, location, date, number of occupants, and then after doing almost all of the call, well then we pass it to the live agent to confirm and book.
Another good example of this would be Penske Truck Rentals. We handle all their outbound reminders on reservations, but it’s not just a dumb reminder. The person who receives that reminder has the flexibility to change the reservation. And so, the virtual agent can assist them with that down the widest path. We can help them change it to a different time, which is part of the wide path. We can help change it to a different day that’s part of the wide path. We can help them change to a different city that’s part of the wide path. But if we notice that the truck is not available at that different time or different city? Well, we know that it just went outside the swim lane. And so, now it gets transferred to a live agent as an exception.
I hope that answered a little bit of that question that you were asking before, Chad, about where do you define of what a virtual agent should handle and what a live agent should handle? And here’s how your role as a contact center leader changes is because now, a big part of your role isn’t managing human capital. It’s now managing the symbiotic relationship that happens between your AI resource and your human capital.
Chad McDaniel: Brian, good example. Yes, it does. There’s been a few number of questions that have come in generated over the last couple of slides. We could try to tackle a few of them while we’re at the moment.
Brian Morin: Sure. Let’s do it.
Chad McDaniel: Again, I just want to be mindful of time. Jack, thank you for the listening. Good to hear from you. I know one of your questions, Jack is how can we use this for non-voice support? Any ideas, suggestions there, Brian or Dan?
Brian Morin: Yeah. It’s the same. We were just pointing out here and maybe talking kind of in voice first just because most customers go down the voice first path. It’s for no other reason that, that’s just where the greatest ROI is. But as soon as you have that application customized for that interaction, well, then you can scale digitally. However, we have some customers that do it the other way. They will start that digitally and then scale out to voice.
Chad McDaniel: Thank you. Bradley, got your question here. I’m going to try to answer and hopefully this is the same answer with you, Brian and Dan, but what percent of clients are notifying callers they’re speaking with AI? For me personally, I hope it’s 100%. But what’s your thoughts on that?
Brian Morin: Yeah, Dan.
Dan Fox: The idea is to never try to confuse a customer or tell them that they’re not working with an automated system whether it is text or voice based. That deception just leaves a really bad taste in folks’ mouths. So you can either say something like, “Hey, how can I help you? I’m a virtual agent who can help you answer any questions.” Because you want to front end that you are talking to an artificial intelligence system because that also directs the way that people talk. If it said, “How can I help you?” And people thought it was a human they’d explain their issue in a completely open ended manner the way that they would a live agent.
That may take away from what you’re able to understand because natural language processing no matter how great it is, is really good at understanding a simple sentence. No matter how many pieces of information are in that sentence, keep it a sentence. If it’s four or five sentences, you’re going to have a tough time extracting what that intent is. So, never deceive the customer. Always, always tell them that it’s an automated system. And that’s really where you’re going to get the best experience.
Chad McDaniel: Yeah, I would say… Chad here, is that there’s unanimous consent amongst the corporate brands to disclose, don’t hide.
Brian Morin: Well, Chad, here’s the here’s the funny thing though on our end. For us, the disclosure piece isn’t the most important piece for us. It’s you have to let them know who they’re talking to. Actually, if we don’t disclose these systems are now so human sounding that as Dan mentioned they’ll go into a five minute diatribe on their problem.
Chad McDaniel: Yeah, that’s good point. One last quick question. I know we got a lot more slides to go through. Thank you everyone for your input. Caleb’s question was one of these two slides we’re on. How do you manage the IVA calls? Are they recorded like regular agents? If a manager notices problems with the call, can they immediately make the connection or is it a lengthy IP process? I’ll stop there to that first part.
Brian Morin: Yeah, we do. We keep call recordings of all the interactions, and of course all the data and analytics associated. So, we’ll usually use the top-down view where we’re combing through data and analytics and then say, “Okay, we need to roll up the sleeves and actually get in some recordings.” But of course, all that’s available at somebody’s fingertips just the same way.
Chad McDaniel: Okay. Well, again, thank you for the continued questions. I’ll try to get to them. Let’s move on, and I know we’ve got a lot more we want to talk about.
