PwC's Tech While You Trek

PwC's Tech While You Trek: Conversational AI

Season 1 Episode 27

Tune into another episode of Tech While You Trek to hear PwC Senior Managers Skylar Versage and Ernesto Valdes discuss Conversational AI as it refers to solutions like chat bots and voice assistants. Users that want to interact with products and services, using natural language as a form of communication, seek to utilize conversational AI. 

Tech While You Trek - Conversational AI

Release Date:  4/19/2021


Adam (00:08): 
Hello, everyone, and welcome back to another episode of PwC's "Tech While You Trek." I am your host, Adam. And today I have with me, Skylar Versage, and Ernesto Valdes to talk about conversational AI. Gentlemen, welcome, and please introduce yourselves for all of our listeners.

Skylar Versage (00:23):
My name's Skylar Versage. I work in the Emerging Technology Group, primarily focused on artificial intelligence machine learning with the focus on natural language processing, which deals with a lot of text-based analysis

Ernesto Valdes (00:37):
My name is Ernesto Valdes. I'm a senior manager at PwC and, as Skylar, I work on the Emerging Technology and Innovation Group. I have been with PwC for more than seven years now. My background is computer science. Since my major, I've always been interested, not only in the how computer works, but also how to build software and how do people interact with software.

Adam (01:01):
So can you please give our listeners an overview of what conversational AI is and some of the technologies that make it possible?

Ernesto Valdes (01:10):
Conversational AI refers to solutions like chat bots and voice assistants. Conversational AI allows users to interact with products and services, using natural language as a form of communication. Done well, this technology can help people interact with complex systems in a faster and easier way, and also help business deliver personalized engagement experience, and support at scale.

The technology side of it behind conversational AI, we think of it in tiers. The first one is the input or communication channel. The main communication types that you can think of for conversational AI will be speech and text commands. You can find this tech today in smart speakers, also mobile devices. You can find it on cars, which is one that I personally love, including TVs, also one of my favorites. The second component within conversational AI will be the brain of the solution. That's the AI part of it, the artificial intelligence, these solutions use data or sample conversations to train the machine learning model that leverage natural language techniques in order to extract valuable information.

Adam (02:22):
That's something that's very interesting to me is that inflection point between a human being's attempt to share an idea and the machine's ability to understand what is trying to be communicated. Talk a little bit about how things are different from digital assistants that we can talk to on our smart speakers and the like.

Ernesto Valdes (02:42):
To be honest, anything that you interact through speech, text commands, solutions that are capable of understanding requests. And to clarify, this is away from rigid, unstructured conversations like, "Say 'one' for this," or, "Say 'two' for that." Because these are easily developed with code, right? You can parse that conversation and understand what the user means. But conversational AI is, any time that the service understand unstructured conversations or requests from the users and is able to identify intents from that conversation using machine learning. The only difference that I will point out is that today's trends indicates that adoption of smart speaker is very high on only keeps growing, but the common use case is still, "What's the weather?" "Play music," which are all within the customer entertainment industry. Where we see the potential at PwC is pushing the boundaries and exploring beyond those use cases. What is beyond playing music? What is beyond asking the weather?

Adam (03:45):
Well, so, along those lines, what are some of the common business use cases for conversational AI?

Skylar Versage (03:51):

Typically, as Ernesto mentioned before, a lot of what people think of when they think of conversational AI is the chat bot. And that is a great jumping off point for businesses who are looking to, one, adopt AI in general, because there's a lot of value to be had but then, two, there's a lot of lower-hanging fruit in the sense of getting people to adopt it and see some quick wins. So with that, we like to organize around three primary areas when we talk with clients.

The first one is kind of what you think of when you think of chat bots, that Q and A. So you come in, "Hey, what are your opening hours?" Or "What's my vacation balance?" Maybe internally if you have an internal chat bot or conversational agent.

The second one, and this is what's coming newer onto the scene is more of a task-oriented. So the conversation keeps a record of what's been going on. So something would be, "Hey, I need to book a flight." All right, that, in and of itself is the intent, that's what you want to do. But also it helps to facilitate the conversation to say, "When do you want to fly? Where are you flying from? Where are you flying to?" And we're seeing that start to be adopted more broadly as some of the machine learning artificial intelligence technologies start to catch up to facilitate a longer conversation.

Adam (05:00):
Well, it certainly feels like we have been properly trained how to speak to certain personal assistants and intelligence is in our smartphones. It's molding us to them as opposed to the other way around. You're talking about the other way around.

Skylar Versage (05:13):
Certainly, especially for the younger generations. There's this example I've recently seen where two adults were talking about, "Well, what time does this store open?" And they were getting their phone out and they were getting ready to search for it. And then their kid, without thinking, asks the smart speaker, "Hey, what time does this store open?" So your point's well-taken because, how we interact with it being a little bit older, we're getting used to it. But for the younger generations, it's just sort of a way of life.

Adam (05:37):
I have a dear friend whose one-and-a-half year-old started responding to the name of the home assistant before he started responding to his own name.

Skylar Versage (05:46):
So the first two, question and answer, the second being more task-oriented, where there's a natural conversation. And the last one, this is a little bit more suited to business than it would be consumer-facing or what we would see in our everyday lives out there, but it's called troubleshooting, or "Next Best Action." And a good example is actually if you go to a doctor and the doctor doesn't just start by going from the head to the toes. He says, "Well, what hurts?" And you could say, "Well, my knee hurts." Well, then you automatically start to ask questions that are tailored to, or localized to, whatever area you have a problem with. So you could think about in a business, especially something like, maybe, construction or an internet service provider, if your modem has gone out and it's not working, you don't just start and ask a whole bunch of non-related questions. You ask questions based upon what the facts and circumstances are right there. So we call that, "Next Best Action" or troubleshooting.

