Less than 10 years ago, having intelligent conversations with a voice-enabled computing device was a familiar feature of Sci-Fi movies, but an uncommon sight in everyday life. A decade onward all that’s changed because as we’ve seen time and time again, there’s no such thing as science fiction, only science. Wherever you look these days, smart speakers, chat bots, and other advanced communication channels employing conversational AI are making information more accessible than ever. Adoption of these technologies is so rapid in fact, that it appears to be hastening the demise of mobile apps, and will perhaps eventually become our primary application interface in the not too distant future.
To better understand this seismic shift in how organizations interact with their customers, employees, and partners, we turn to Matt Smith, Conversational AI Practice Leader & AVP of Cognizant. Matt shares with us some of the fascinating use cases he’s seen companies deploy, as well as where he thinks conversational AI will go over the next 3-5 years. Along the way we’ll find out what use cases are best suited for conversational AI, key factors in creating a successful conversation design, and how conversational AI could even change the way you order fast food in a drive-thru.
Guy Nadivi: Welcome everyone. Our guest today on Intelligent Automation Radio is Matt Smith, Conversational AI Practice Leader and AVP of Cognizant.
Matt has had an illustrious career in robotics process automation, and IT outsourcing. He’s served on the leadership board of the International Association of Outsourcing Professionals. And as an industry leader, in the field of conversational AI, we wanted to get his perspective on this disruptive technology, which is really taking a number of industries by storm right now.
Matt Smith, welcome to Intelligent Automation Radio.
Matt Smith: Thank you very much, Guy. It’s a pleasure.
Guy Nadivi: Matt, you run Cognizant’s conversational AI practice. Can you speak a little bit about what is conversational AI, and differentiate it from the other AI we hear about out there?
Matt Smith: Yeah, of course. The way that the term is generally used right now, conversational AI, it actually refers to a number of different ways for brands to build a new way of engaging with their customers. Their customers could be internal workforce, it could be actual customers of theirs that are already buying something from them, it could be prospects. But conversational AI has become an umbrella term for everything that could be, for example chatbots, or intelligent speaker devices, or hands-free speaker devices, virtual agents, and even more and more examples of conversational AI we’re seeing are applications within a vehicle, or in locations within a home where you can speak your instructions to a device, versus a traditional way of using a keyboard or some other form of interaction.
Conversational AI, it really is a subset though of the broader category of artificial intelligence technologies. You can’t do true conversational AI without a couple of the key building blocks that are part of AI systems today. Natural language understanding is a critical component of a conversational AI solution. Machine learning is a very important component of conversational AI. Those are really part of the engine, if you will, that makes these technologies able to do what they do, and seem very seamless from an end user’s perspective.
Guy Nadivi: You mentioned that it’s a replacement for keyboards, and you’ve described conversational AI as hands-free computing or screenless computing. I think most people personally experience conversational AI when interfacing with a chatbot, or intelligent speakers like Amazon Alexa. For organizations that have implemented conversational AI for their customers, how have you seen that it’s given them a competitive advantage today?
Matt Smith: If you look at some history examples of where a new user interface has emerged, and that’s really if you think of the idea of screenless interface, or screenless computing, it’s a new user interface. If you look back in some past eras, we had the website era in the ’90s, and companies – initially it was seen as a research tool, or it was seen as an extension of branding, and a lot of companies just took their brochures and converted them to websites. But eventually got to a point where they realized, there’s a lot more here and customers are interested in doing a lot more than just flipping through something electronically that they could previously have done with a catalog. Much the same happened a decade later with mobile apps. Then it was a little bit of a similar experience where companies then just took their websites and converted that to a mobile app.
But in both those cases, a learning curve happened where the idea of these new user interfaces started to catch on with marketing departments, with sales organizations, with customer services teams as a new way to compete for customers, as a new way of differentiating what they can represent to their marketplace. I think we’re seeing that very much play out now with conversational AI, Guy, where people are experimenting with the technologies, and very quickly starting to learn that this is a new way to differentiate in the marketplace.
