We've already gotten comfortable with digital assistants in our everyday lives. Raise your hand if you have an Amazon (NASDAQ:AMZN) Alexa or a Google (NASDAQ:GOOGL) Home sitting within 20 feet of you right now.
These voice-based digital assistants are becoming more commonplace, but they still run off of simple, rule-based systems. The machines listen to the amplitudes of your voice, convert your sound waves into text, run structured queries to understand what you're saying, and then run software to provide the answers you seek. Behind the scenes, Wolfram Language is the secret sauce that bridges your human thinking to the machine computation.
But there are also limitations, which anyone who owns an Alexa or Google Home is already well-aware of. To be frank, they're still pretty "dumb."
The current wave of vocal assistants have no problem with simple tasks, like playing songs or telling you about the weather. However, they struggle in responding to more complicated questions, or understanding if you try to speak in a way that, well, humans speak. We use idioms, figures of speech, and context, and those tend to get lost in translation.
The bottleneck in the process has been conversational artificial intelligence, which is what would make machines able to understand humanlike conversation. Some would say that the lack of progress here has been what's holding back the adoption of digital assistants in the workforce.
Companies would certainly like to unleash AI on resolving customer issues, or in providing automated sales quotes. But there's a huge price to be paid if the bots get things wrong. Irritated customers with unresolved claims could take their business elsewhere, and inaccurate information on a sales bid could lock a company into unwanted financial or legal liabilities.
If you build it, they will come
But even with the challenges, there's a huge opportunity for AI in the enterprise. Gartner predicts that by 2020 85% of customers will manage their relationship with a company without actually interacting with a human being. A recent Oracle (NYSE:ORCL) report said that 80% of large companies plan to have chat bots implemented in the next two years.
This is a huge problem that needs an innovative solution, and companies are beginning to rise to the challenge.
Artificial Solutions is one company addressing the opportunity, providing a conversational AI platform that can better communicate with humans and is ready to serve the needs of enterprise customers. Those commercial needs could include being deployed in multiple languages, or the ability to mine actionable insights from hours of natural customer conversations.
I recently spoke with Andy Peart, who is Artificial Solutions' Chief Marketing and Chief Strategy Officer. Andy describes conversational AI as "having a focus group of millions of people, 24/7 and 365 days a week." That's certainly a powerful concept that could have incredible business implications.
In our conversation, Andy defines where conversational AI is being deployed, describes how smaller companies could actually have an edge over larger rivals, discusses the role of data privacy, and points out a few things that investors should be watching.
A full transcript is provided below.
This interview was originally recorded on August 14, 2018.
Motley Fool Explorer Lead Advisor Simon Erickson: Hi, everyone! Motley Fool Explorer Lead Advisor Simon Erickson. This morning, we're going to be talking about conversational artificial intelligence.
Just to frame this a little bit, 85% of customers say that they'll manage their relationship with a company without interacting with a human being by the year 2020. On top of that, an Oracle Report says that out of 800 large companies polled, senior leaders say that 80% of them plan to have chat bots implemented in the next two years.
With that frame in mind, my guest this morning -- or afternoon, if you're in the United Kingdom -- is Andy Peart. Andy is the Chief Marketing Officer of U.K.-based Artificial Solutions. Hey, Andy, thanks very much for joining me today!
Artificial Solutions Chief Marketing Officer Andy Peart: Thank you for inviting me, Simon.
Erickson: Andy, let's start this by talking a little bit about conversational AI. You all have an AI platform called Teneo, Artificial [Solutions] does. It's a natural language interaction interface that can communicate with human beings in a variety of different ways. Can you start us by talking a little bit about what that is?
Peart: Yes, absolutely. Well, conversational AI is a form of AI, and [it] basically allows people to talk to devices, to applications, to websites, in fact a wide range of different channels, in a very human-like, intelligent, and conversational manner. And that conversational bit is absolutely crucial, because we as humans love to use our own terminology. We like to go to the different tangents. You need a system that can understand you as a human, and that's exactly what conversational AI is all about.