Adam (06:34):
So talk to me then about some of the challenges that companies are typically running into as they try and adopt and implement this kind of technology. Go down a pros and cons list for our listeners.

Ernesto Valdes (06:44):
It is very easy to run into challenges while building conversational AI. I'll start with two, to get the conversation going. Having little data. Building a virtual agent with little data is one of the main challenges that we see within the enterprise. AI requires data in order to be trained and it's common to find use cases where people only have one question and answer per. It is ideal to have a variety of questions so that you can diversify the training for the AI in order to understand, successfully understand people once it's deployed.

The second one is design. Conversational requires simplifying the information. You don't want to push to a user a 24 pages content as part of a response through voice. Also, conversational is more a way of interaction. It's not a one way conversation. You want to make sure that you build solutions that are engaging with the user as Skylar mentioned. So the conversational assistant should be developed and designed in a way that considered many challenges within the conversational path.

Skylar Versage (07:50):
Scaling's an interesting one because it's so easy to get out of the gate and see quick wins with conversational agents, whether you're using some sort of cloud-based tooling or whatever, it's easy to fall into the trap of thinking that, "Well, if I can work this with a hundred people, this will work with a thousand people." Not always the case. A thousand people, as Ernesto articulated with the data and the design, people can ask things different ways. So the way that they would articulate a question, if you've got a new way someone's articulating a question, when you expose it to a new audience, your system's going to trip up. And then when it trips up, they end up calling the live agent or writing in and...okay, well now it's a degraded customer experience.

In addition to the scaling piece of it, is you need to account for how you actually want to adopt it and make it more pervasive across your business. Up front, you might just say, "I want to build a chat bot to allow someone to book air travel." Okay. Well, what happens when you want to expand to trains or buses? If you've built it up front, just to say, "I want to book air travel" but then in a few months you want to add trains, if you haven't designed your system to be modular, in a way, it really starts to trip up and then your system can get confused very easily. So, that's the scaling piece of it.

For the trust aspect, it's particularly germane for conversational AI, although it's a hot topic across, really, any kind of area where there's not a human behind the wheel, if I can say it like that. As the technology gets better, and people are aware that they're interfacing with a conversational agent, how do you appropriately call that out and foster that trust so that way you don't get into this gray area where all of a sudden the customer's saying, "Wait, what the heck was that?" or "What are you guys recording?" or whatever the case is.

Adam (09:27):
Talk to me, then, a little bit about how conversational AI has changed just in the last year as we've all had quite a year.

Skylar Versage (09:37):
So, within the last year, lots has changed, right? There's been a few paradigm shifts. People are working from home more often. And so with that, the way that people would have perhaps historically gone into a physical brick and mortar location, maybe now they're trying to call or contact in a different way. So it's really had to expedite this shift to adopt conversational AI to, one, keep that customer experience something that is not a point of frustration.

The second piece, and this is always going to be the case, but particularly in the last year for this natural language processing/conversational AI is, the technological/algorithmic advances are just starting to exponentially advance. And so, before this, what had really gone through sort of a reckoning was what's called, "computer vision." So, if you think about any social media app where it'll put a box around a face and perhaps suggest who that person is, NLP is going down that path now where academia is starting to offer a lot of technological advances, which is enabling.

And then, lastly, the way that younger generations are starting to adopt and have conversational/virtual assistants be in their life is sort of ubiquitous. I think that's really going to impact the arc in which it's adopted.

Ernesto Valdes (10:44):
Another change or evolution of conversational AI, it's intersection between conversational AI and other technologies. The first one, think of Internet of Things. Devices keep getting smaller and smaller and the interface, the way-providing IoT for users to interact is becoming more voice-first than before. But the other one, and this one is the one that I'm very passionate about, is virtual reality. For a while now, since it started, VR has been all about controls, but now we are moving towards that conversation in order to make it a more immerse interface. So, there is a bigger intersection now between conversation and multiple other technologies as the favorite way of communication.

Adam (11:37):
Well, listen, gentlemen, before I let you go, I've been kind of firing off a fun question at all my guests this season, you guys got time for one more real quick?

Skylar Versage (11:43):
Oh yeah.

Adam (11:44):
What would the "you" of, say, 10 to 20 years ago be the most surprised that the "you" of today is using from a technological standpoint?

Ernesto Valdes (11:52):
20 years ago, I switched from my four-year major of mechanical engineer into computer science major. And I knew that there was a huge potential between physical software and experience work colliding, but I never thought voice was going to be that medium. But more than that, I'm just surprised that I'm part of it. I have an opportunity to make an impact within PwC and within that domain.

Skylar Versage (12:19):
For what has surprised me the most, I would say it's virtual reality. And I was even a, probably, a skeptic as of a year and a half ago. I was skeptical of what value it brought over and above even a video meeting itself and using it for meetings, but then also seeing some of the implications it could have for education and how it can bring quality education, or even just quality content, without having to physically be somewhere. I was a convert.

Adam (12:42):
Listen, thank you guys so much for your time and for stopping by today.

Skylar Versage (12:45):
Thanks, Adam

Ernesto Valdes (12:46):
Thank you, Adam.

Adam (12:47):
This has been another episode of, "Tech While You Trek." I have been your host, Adam, and we will talk to you again next time.

Speaker 4 (12:57):
This podcast is brought to you by PwC, all rights reserved. PwC refers to the U.S. member firm or one of its subsidiaries or affiliates, and may sometimes refer to the PwC network. Each member firm is a separate legal entity. Please see www.pwc.com/structure for further details. This podcast is for general information purposes only and should not be used as a substitute for consultation with professional advisors.