The whole idea of conversational AI is to create a better ease of use, a greater convenience for customers, the idea of personalization, and being contextual to a situation is very much a part of it. Again, that’s where some of the AI technologies come into play, but it’s the new competing space for brand differentiation, is really what conversational AI has become.
Guy Nadivi: In broad terms, what are some of the lowest hanging fruit best suited for conversational AI applications within an organization?
Matt Smith: Well there’s three that we see right now, Guy, with the clients that we work with. The first is if you think about very high frequency, highly repetitive informational kinds of requests. Things like frequently asked questions that today might be on a website location or embedded within an app, finding forums, and finding locations or branches, or store sites, or looking up menu items. Those kinds of very high frequency, highly repetitive kind of interactions, that’s one area that we see a lot of organizations starting with first.
As far as some of the use cases, lead development and lead conversions is definitely an area of focus for a lot of conversational AI solutions. As a way of reaching out across a multitude of channels, and engaging where your customers are, or where they would prefer to engage with you is a great place to look for conversational AI low hanging fruit or early opportunities.
The third area is lower complexity but transactional inquiries. It’s a step up in complexity from just information requests, and it will typically require some kind of login, or credentialing, or authentication of some sort. But for example, being able to look up an account balance, or update missing miles on an application, a mileage app for an airline, filling in applications, or initiating the ordering of products, those kinds of today, fairly simple transactional activities are also become more and more popular for companies that are looking to implement conversational AI.
Guy Nadivi: To refine that a little bit further, can you please share with the audience some of the more intriguing use cases where you’ve seen conversational AI impact an organization’s operations?
Matt Smith: The main area of focus right now for most enterprise kinds of clients, Guy, is in the idea of contact center transformation. If you think about the way contact centers had evolved over the years, and all the elements of technology today that exist, and some of it has been very beneficial for customers, and some of it customers we all have come to despise, the IVR or having to repeat your account number every time you get transferred to someone else. Those are areas where organizations are focusing today on conversational AI, and again, how can it create a better customer experience versus the ways that organizations have used some of these other technologies? We see a lot of activity in contact centers, and it could be voice interface, it could be better versions of web chatbots, it could be an integration of those technologies, it could be intelligent messaging as an option.
We were working with one client, and they actually started to experiment, Guy, with the idea of offering customers that were on hold, the option to try and resolve the question themselves using a messaging solution. They didn’t have to download an app, they didn’t have to go to a website. It would embed a functionality within their existing messaging app on their mobile device, and they had an 80% acceptance rate of customers that chose that option to self-serve rather than waiting on hold for the next available agent, and were able to resolve their own issues. Those kinds of ideas, I think, are really gonna start to catch on over the next 12 months or so.
Then some of the other areas that I think are more from a user-oriented perspective that we’re gonna really start to see a lot of examples, so we’re doing some work in the fast food industry. If you think about the way drive-thrus work today, it’s a very linear process. You wait in line in your vehicle to pull up to the sign, you tell the operator on the other end what you’d like, they read it back to you, they tell you what you owe, you pull up to a window. It’s not really ideally for the customer or for the restaurant because they can only sell as much traffic as they can run through the drive-thru. The ability of conversational AI to really extend the footprint of the drive-thru concept, but be able to leverage that around vehicles that could be anywhere in the area to place orders, and to do it not necessarily through a person, but to be able to order their items or their meals through conversational AI, and then just pick it up if they get anywhere near the building, another cool use.
Then we’re seeing, for example, in retail environments. We’ve all been in stores, grocery stores, or big box stores, or big hardware stores, and you simply can’t find the item you’re looking for. We’re helping a couple of companies that are operating in the grocery space to put intelligent agents on top of their mobile app. You can simply ask where an item is located, and not only does it tell which aisle and which shelf, but it can bring up a visual representation. As you get closer, it uses echolocation to show you exactly where your item is gonna be. I think those are some of the things that we’re seeing right now are really gonna be, before much longer, just standard for a lot of the interactions that we have.
Guy Nadivi: I love that last use case, especially at Costco where they change things every week. I could use that right now.