Now, we have chosen to address this rather thorny technological issue by building a platform. As you rightly said, Simon, it's a platform called Teneo, that allows our customers to build these conversational systems. Once built, they can be deployed in multiple languages, across multiple platforms, and supporting multiple channels.
But there's just another point that I'd like to flag, and that's to do with the conversational data. Because when you and I, and many other millions of people speak in conversational ways, they reveal an awful lot about what they're thinking. If you can capture that information and retain the context of those conversations, it's like having a focus group. But not a focus group of 20 people for an hour. Potentially, a focus group of millions, 24/7, 365. And you can start to get a real insight as an organization, into what your customers are looking for, what they're thinking, what they're feeling, so that you can then start to really target your products and services much more accurately.
And that's what conversational AI is all about.
Erickson: Sure. Andy, I'm really glad that you brought up the part about it being conversational. Because this isn't just crunching numbers. This isn't just listening to customers in an NPS value score, or customers having a drop-down menu of what they want.
This is actually conversation. This is actually listening to customers, in their own words, about what they're saying about businesses, which is a different form of analytics than I think that we've gotten used to.
One of the initial early adopters of this is chat bots, correct? Can you tell us a little bit about the intelligent assistant market?
Peart: Yes, absolutely. I mean, there's lots of discussion around what you should call these things. And to be honest, I'm not so fussed about that [laughs]. Many people are now kind of falling into the term "chat bots," as kind of an all-inclusive term for virtual assistants, digital employees, chat bots, whatever you call [them]. But basically, these are automated capabilities, devices, that can understand you and I as humans, will run across different channels, and be able to do things, be able to answer your queries.
Now typically, they fall into three broad areas. One is using them for customer service-orientated activities. That tends to be the most popular at the moment, where you can automate a proportion of the calls that would otherwise have gone through to a live agent, and handle them through automated channels in an intelligent way that users are happy with.
The second area is to do with online sales support, where you can use sales advisors -- that, again, are these automated capabilities -- that will help give input and sales recommendations in a natural, human-like way.
The third area in which this sort of technology is used is providing voice-enabled interfaces. So you're really differentiating your product or your service by providing a different form of interface. And those are the three broad areas in which this sort of technology is used.
Now, chat bots typically tend to be deployed across messenger services. And, by that, they tend to be more text-based rather than voice, and unfortunately, can be a little bit dumb. They're built to follow a linear process, which is fine if you as a user follow that exact same process that the developer designed the chat bot for. But as I said earlier, we as humans tend not to do that. We branch off. We ask questions. We want to come back. And that's where these initial, stage-one chat bots have tended to fall down, and maybe give the whole sort of industry a bad reputation.
Erickson: On that note, I would definitely echo what you just said. We have an Amazon Alexa in our home. It seems to be much more of a vocal command prompt than, I would say, a conversational artificial intelligence platform. But I would say that I'm not a unique case in that. You know, a lot of us have gotten used to the frustrations of not saying things correctly, not triggering or getting the information that we would want to out of one of these vocal assistants (or virtual assistants), as you said.
But there are a lot of big companies that are in this space, Amazon being just one of them. Facebook has Messenger. Microsoft, Cortana. Apple's even got the Siri technology. These go back for years, Andy.
These companies have got very large user bases of over a billion people, oftentimes. And now they're starting to go after enterprise customers as well. But as you said, they've kind of got this -- stigma, maybe, is the word -- that's already been established from the early wave, that aren't working as well.
What is the edge, if you believe there is an edge, of start-up companies and smaller companies that are innovating in this field, as opposed to those larger tech companies we've gotten used to and are familiar with that have those large user bases already established?
Peart: Well, a really good question, and most definitely there are edges that the start-ups have.