Brian Morin: Yeah. We’ll try to speed through this. You may be asking yourself, “Well, what is the threshold for change?” When should we look at a particular interaction, and think that this is a candidate for automation. And so, as a rule of thumb, here’s what we use is if we use a minimum baseline for automation as 25,000 minutes a month. If there’s an interaction that your live agents are handling that’s equating to 25,000 minutes a month, that is your transactional or linear or near linear nature, well, then it’s a great candidate for automation. Kind of like that purchasing power example I shared earlier, you are likely to recoup the ROI on that within your first three months. If it falls under that threshold well then we look at that and go, “Is it really worth the cost and effort to customize the conversational AI application to handle it?” It probably isn’t.
So, I just wanted to show you, really, if you’re kind of making your first look at this, usually where most folks start is choosing the reference architecture. And there’s really only two reference architectures to choose from. I alluded to this earlier. This first one scenario A, this used to be 100% of our customers. But what we have found is that over the evolvement, only about 20% of our customers actually fit in this scenario. In this case, the caller calls in. They’re going to be greeted by the IVR having that DTMF touchstone experience. And then if it’s interaction like… Well, let me give you an example.
For instance, AAA, we handle all emergency roadside assistance for them, and we do that whole process end to end automated. However, when they call in, they are greeted DTMF, but if it’s… I think it’s a press two for virtual agent then they’ll get passed to the virtual agent and the virtual agent just like a live agent is going to ask, “How can I help you today?” They’re listening to one of several intents of different reasons why they might have trouble with their vehicle. We’re finding their location, dispatching the nearest tow, and communicating ETAs.
Another good example here would be Office Depot. They’re not using virtual agents on the front and they have DTMF at the front end. But if it is related to returns, then it’s routed to our network where we handle that in natural language. And so, then you can see the next reference architecture. This is where we have seen over time, most of our customers evolve, where they’ve actually taken the virtual agent and they’ve put it in front of their IVR. So that way, whenever the customer calls in, they’re greeted in natural language, how can I help you? Extrapolates intent, authenticates, and then routes to the live agent or if it’s intent to be handled in automation, self-service then will stay in the virtual agent.
Now there will be some that will still keep some of the simple automation. They have their IVR like payment. So if they come in and we understand that their intent is payment, they might like how their payment app works on the IVR, and in those cases the virtual agent would transfer that to the IVR for that particular application. So the next consideration on your approach, Dan’s going to take over.
Dan Fox: Sure. So, if you have looked at automation and this new wave of virtual agents, you’re probably wondering, the same question that a lot of folks who go into this are asking, “Should I do it on my own, or should I partner?” The good news is that there’s no wrong answer. It’s just kind of what works best for your organization. There’s definitely a slew of a lot of technologies out there that allow you to build your own conversational AI. Google, Microsoft, Amazon, all of them have these amazing tool sets. Be careful, right? You want to think about a do it yourself approach as making sure that you have the right team in place to develop these tools because at the end of the day all they are, are tools.
I think this conversation for conversational AI is super similar to web development. You can go on websites like Squarespace and Wix to build a website in one day using a template, and that works for some people. If you are an enterprise, when you approach web development, you’re picking a platform, you have someone to write in copy for it. You have someone designing it. You have someone with user experience background. Before you know it, you probably have a team of five or six people who are building that website for you. So, in that approach, if you’re ready to build out a team for conversational AI, you can do that for sure. Just make sure you have the team in place.
The other option is to partner. Pick an organization, a vendor who delivers it in a fully hosted manner. SmartAction is one of those, there’s plenty more out there that allow you to pick a partner that handles everything for you. So that’s conversational design, implementation, management over time. So it’s basically do you want to handle that development or should you trust another partner to? And at the end of the day, our recommendation is, if this is an extremely strategic operation for you, if you are on the border of becoming an AI company then do it on your own. That’s Bank of America when they deployed Erica, which is their own virtual agent through their application. Bank of America harnesses so much information. They’re a giant financial institution. They can build out the team to do this.
For any of our customers. If you look at someone like DSW, this is a company that wants to focus on selling shoes, and having that great e-commerce presence. When it comes to conversational AI, they said, “Hey, I have five million calls. I want to automate a bunch of those.” And they don’t want to be in the business of delivering and building conversational AI tools. So they partnered with us to do that.
So, just to jump in on the next one. The next step here is really, what does that team look like? So, we hinted out a little bit on the last slide. What we wanted to do is just talk to you about conversational AI, not really being a product, something off the shelf. At the end of the day, conversational AI is full of a lot of technologies. And that’s things like speech recognition, natural language processing, machine learning, but at the end of the day, these are just tools. What’s really important are the services that wrap around it because you have to know going into it, do I have the right people in place to create something with that technology?