Matt Smith: We all could. Think of the time it would save you and the convenience too.
Guy Nadivi: Matt, in a recently published article, you state that, “Conversational AI is a top 2019 priority on the strategic plan for nearly every company we’ve talked with.” Is this an indication conversational AI has crossed the chasm from innovators and early adopters, and is now being embraced by the early majority to become a mainstream technology? Or is it still somewhat experimental?
Matt Smith: Yeah, it’s straddling those areas. I think it somewhat depends on the industry, Guy, because there are certain industries where it’s somewhat easier to implement. For example, the technology can be used equally in a retail environment to do simple transactions, like we talked about earlier, or in a healthcare environment. But if you put on top of the healthcare environment the complexities of security, and privacy, and compliance that exist, it suddenly makes it potentially a lot more complicated, so the functionality could be there, but there’s other drawbacks that make it more challenging.
What that means is we’re seeing certain industries move faster with conversational AI-based solutions than others, but there’s definitely this effect from a customer point of a view, a consumer’s perspective. If I can use this technology in a certain way over here, why can’t I use it the same way over there? It’s pushing the whole pile forward. I think it is creating this … We’re crossing over into the next era.
We think that we’re in this, what I would call the third microphase, or the third phase of conversational AI. It really is now about this idea of starting to put more and more focus on a multitude of ways to engage, connecting these projects, and having a strategy for scale. I think we are at that point, and just if you look around, all the examples you see as a consumer of conversational AI coming into play, I think we can all sense this has happened.
Guy Nadivi: Is there a single biggest factor driving adoption of conversational AI among your clients?
Matt Smith: Yes, there is. It’s really two dimensions. The drivers are the idea of commerce and customer care. Using these technologies can extend, or enhance, or improve the idea of both commerce within customer care and customer support. That’s one dimension as the driver. Then the other is really a need for scale and for strategy.
Again, I’ll draw a parallel to where websites were in the ’90s. When they first started to gain a degree of popularity, there were a lot of companies at the time that said, “What do I need a website for? That’s for such and such, a kind of brand, or for that kind of an example or a use. We don’t need a website, our customers wouldn’t go there.” Then suddenly those same companies found, in many cases they had hundreds of websites that had popped up as different teams around their organization were experimenting with this technology, then they were building little, today we would call them microsites, but building little websites for their department, or their function, or their operation. That’s when this idea of scale and strategy suddenly hit IT leaders, and brand and marketing leaders. Back then they said we need brand standards, we need the architectural standards, we need strategies for security.
That’s what’s happening right now with conversational AI. We’ve had these two earlier phases of a lot of experimentation, and a lot of small pilots, and small deployments. Now suddenly the IT leadership, marketing leadership, chief digital officers are all stepping in and saying, ‘We have to get our arms around this quickly and have a good, smart strategy for how we’re gonna to go about it.’
Guy Nadivi: I read in 2017 that Gartner predicted a couple of developments that I think many people in our audience will find startling. They forecasted that, “By 2019,” this year, “20% of brands will abandon their mobile apps,” and, “By 2020, 40% of all mobile interactions will be via virtual assistants,” which I think of course means conversational AI, or at least partially conversational AI. Matt, my question for you is, have we reached peak app and started transitioning towards conversational AI as the predominant user interface going forward?
Matt Smith: I think if we look … Two years ago, making this prediction, I think it was a realistic way to look at the way things were trending. I would flip it around today though, in the middle of this first wave that Gartner was predicting. I don’t know that it’s the brands are necessarily abandoning their apps, it’s that the users and customers are looking for alternatives to apps. We’ve all seen the stats of how many unused apps everybody has on their phones, or their tablets, how much functionality within a given app people never even are aware exists, much less take advantage of. I think users have just reached app fatigue, and they’re looking for or hoping for a better way to engage.
