Chief Content Officer,
Techstrong Group


In this Digital CxO Leadership Insights Series video, Mike Vizard chats with Mahesh Ram, head of Contact Center and digital customer experience for Zoom, about how advances in conversational artificial intelligence (AI) are changing how customer service enabled by virtual agents is provided.



Mike Vizard: Hello, and welcome to the latest episode of the Digital CXO Leadership Insights video series. I’m your host, Mike Vizard. Today we’re with Mahesh Ram, who’s head of digital customer experience for Zoom. And we’re talking about virtual agents and advances in AI and how that’s all coming together. Mahesh, welcome the show.

Mahesh Ram: Hey, thank you so much; my great pleasure to be on with you.

Mike Vizard: We have tried to improve the customer experience a number of times over the years, we’ve had mixed success. A lot of folks, you know, appreciate the fact that the cost may be lower, but the customer experience always seems to have been sacrificed in the name of that cost. So now we’re looking at these new generations of virtual agents enabled by conversational AI. So how is this whole equation about to change?

Mahesh Ram: So I think, first and foremost, like most things, it’s about thoughtful design. And I think if you start with the premise that the user experience is the single most important thing, it’s important for the user, and it’s important for the brand, or the company; I think good things tend to happen. Because I think today, the available technology is really at a place it has not been in history, which actually allows you to create what I think of as being effortless experiences for the user, that are also beneficial to the brand. And I don’t think that that was possible a decade ago, but today it is that you can use conversational AI, technologies, techniques, natural language understanding, to actually understand what the intent of that user is in that moment. And then be able to take them on the right journey. And sometimes that journey is about self service and automation and support. Other times it’s about getting them to the right human agent or the right human expert who can help address the issue. And it’s about – I sometimes liken it to building the best digital concierge experience you can you can get. And you think about it from that perspective. I think opportunities are are plentiful.

Mike Vizard: So what is the quality of that experience? Can you describe it? I mean, short of giving a full blown demo, but you know, what should people expect?

Mahesh Ram: So I think people should expect a few things. One is that I think AI and machine learning techniques allow users to express themselves in everyday natural language, they come to a brand, it could be on the mobile website, it could be in social, it could be in a lot of different places. And if they’re interacting with a Zoom Virtual Agent, for example, they would just express their issue in everyday natural language. They wouldn’t try to speak the way a knowledge base or an FAQ would expect them to speak, they would just express the issue the way they would if they were talking to a human agent. The system in the AI then is able to use intent understanding, which is a field of AI and machine learning to properly understand and categorize the issue. Now, once I understand what you’re asking me about you, you say that you want to stop a meal service; I know that it’s a cancellation of a subscription intent, I can then take you on a very specific journey that could either could take one of two forms. Either I’m going to immediately give you an answer. If you ask me what store hours are for a particular branch in a particular city, I should be able to answer that question for you. From a knowledge base or an FAQ, I can use AI based question answering to immediately extract that specific answer from that knowledge base or FAQ presented to you and say, “Did this help you?” And if you say no, then maybe I escalate you and I ask you a follow up question or I get you to a human agent. But in other cases, I might need a little bit more information from you. So I might have a little bit of a back and forth with you, as the virtual agent might ask you, “Hey, what’s the order number you’re asking about? What’s the issue you’re having?” And then based on those criteria, it might look up some information from a back end system, it might make some decisions, it might even refund you the money right then and there if you’re qualified to do so. So if you think about that, what I just described for you is a set of journeys, but the beauty of it is the consumer. The user doesn’t have to know which journey is right for them. When they express the issue. The system can understand it can do it. I’ll give you the extreme example; if someone comes in and they’re, they’re feeling unsafe in a ride, share the intent understanding can immediately say, this is not something we want to self-serve. We don’t want the chatbot to be there, we want to get you to a human being who can get you to a Trust and Safety agent who is best qualified to help you immediately. So what I’ve just described, I think is an elegant experience that people actually treasure and they value, and that’s what our recent survey that we did showed is that when people have that experience, they actually come back to the brand with a with a heightened sense of loyalty and trust.

Mike Vizard: So do you think the customer will ultimately be happier because the system is essentially triaging the calls, and today it’s more of a first in first out kind of play, and you may have a simple question that you don’t want to spend that 45 minutes waiting on a phone call, because the 20 people in front of you have complex questions? So do you think that there is is an opportunity to kind of change the way people do engage with companies?