I think a good analogy for that is kind of like painting. If you gave me a bunch of paint brushes, and you told me go make some beautiful piece of art, I’m not going to make you anything nice. It’s going to look like macaroni art at the end of the day. And the same thing with conversational AI. If you give someone a set of two paintbrushes, and they’re a painter, what they’re going to create is something really beautiful, something that you really enjoy, and that’s the same with conversational AI. So keep in mind, we’re not talking about a product, we’re really talking about technology wrapped in services to create something that’s truly a great customer experience.
And just we thought we’d throw this out there as an example. This is what our CX team looks like. So if you are doing it on your own through that approach, you’ll want to assemble a team that sort of looks like this. What we have in place is seven CX disciplines that we’ve focused on. So every time a customer comes to us, we have everything in place to support them over time. So that’s figuring out what do I want to automate? What’s the return going to be? Do I have someone managing this project to make sure that it gets out in a timely manner. Human experience design, so having someone in place like Mark who knows how to design for automation because it’s very different from writing a live agent script. The ways that people interact with conversational AI on voice and chat and SMS, this is an expertise. It’s really a dedicated focus.
Making sure that you’re delivering quality. Making sure that you’re reporting on the right thing and able to tune it. You may build a system that can take someone’s order. So, when someone says, “Where’s my order?” You can automate that today. What happens tomorrow if someone says, “Where’s my stuff?” And you didn’t prepare for that when you’re building the application. You need to make sure that there’s a feedback loop to keep that application up to date and accurate. And on top of that, making sure that this application is successful over time.
Brian Morin: So Dan, why don’t you just take me to the next step slide at the very end, and then we’ll just transition right into Q&A with Chad. The slide after this. So, if you are interested in next steps and have any interest in maybe finding out, A, why customers are choosing SmartAction for this approach to step in as a partner that brings your technology stack, plus any services, you can hit us at that email on screen email@example.com.
Most start with either one or two approaches. One is you’re requesting a demo, which is finding out what are we doing for other customers that are in your same space because there’s a high likelihood that we’re already automating a lot of the interactions that your live agents are handling. Or the other piece of it is, hey, let’s step in and do an AI readiness assessment, which is actually just a free consultation. It’s a very consultative approach to sitting down with you and saying, “Is there a perfect fit for AI automation in any of your interactions?” Let’s roll up the sleeves. Let’s look at those interactions. Let’s understand the traits, the conversation flow, understand how linear it is, do you have the data? What business rules are applied? So we can sketch out some type of expected ROI, and CX experience? So that’s something you can engage us about.
So Chad, what I’ll do is I’ll throw it over to you to wrap up. We ran just a little long past the bottom of the hour. But I know that you want to end with your own housekeeping and then we can move into Q&A.
Chad McDaniel: Yeah, thank you, and to that last slide. First of all, Brian and Dan, thank you both of you. And I know we’ve got a number of questions. I’ve got things I want to bring up within our corporate leadership group. For me if I was sitting in the chair like most of the listeners responsible for this stuff, I would be very diligent about self-education and learning. So, if I could take you up on this offer, Brian, so be it, great, or what other offer it is, but we’ve got to continue to self-educate and do things in trial runs. Because consumer preference is changing. We’ve seen it, we’ve known it over the last X period of time. And that’s going to continue down the pendulum so we’ve got to work towards it.
One thing I wanted to bring up, though, before I get into questions, and thank you again, for all the questions that have come through. Brian, where do we run data to predict intent or do we have the right data to build a virtual agent. Getting data in the right format, really just sort of speaking out loud here, but it’s that constant data gorilla data challenge? What advice, suggestions do you have with that sort of data set in mind?
Brian Morin: Yeah. Well, so I’ll actually throw that one to Dan first because a large part of what Dan does is this consultative effort around this piece. Dan, what do we do?
Dan Fox: So, if you’re looking at data within your virtual agent, it is essential to build with data. Like we referenced on the earlier slides, you don’t want to build this in a silo, and it has to be integrated. And we understand when we talk to organizations that a lot of you are undergoing a process to unify that data in a centralized place, and that is the ultimate goal. Unfortunately, that can be a yearlong, two-year process. And maybe when you get that data consolidation effort completed, there’s still things that are missing.