You parallel that with the use of messaging as a preferred way to communicate for a variety of situations. Certainly a generational or demographic trend favors messaging for lots of users. But even in the case of, like I talked earlier about that contact center scenario, where users of any demographic were given the choice on their own to opt into a messaging approach, versus waiting online, or maybe fumbling through the mobile app. The uptake was 80% to shift to messaging. There’s plenty of data now showing that messaging is becoming rapidly the predominant way for engaging with customers in this asynchronous method. I think we are seeing that really start to play out.
As far as calling conversational AI the dominant user interface, I don’t know that I would say it’s gonna become the dominant interface. I think it’s gonna be one of the primary ways that people will interact with technology going forward. We’ll look back on this 25- or maybe 30-year era where the way you would get information into or out of technology was with a keyboard, and then a mouse, and then a touchscreen. We’ll look back at that and say, “Well boy, that was clumsy.” Much like we look back at blackberries today and go, “That was the way you would send a message?” And we chuckle about it, but it’s in hindsight.
I think conversational AI will be the same in some situations where it does become predominant for certain uses where it makes the most sense. In other cases we’ll see other forms of probably AI-driven user interfaces emerge as dominant in those eras.
Guy Nadivi: Do you think conversational AI can eventually extend beyond traditional computing devices? Could we see voice enablement eventually found in everyday things like product packaging?
Matt Smith: You don’t have to go any further than the Consumer Electronic Show this year. CES in January 2019 was all about conversational AI. It wasn’t always called that, but that’s what was the dominant theme across CES.
It’s happening now. Whether it’s in home, all the ways that you can drive your home automation, or home security, or air conditioning and heating systems, or lighting is voice controlled in more and more situations. Or you look inside of vehicles and the idea of avoiding distracted driving using voice assistants for everything, from navigation to selecting the media that you listen to on the dash, to conducting basic transactions perhaps on like we talked about earlier, information requests, or placing orders for fast food on your way to work or to run an errand. We’re seeing conversational AI happen there.
Then in workplace solutions. There’s going to be more and more examples of voice replacing other forms of interaction. If you think about the way most of us would get into an office building today, where we would go to work, you swipe your badge. Well that could easily be replaced with voice identification instead of having to tote the badge around, or looking up the information in the conference room, for example, and connecting to all the systems. We’re putting those solutions in place for clients today.
In healthcare there’s lots of examples of voice automation starting to make its way into patient rooms, or even operating rooms, or labs or clinical environments where you want to be able to operate without your hands, where you want to engage with the patients, or you want to maintain a sterile environment. We’ll see lots of those kinds of big examples too become more and more prevalent as voice starts to make its way. The idea of extending the untraditional computing is happening really fast.
Guy Nadivi: Matt, you’ve talked about some amazing advancements today, but AI and automation sometimes frighten people from a job security standpoint. What are some ways you think conversational AI will make employees more valuable to their organization?
Matt Smith: Well it definitely does, but I think it’s interesting that the term AI by itself for a lot of certainly popular media, it’s this dark robotic future, and it speaks about the presence of these robots everywhere, and job loss, and loss of control and personalization. But when you put the word conversational in front of AI for just about every situation, it seems like it becomes a much friendlier situation.
I think we see lots of examples, Guy, where the idea of conversational AI makes a workforce far more productive and able to do things that are much more meaningful and enjoyable, just like if you look at intelligent automation. What that’s been able to do with very menial tasks like data entry, or screen scraping, or mouse clicking between forms, conversational AI can automate much of those same kinds of processes. The difference is, in a lot of cases, intelligent automation or earlier forms of RPA required very structured rules-based processes, and structured forms, and structured data. Conversational AI can exist in a world where there’s not as much structure to the process. It’s natural language versus structured language, and so it gives us more flexibility and creativity.
But we’re seeing lots of examples. Back to the contact center scenario, conversational AI can be just as beneficial for a contact center agent to deploy in the background of supporting a customer so they can anticipate better what the right response or the right answer is to a question, or they can have a conversational AI system working in the background, pulling up information for them, or pulling up data that’ll help them provide a better solution while they are talking to the customer and understanding exactly what it is that they need or what challenge they’re facing.