Mahesh Ram: I think it’s a generational shift, too, I think what we’re finding is that an entire generation of users today, doesn’t use the phone. You know, they, they prefer to get instant answers. They like to have things that are – they prefer the immediacy when it’s accurate. So you have to have immediacy combined with accuracy from a user perspective. But as a user, there are some times when I know I want to get to a human being and the issue type is such that I have to. So that’s kind of what I described earlier. But if we should also think about it from the perspective of the businesses that are deploying Zoom Virtual Agent, what are they looking for. They have a group of agents who are experts in a particular domain, if you can get the the issues that should not be in front of those agents cleared, what you then end up with is an opportunity for those agents to handle the more complex issues, the edge cases, spend time with more empathy; they have a little bit more time to listen to you process, do the things that you need to do. And I think if you’re a business, that’s a really positive outcome, serving the consumers the way they want to be served. I’m also making my agents better, smarter, faster, happier. And I’m making the entire business more efficient for everybody. I know it sounds like a perfect outcome. And but as we all know, you know, it takes thoughtful design to get there, it just doesn’t happen overnight. The technology is great. But a lot of it is how you actually deploy it. And that’s, I think what we’re very proud of is the way we we’ve done it with Zoom virtual agent which is really user centric design, I think, in the way we put this together.

Mike Vizard: I think one of the things that people don’t often appreciate is the burnout rate in these customer service organizations is pretty high. So as we move down this path, do you think that that will get reduced because a lot of the routine trivial stuff that essentially bores people to tears will be automated?

Mahesh Ram: I think a few things. I think this is certainly a contributor to that. I think a few other things are contributors to that as well. One of the sister products of Zoom Virtual Agent is Zoom Contact Center, which was released as an agent facing technology in early 2022. And I think when you think about that design of Zoom Contact Center, taken together with virtual agent technologies, what you’re seeing is an emergence of saying, how do we make agents feel more like experts? You know, I think all of us want to be seen as experts in our domain or our field. And the more we’re empowered with tools, and technologies that that, that make us more capable, and get a smarter, the happier we are in our in our line of work, right? The less we feel like robotic people doing the same transaction the same way over and over again, the more likely we are to stay with the employer. And I think that’s what we’ve done with a combination of Zoom Virtual Agent and Zoom Contact Center. So imagine, Mike, this specific example: You interact with Zoom Virtual Agent, it collects some information from you, because it understands the intent of the question. It takes that information and says, I still think you need a human agent, it connects you to the agent whose best fit to answer that question. It gives that agent all the context for the conversation that Mike’s had with the virtual agent. So you’re getting to the right person with the right information, and that person is not making you repeat yourself. So the consumer is happy, the agent can get right into helping you can pull in data from other systems that they need to answer your question. You can just imagine what a better agent experience that is. It’s just so much better than having to make you repeat yourself. Now you’re irritated as a user; I as the agent, I’m not doing anything with my time that’s productive for you. And it’s all avoidable. I mean, that’s the great thing about using available technologies, is these are all solvable problems. You can do them today.

Mike Vizard: A lot of organizations will try to upsell products through their customer service arm because they have the person engaged and there’s typically an issue at hand. Is that something a Zoom Virtual Agent can do and can understand what other things in the portfolio might be applicable?