So the important part is to visualize the virtual agent. Less of a technology that needs to replace or integrate into a bunch of places to knock things out but really adding an AI layer to your contact center. So you have your contact center technology, you have your agent desktop, and then you have your live agents, and now you have a virtual agent. And that virtual agent can seamlessly connect into all those disparate data systems. And that’s why when you’re partnering with an organization who does it for you, that’s our role. You present to us what your customer relations management tool is, where your agent desktop is, what your telephony infrastructure is, and then we do those integrations on your behalf. That doesn’t mean that there’s no IT resources required, but it alleviates the burden quite significantly.
Chad McDaniel: Yeah, I brought to the audience here again, when it gets into the data, the data gorilla issues if you need further ideas, suggestions, feel free to reach out via chat, Execs In The Know. We can put you in touch with other brands that are on our online community that can give you some ideas of how they’ve approached it and outside of what’s already been said. But that’s going to be a continuation. I don’t see that need an issue changing. So let’s start tackling it.
One other quick one, Dan and Brian, I want to throw out what I see very consistently in the corporate community on our side outside of the predicting intent, and data is really getting to personalization and segmentation. Now, some of this, you’ve covered off in the piece there. But are there any suggestions or ideas when we think about continued strategies on personalization and segmentation strategies?
Dan Fox: What we find is that the more personalized, the better. The more you can predict why customers are calling, the better you’re able to drive an experience. For a lot of our applications, we have the capability to say, “How can I help you?” But if we know anything about that customer, we want to be predictive about the discussion. Like if someone calls Choice Hotels, and we know that they have a reservation on file, we know they’re probably calling about that reservation and we can ask a pointed question, which allows us to drive that personalized behavior. So in our mind leverage the data however you can to drive a personalized course of action.
Brian Morin: Chad, on the segmentation piece here we have customers that you’d absolutely have dug a layer down and they identified different segments within their own customers based on the lifetime revenue of the customer, based on their rewards level, on the financial side whatever card that they’re on. And so those customers are treated and handled differently than others depending on the segmentation.
Chad McDaniel: But we would all agree as three panelists that these things are fundamentally critical, we’ve got to put them on our roadmap to solving, looking at, and considerations.
Brian Morin: So Chad, the answer is yes. I think that a lot of those listening and say, “Well, how do we actually get momentum behind this?” It’s the same way around the data question that you brought up earlier, and so even for a lot of our customers who typically need a lot of collaboration with IT, and in every organization that seems to be one of the most limited resources. There has to be an impetus that starts that initiative within your organization. And so, for in our case, we can oftentimes end up being the Trojan horse where even if you’re not implementing an AI vendor, it’s that you’re looking to implement an AI vendor, you know that you’re going down that that route. And so, that needs to be an impetus or reason you as a compelling event to drive the action within the organization so that you can take advantage of the potential that new tools offer.
Chad McDaniel: Well, I know I always put myself as a consumer whenever I’m calling an organization. We’re all consumers and we definitely know what various experiences look like and when we’ve chose to remain with that brand or not in service being a competitive differentiator. So we as service leaders believe in that that motto and passion. All right. You’ve been very patient. I want to get to some of the questions here. Justin, thank you. The question from Justin is any idea the percent of people that opt for a live person versus AI?
Dan Fox: I think it I hate this answer because it’s like the typical consultant answer, but it depends on what the use cases. Our job it’s probably around 10 to 30% on average for most organizations, the customers who aren’t going to work with automation. What you have to do is design for three unique ways that people interact. You’re always going to have a demographic of customers who don’t work with automation. The ones who no matter how great the AI that you put in place is, they just don’t want to use it. And that’s fine and let them go over to an agent.
There’s also people who are like, “Great, I want to automate more than ever. If there’s a self-service opportunity, I’m going to use it. I don’t want to talk to an agent if I don’t need to.” And then there’s that second bucket, or the third bucket, sorry, the third bucket are the customers who aren’t expecting automation to work. But if you can deliver them an experience that works for them, like greet them by their name, ask are you calling about that bill. And if you’re able to do that, you might just be able to get them the self-service experience that they’re looking for that they didn’t necessarily expect to get.
So rather than focus on, yeah, 30% of my customers will never work on it. Keep in mind that that second area is growing. People are becoming more comfortable with conversational AI. They use Alexa in their home. They use Siri and Google Assistant, and when they contact a brand, they’re not shocked anymore when you say, “Hey, how can I help you?” Because they’ve interacted with those systems before. So, the biggest thing is like, okay, I have one shot to get this right. And if my conversational AI delivers something that provides value, they’ll use it, and next time they call in they’ll use it again.