Those are some of the examples. If you think about the insight and the analysis that conversational AI can start to open up for organizations, because if it’s hard, it’s a significant amount of data. Within that data is often times it’s better solutions for customers, it’s faster resolution, or it’s a quicker identification of the right product or service that somebody needs. We see that working in tandem now with a lot of workforces.
As customers, we may not even know that there’s a conversational AI system side-by-side with the agent we’re talking to, but increasingly it’s there, as part of the background supporting them.
Guy Nadivi: Now you mentioned structured and unstructured processes, and at Cognizant, you’ve developed some best practices for conversation design. Can you expound a bit on what factors differentiate a successful conversation from one that leaves customers frustrated?
Matt Smith: Yeah, there’s a few that we think are really important upfront. One of the early lessons that we learned, if you look back a couple years ago when conversational AI was just getting started as a business idea, it was usually technology first, and customer expectation much later in the process. One of the things that we stress is you have to start first with what is it that your customer wants to do, or where do they want to engage, and what are their expectations from a conversational AI solution. That’s one. The data availability for training the solution is critical. You could have with my team, on the surface a really cool use case, but if you don’t have the data to train the ML, if you don’t have the data to enable the chatbot, or the voice assistant to have a deep and broad enough data set to reference against, it won’t be a good solution.
Another key factor, Guy, is what applications in the backend of our enterprise will we need to have access to, and how easy or how complicated will it be to connect a conversational AI system to these backend applications that run our warehouses, or run our inventory systems, or run our invoicing and billing, or run our customer care? Those backend enterprise apps are still driving the organization, so you have to connect to them.
Then finally, I mentioned early, in certain context and particular in certain industries, you have to have a discussion around security and privacy at the very front of the project, and makes sure that that’s not something that gets neglected, or passed over until you’ve done too much work, and then you’ll find that you’ll have to do a lot of backtracking. Those are some of the real key factors that we see.
Guy Nadivi: At this point in the evolution of the conversational AI market, there’s multiple consumer platforms like Apple Siri, Google Assistant, Amazon Alexa, etc. Do organizations have to conduct separate development efforts for each platform, or is there a common API they can write to?
Matt Smith: The answer is that there’s parallel work efforts that will have to happen as organizations look into addressing different devices or different brands. Today, that’s simply one of the limitations or challenges of implementing conversational AI. It’s an important factor because again, it comes back to where are your customers? Not everyone uses Siri, not everyone uses Alexa, not everyone uses Google. Some homes have a multitude of those devices, others are built around just one, and some people don’t use them at all, they prefer to use messaging or chat.
The point is, you have to have really a multi-device strategy, a multichannel strategy, and increasingly with conversational AI, a multi-modal strategy. If you think about some of the newer voice interface devices where it’s not just an intelligent speaker, but there’s some images that are included as well, Echo Show or Google Home Hub are great examples on the consumer side where it’s multi-modal now. Voice-driven, but you have the ability serve up and show content.
Some of the work is reusable. Certainly the strategies are reusable. Customer expectation, that kind of work is reusable, but when you get down to a code level, often times you’re going to be coding for different channels, different devices. And like mentioned, certainly different ways or different modalities of interface. You have to think through all of that before you undertake a project.
Guy Nadivi: Could be a great market opportunity for someone to come along and….universal interface for all of the …
Matt Smith: Yeah. Well, and that’s a good point. I think we’re going to start to see, maybe not on the coding side right away, but from a user’s perspective, we’ll have the lessons learned from the way consumers are looking at these technologies too. Eventually they’ll say, “Well I don’t want all these different. I just want my favorite virtual agent, so my agent will drive all these other different technologies or interfaces.” I think we will see some of this start to simplify itself from a consumer perspective for sure.
Guy Nadivi: What are some of your predictions for conversational AI over the next three to five years?