Mahesh Ram: It can do a lot of different things. And I think that leads me to, to maybe a subtle point that may not be obvious with what Zoom Virtual Agent offers, is we’ve actually made it possible for non-developers, subject matter experts in these companies to design the experiences that best match the business outcomes they want to accomplish. So let’s use the earlier example I gave about somebody who wants to cancel their subscription, right? That’s a moment of truth for a brand because if they don’t do the right thing, they’re probably going to lose the customer. And they might have to lose them, but there’s a difference between not trying to lose them. I’m doing something about it. And let’s just assume that, Mike, you are the customer and you’re a VIP, and the brand knows that and you’re you’re looking to cancel. Instead of simply just allowing you to cancel or connecting you to an agent who then canceled your subscription, I might artfully design an experience that asks you why you’re canceling. And it might be that you’re taking a vacation, you know, maybe you’re going to Rome for two months. And then you don’t need that subscription for the two months. Well, then I can take you down a different path. And I can say, “Well, hey, we have a pause option for you. And if you’d like to pause, we can just resume it when you’re back.” Or you might say I have too much of your product. And that’s why I want to stop the subscription. And we might say, well, that’s okay, we have ability to skip this month, and we can keep going next month. So it’s an example of taking what would otherwise be a churn 100% of the time, to turning that into into a retention, right? So I’m using a retention example. But other examples exists today. But I think, where you actually then can do an upsell, you could say, “Hey, you’re having this issue with the product or service that you purchased from me – consumer electronics device.” All of us have that where you have an issue. I don’t know how to do X and Y. And I solved the issue for you first. So first and foremost, I must help you find the solution to the problem. I can’t upsell you, or convince you of anything else until your problem is out of the way. But let’s assume I do that successfully. I might then come back to you and say, “Hey, did you know that we have a repair service that we can offer as a subscription service, and you can pay $3.99 a month and you could have somebody help you with this. And you could get instant access to a hotline for repairs, if you wanted. Would that be of interest to you?” Think about where I’m placing that in the journey. I’m placing that after I’ve understood your issue, after I’ve diagnose your issue, after I’ve resolved your issue, and then I’m presenting it to you completely as an option that you might want to you might want to take up. It’s a very elegant experience. And it doesn’t leave the brand feeling like there’s pushing or selling all the time because I think consumers really dislike when they’re being sold something, and I’m not here to be sold to. I’m here for my problem to be solved. So I think it’s that sequencing and mental sequencing that matters. But if a brand wants to do that, they can just do that without calling a developer. The thing I would say, for your viewers, I think that is the most exciting thing that I’m excited about Zoom Virtual Agent is that you don’t need any developers to build any of what I described. You don’t need any engineers, you don’t need any developers. And you as a brand can control those experiences, build them, change them, connect them to your back end system, all without actually even calling us in. They can build them themselves – that people building these are, you know, folks who who kind of build these experiences out. And that’s, I think, that’s the power of what is possible today, we’ve got I think of them as citizen, citizen builders, you know, citizen experience builders. And that’s, I think that’s the future for a lot of things in tech. And it is enabling non-developers to build these things. And we’ve certainly done that.

Mike Vizard: Right. And a lot of that also has to do with those citizen folks are closer to the process, a lot of times the developers are a little bit too removed. And you wind up with a lot of back and forth trying to create something that if the citizen person who knows the process just created themselves, we’d get to the end result faster.

Mahesh Ram: That’s a great point. And it also dovetails with a very important fact, which is that I don’t think there’s ever been a time in history when businesses are more dynamic than they are today. You know, companies are introducing a product or service that seems like every week, every month, and the pace of that change is so fast that you almost need to have a technology wave for a citizen, citizen, expert about that product, new product or service to be able to go in and quickly say, “Hey, starting tomorrow, people are going to be asking me a lot of questions about this new service I just released. How would I best be able to answer that so that I wouldn’t get flooded with calls?” And so if you were to rely on developers or wait for developers, you might be waiting a month. And you might lose the window when when a product is fresh. And everybody’s coming in and asking lots of great questions about it. But imagine being able to go in there and immediately be able to set up a virtual agent bot flow specific to that new product or service and understand the intent knows why you’re here. It answers the frequently asked question, but it’s also listening, Mike. And this is the other part that I think is subtle and not obvious, is with Zoom Virtual Agent, because we see all the conversations that are coming into the brand, we can actually start to tell the brand things that they did not know before. We can tell them people are asking questions about these topics. You don’t have a good answer or response to this. You can quickly go in and add one. And the next set of users who ask the same question will now get a great resolution. Or here’s a bunch of things that maybe are business policies that You’re missing, no answer in the world is going to keep this customer happy, you might need to change your business policy. But I’m telling you that this is coming in fast and furious. And we actually had a real example of this with a predecessor product, to creating Zoom Virtual Agent where we had a food delivery business. And we were telling them quickly that they were having a lot of problems with ice melting in the meal, that they were being delivered to the home. That’s not a support issue in the sense of like, the support team can’t do anything except give you a refund, which they can do. But what we were able to tell them was something’s going on with their supply chain, maybe you need to fix something in the warehouse in a shipment. And that’s a powerful piece of data. I couldn’t self serve those users, because all they can do is get a refund. I can self serve them by giving them money. But I’m also telling the brand, something they didn’t know. So insights and analytics, when you can listen to that. That firehose of conversations that are going on between the brand and the consumer, you actually start to see things that a single human being inside that brand would never see. I might see 10,000 of those conversations a day and the virtual agent is able to process them and make sense of them. And allow them to take action. I think it’s about action, right? It’s taking better actions improving constantly.