Brian Morin: I would say that if we were to give you a stat to it, I would say now in our current climate, just across the board with customers, we would see that number less than 10% you’re opting for a live agent. But however, if we were to rewind this conversation to let’s say you have four or five years ago, those numbers were a lot higher. And you might ask what’s changed? It’s that not just that people become more accustomed to it, two, they don’t want to wait on hold for anybody. They want to be able to self-serve at their convenience. But frankly, a lot of it has to do with how the system sounds. If you have a system that is human sounding, and it shows some intelligence on the front either knowing why you called or being personalized knowing your name, you’ve now just dramatically pushed through your success rate at those who want to interact with the system.
Chad McDaniel: That’s an excellent point. I’m going to give it a little bit higher percent just because I think some of us are still doing not a great job with it. But at the end of the day, we’re learning and we’re learning very quickly. So for the audience, this is our chance to learn together. So if you have any comments that you want to throw in, or any tips, recommendations, best practices, or questions, please throw them in here, and I’ll try to get them out there. Stephanie, thank you for your question. The question is, how does the AI do with names spelling? I don’t know if that’s incorrect spelling or what Stephanie’s asking specifically, but how do the AI, names do with spelling?
Dan Fox: I can take that. Names are actually a pretty capable thing for an automated system. We typically get high success rates on capturing names because they are all letters, and if people say it in a certain pattern, you have a pretty high chance of capturing that correctly. An interesting thing that you can also do with named capture, something that we do with Electrolux. When we do a product registration for them, what we do is a 411 lookup by sending that phone number that they’re calling from to a database to retrieve what we think that their name and address is. And then rather than saying, “Hey, what’s your name? Or hey, what’s your address?” We say, “Hey, I think from your phone number, your name is Dan Fox, is that correct?” And what we can get is a really high accuracy rate out of that. So that’s just a little trick that Electrolux uses.
When you move outside of names and simple stuff, it gets a little more complex. If you have alphanumeric numbers, you can get a high accuracy if you know what you’re looking for. So if everything has the same pattern, like a VIN number, because we do a lot of VIN numbers for AAA, we know we’re looking for a certain pattern that goes like letter, number, letter, letter. And so, we’re able to get a high accuracy rate for that. When you look at something like email it’s pretty complex. You don’t know how many characters it is. It could be numbers, it could be letters, That’s something that you probably don’t want to automate and that you’re not going to get the best experience from. So yeah, if you’re looking for names, you got it. If you’re looking for something within a defined set of characters that are letter and numbers, you can handle it. Just freeform email addresses probably avoid.
Chad McDaniel: Yeah, I think, Stephanie, the answers remain the same. For clarification was being able to spell complicated names, but I think the answer remains the same. Richard, thank you for your question. Richard’s question is how does or will the virtual agent work in healthcare due to HIPAA privacy concerns? Great questions.
Brian Morin: Sure. Dan, do you want to take that then I’ll play cleanup?
Dan Fox: Sure. Yeah. And anytime you’re building a virtual agent, you have to keep in mind what regulatory environment you’re in. Whether it’s federal requirements, HIPAA requirements, capturing credit card information, and maintaining PCI level one compliance. Same thing with HIPAA. When you develop a virtual agent, you have to make sure that it has compliance in mind, which means redacting any information that is deemed as sensitive. Making sure that you’re storing that information in a HIPAA compliant manner. So, our solutions are HIPAA compliant. That’s something if you go down the build it yourself road that opens up a little bit of a tricky situation. But yeah, if you’re working with a vendor, they should be able to pretty quickly tell you if they’re HIPAA and PCI compliant.
Brian Morin: Yeah, and where we see is kind of like the low hanging fruit where we see most go on the healthcare side. Certainly, it’s the HIPAA compliant authentication that can take because it is low longer, more extended. There’s more average agent handle time involved in that, that you can offload, but we see them moving down the path of their outbound reminders related to appointments. Now you have the ability to actually do some rescheduling, cancel and reschedule. And then of course, everything that’s coming in inbound, you’re capturing intent, and then routing or keeping within automation for self-service.