Matt Smith: There’s really three big predictions, or three things that I think we’re in fact seeing them happen right now. One is virtual agents will become the preferred interface versus apps. Whether it’s Siri, Google Assistant, Cortana, your virtual agent will be increasingly personalized to you, and to how you go about your day, whether it’s your work day, or your personal day. That virtual agent will engage initially with the apps on your behalf, but at some point I think we’ll see a crossover where virtual agents will engage with other virtual agents or directly with systems. The app will fade more and more into the background, and maybe become just a specialized tool.
The next is, as far as big trends, is the idea of contextualization of the process. That’s where the capabilities of AI, and all these super computing solutions really come into play, and we see it today. We hear examples of people say, “I got in my car and Siri popped up and said, ‘It’s gonna take you 25 minutes to get to the office today,'” and you go, “Wait a second. How did it know I was gonna go to the office?” Well it’s contextual. It’s running in the background and it knows that on Monday through Friday, Guy gets in his car at 7:00 a.m., and he goes to work, and it knows where you start and where you stop. Increasingly, it’s telling you, “Well look, today there’s construction. You need to go a different way.” I guess contextualization is gonna become more and more accepted and expected.
Then the third is we’ll see conversational AI everywhere. In a sense, we won’t see it at all because it will become more and more transparent. The idea of voice interfaces will just be the way that we engage, whether it’s with our technology, or our homes, or our vehicles, and even in our workplace. It’ll just become part of the backdrop of how we engage. Contextual, everywhere, and virtual agents I think are three trends to really watch for.
Guy Nadivi: For the enterprise IT managers, who have never dealt with conversational AI, what should they know before deploying it?
Matt Smith: Steve Jobs probably has the most appropriate quote, and he was talking for a different era, but I think it applies really well here which is you’ve got to start with the customer experience and work backwards towards the technology, not the other way around. I think a lesson that we would impart to, whether it’s IT leadership, or marketing leadership, or brand digital officers is just that. Focus on the customer first. That’s one.
Link your projects as much as possible, which means for leadership, bring together different teams, business teams, IT teams, cross functional groups that all will have their perspective on where conversational AI can come into play. Again, think beyond the idea of chatbots. Most companies, I think, are today thinking past the point of a website, text-based chatbot, and think about other ways that you’re going to be using conversational AI. Again, it’s that omni-channel, multi-device, multi-modal future that is becoming the new user interface.
Another important point I think is look for innovation partners that can help not just from a project delivery perspective, but also in terms of supporting the strategy of how these technologies are being deployed, and where they’re going, and can help think about the road maps that should be built along the way. You’ll save a tremendous amount of time if you’ve got those right partners in the front end, and I think you “de-risk” your projects in many ways by doing that with partners that can help from an innovation perspective, as well as do, call it basic deployment capabilities.
Guy Nadivi: Matt, you speak with customers all day long about conversational AI. What is the one take away piece of advice you have for CIOs, CTOs, other IT executives considering or already deploying conversational AI in their environments.
Matt Smith: I’ve said it several times, I think, but it’s really the voice of the customer. The projects we’ve seen, Guy, that have the best success started with the customer, not just thinking about, but talking to customers, and looking at data of past customer interactions, and testing out in really small MVP, so Minimal Viable Project or Product approaches before trying to scale and get too complex. Customer first, and go bite-sized pieces, very short spread, and iterate as you go. Those are where the companies that we work with have followed that plan. They have very good success. When they don’t, they’ll see the adoption just isn’t what they expected, or even worse, the reaction from customers is to let them know on social media or other places that they hate their new chatbot. If you start voice of the customer, I think you’re gonna be very well served in this new user interface era.
Guy Nadivi: I think that’s excellent advice, even outside the realm of conversational AI.
All right, looks like that’s all the time we have for on this episode of Intelligent Automation Radio. Matt Smith, thank you very much for joining us today, and sharing your thoughts about the current state of conversational AI. We’ve really enjoyed having you as our guest.
Matt Smith: Guy, I thoroughly enjoyed it, and look forward to talking again.
Guy Nadivi: Matt Smith, Conversational AI Practice Leader and AVP of Cognizant. Be sure to get in touch with Cognizant to learn more about their services.
Thank you for listening, everyone, and remember, don’t hesitate, automate.