Mike Vizard: We hear a lot about all things generative AI these days. So how smart can this get as we look forward and kind of start thinking about additional capabilities? Because well, you know, my brain is getting smarter at a very slow rate; the machine is getting smarter at a much faster rate.

Mahesh Ram: So I have no doubt you’re going to ask me about generative AI today. I think it’s the flavor of the day, week, month. And I do think it’s transformative technology when used appropriately. I think when you think about things like large language models, you know, which is an area that we do a lot of work in, in Zoom. Certainly the advances, the amount of data available, the advances in the technology allow us to innovate faster and better. But again, just like just like chatbot design, you know, it has to be thoughtful. So we have to look for, we have to say which datasets are most appropriate to solve the local problem we have. So like most AI, generative AI things, it comes down to the problem you’re trying to solve. So in the case of Zoom Virtual Agent, let’s be very specific. You as a user are coming to a brand with an issue that you need resolved, right? You want to probably have a problem or an issue that you want to resolve, it’s usually very specific to that brand, the policy for returns from one band to another might be different. You know, the troubleshooting for an iPhone is different than it is for an Android. All of these subtle differences between brands and between experiences. We can use generative AI to get better at the accuracy and the prediction and consistently, consistently, you know, get it right. But when it comes to the resolution path, it often goes outside the realm of generative AI, because it might require looking up Mike’s order information, it might require making a business decision about whether Mike is entitled to a refund or not. Maybe you’ve cancelled five times in the last month. And, you know, for that reason, I’ll make you talk to an agent. You know, that’s a possible thing that’s not a generative AI, you wouldn’t want a generative AI chatbot to make that decision today. Not capable of making that decision today. But we also don’t want to ignore the enormous possibility when it comes to content creation. When it comes to the ability to get better at having a conversational experience. We are using many of the techniques in our proprietary models to do this. At Zoom. And and I think that’s going to continue. I tend to be an optimist about such things. I think generative AI is today is still in the, maybe the the crawl, learning how to walk phase, but it’s going to be running very fast. And if we don’t think about how we can use it – thoughtfully eliminate the bias, make sure it’s accurate, and make sure it’s actually fit to the purpose that the user needs at that moment, then I think it’s transformative. But that’s a lot of caveats, and there’s a lot of people who are going to do it poorly. And that’s the other thing is they have to be done with care.

Mike Vizard: Do you think ultimately, we could change the way many business leaders view customer service, because for years it was always about reducing the cost as much as possible. And then everybody came up with these kinds of playbooks. And anything that went outside of that became a problem. So can we rethink customer service now and say, “Well, if the cost of the service is gonna drop, because we can automate a lot of it, maybe we can really focus on the quality of the service?”

Mahesh Ram: We’re already seeing that a lot of our customers have been able to actually retrain because they’ve been able to free up agent time. They’ve been actually able to retrain some of their customer agents into being more consultative experts, as opposed to, you know, take a number and we will process your issue type of order takers or whatever it might be. And I think this is just the beginning. I think that the other, you know, the other parts of this are, you know, in Zoom Contact Center, for example, we’re making intelligent recommendations to agents about what they could be doing right and pulling data in from the various sources. Where, you know, I  know that you’re calling from so and so I know where your order information is. I know what your last three issues were. I know that instantaneously. So I’m able to have an intelligent conversation with you. In the moment, in earlier days, it might have taken me eight, nine, ten mouse clicks, moving to multiple systems to be able to pull up enough information to have a cogent conversation with you. So I think all of that does improve the user experience and the agent experience. But in terms of how brands are thinking about it, I had one chief customer officers say to me the other day, “I see it now, we’re going to be able to take customer support from being on defense to going on the offense.” And I thought that was the best way to summarize it. And offensive doesn’t mean selling you recklessly. It means actually being able to listen and anticipate what the consumer actually might want next week, next month.

Mike Vizard: Alright, folks. We’re a long way from push one of the following six buttons and hope that somebody actually knows what’s happening on the other end of that call. Hey, Mahesh. Thanks for being on the show.

Mahesh Ram: Thank you, Mike. Appreciate it. And thank your audience as well.

Mike Vizard: And thank you all for watching the latest episode of the Digital CxO Leadership insights series video. You’ll find this one and others on the website. We invite you to check them all out. And once again, thanks for spending some time with us.