Chad McDaniel: Yeah, thank you. I’m going to continue down the questions. Paul, I saw your tip. I’ll get to it in a minute. But I want to get to Caleb’s question. I’ll answer this one first. And then Brian, Dan, you guys can agree disagree. We can debate. Caleb says, please address the future of call center jobs. Will there be any with this type of technology coming in and improving? Obviously, this is a reduced role for human agents. Will there be any roles left for human agents? It’s 2020. As far as remainder of my working career, 2030, whatever that’s going to be. Yes, there’s going to be a big portion of call center agent jobs for a number of different reasons. Maybe by 2040 be different world but I would say in the next 10 years for a number of different reasons live agents are going to be predominant. I don’t know. You can disagree.
Brian Morin: Well, I mean, so we’ve been sitting at the front seat of this watching this trend emerge. If we were to look at our customer base just across the board and kind of make a broad based statement, I’d say that we’re automating for most. Roughly around 30% of the interactions that are occurring. For others, it’s much higher than that. Ultimately, it comes down to what can you and what can’t you with a virtual agent? If it’s something that’s going to require judgment or complex critical thinking, that stays with the human, but frankly there are other pieces and elements that, yeah, a virtual agent could handle it, but you’re not getting enough volume where it’s actually worth the effort. You have to customize that app for them. For that reason, a lot of your corner cases are still going to be transferred and handled by live agents.
So, we frankly, we just don’t see live agents going away. But what we don’t see and I think a little bit the misnomer is that we don’t see customers implementing conversational AI and then just going through their contact center, and then wiping out customer service rep jobs. All that they’re doing is just having to let their own organization go through their own natural process of churn. And in the course of their own process of natural process of churn, they’re just not replacing those FTEs until they’ve right sized as an organization.
Dan Fox: I’ll just add one thought there. You have to keep in mind that what are your agents doing to help improve their quality of life because employee satisfaction becomes even larger when you’re talking about artificial intelligence. There’s a process that Forrester refers to as flow where it says employees work best when they know that they’re contributing to an organization. So if you have a revenue generating contact center, and your agents are getting bogged down with this stuff that’s handled by automation, or that could be handled by automation, then it reduces their sense of flow. If you have more agents that are focused on those true human only tasks, it gives them a higher elevation, higher job satisfaction. So, in a lot of ways, this wave allows you to get higher performing, more involved agents.
The second thing I’ll put forward is that as our artificial intelligence continues, the job of a call center agent may shift a little bit because in 20 years, maybe we’re not having the same conversation and maybe call centers are shifting. But keep in mind that all of this technology requires humans at some capacity. So maybe those contact center agents become more of a blended role, where they’re aiding the artificial intelligence engine and making it get better by this human in the loop process. So, there will be a shift I’m sure but no time in the near future are we getting rid of contact centers.
Chad McDaniel: You alluded to it. I’ll just make a final comment. We’ll move out and wrap up here is that there are many use case studies of how automation is enabling agent experience success and actually automation has been very well received and welcomed by agent experience. So it’s been a good thing not a bad thing. Quick question here, Maria, is your solution PCI compliant?
Dan Fox: Yes.
Brian Morin: Yeah.
Chad McDaniel: Okay. I’ll leave this on pause. I know we’re getting to the top of the hour here. Emily, Paul, thank you for your suggestion. Paul’s just recommending to us as a group with this conversation is have a clear goal for the IVA and let the user know what that is stated clearly. I would say amen to that. Thank you. And I agree. Brian and Dan, this has been very helpful. Thank you. And obviously I think within the Execs In The Know corporate community, we need more of this conversation. We need more learnings as you’re continuing down this path. So, anything your organization can do to share the learning that you’re seeing and use case studies, please come back as our guests. We’d love to have you again.
Brian Morin: Absolutely, Chad. It’s always a pleasure. Appreciate speaking with you and of course your audience that you’ve curated here. We always have good discussions with this group.
Chad McDaniel: Well, on that note, I want to thank you for your listening. If you could get to our next event, which will be in Hollywood, Florida. We’re going to be having two days of this type of conversation and many other data sets of conversations. So, come out, be with us meet and greet a lot of the groups that’s coming there. So on behalf of myself, Chad McDaniel, Brian, Dan, thank you again so much. You’ll get a recording link of this tomorrow and you’ll get a survey link as soon as we conclude here shortly, if you could take a minute to conclude. So keep the motivation for customer first, and have a great rest of your day. Thank you, everyone.
Dan Fox: Thanks Chad.