Firstly, it's worth pointing out that the likes of Apple with Siri, and Amazon with Alexa, and some of the others, have done a great job at opening up this technology. As you said, there are billions of conversations taking place now, and it's ensuring that people now understand the sorts of things that are [possible]. Indeed, we did some independent research recently about the use of voice assistants and found that 69% of people surveyed -- and this was a fairly meaningful survey of 1,900 different users -- 69% are already using voice assistants at least once a day. And that use is growing. 49% are now using it more than when they first started. So, a lot of positives in there.
But the research also pointed out a bit of a failing, and it pointed that 70% of users would like their systems to be much more conversational. That's what you alluded to in the question, Simon. That Alexa and others are still fairly dumb. It's OK if you're saying, "Play this," "Turn the music up," but they're not conversational. You can't say, "What was Ed Sheeran's latest album?" "Such and such." "Oh, yeah. Play the second track of that." It just wouldn't understand, and we as humans expect it to retain the context.
So, what can start-ups offer? What can Artificial Solutions offer?
Well, one of the areas is to provide a much more compelling conversational capability. But, what they can also do is focus. We have, for example, chosen to focus on the enterprise. Now, by that I'm meaning building both enterprise-strength solutions, enterprise-strength conversational systems. But also, the platform Teneo is enterprise strength. It has the sort of features that you would expect of a CRM system or a financial system. But because we're operating in the new technology, the majority of the solutions out just haven't addressed this yet.
Let me bring that to life by giving an example. At the beginning, I talked about the fact that we support multiple languages. Cool. That's great. That probably means that you are deploying, as an enterprise, your solution, across multiple geographies, in multiple languages. Hence, you're going to have distributed work groups, so you need the ability to be able to allow them to check in and check out part of a solution, part of a flow, and lock it while you're working on it, then release it. You need things like automated testing. You need the ability for things like rollback. If something goes wrong, what do you do? You need version control. You need the ability to be able to host those on-premise, and if you wish, in the cloud. Because security is crucial for these enterprises. You need to think about what is happening to the data that's collected. Who owns that data? Who's learning from it? These are just a few of the sort of enterprise-strength considerations that need building into a solution, the sort of things that large enterprises are looking for in a conversational platform.
So, your question was, how can you differentiate against the FAMGA, some of the big players in the market? Well, one of them is to be very, very focused and really understand the needs of your audience. And in our case, it's those large enterprises that need enterprise-strength capabilities.
Erickson: Sure. And then, how do the enterprises act on this data? You know, in addition that you're collecting so much information, the insight is probably the most important thing, if you're a large enterprise customer.
Is it dashboards? What's the interface that you guys offer to the enterprise to make sense of all of these billions of conversations that are taking place out there?
Peart: Well, again, that's a recognition of the sort of market that we're targeting. So, enterprises almost certainly will have a range of BI tools already in place. So what we do is focus on the bit that we're good at -- the conversational analysis, the understanding of that data -- and we can then surface it through a range of different BI tools, whether that be Tableau or Qlik Tech or Excel. It could be a range of different tools.
So, we don't get hung up on the presentation layer. What we focus on is the ability to interpret conversational data. Which is pretty difficult to do, because you have to retain the context of that conversation. "I love that." Well, what are you talking about? Are you talking about that red shirt that you were looking at? You know, am I looking out the window and talking about the weather? You've got to understand the context of the conversation, to then be able to bring meaning to it.
And that's not easy to do. You need to understand things like sentiment. You need to understand things like synonyms. So if, you know, "That's sick." "Oh, dear. That's not very good, then." "No, of course. It's great." You know? You need to be able to understand language, and that's what we're able to do, and present it through the relevant tool set.
But just to jump in, Simon, one other area: We talk about how it brings insight, and that is absolutely crucial. But, this conversational data can be used in other ways. It can be used to then start to tailor the responses that are given back to the user. So you're starting to personalize the response at a conversational level. That is absolutely fundamental for enterprises, because if they're wanting to build a relationship through automated channels with their customers, you need to get personal. And that's what data can do.
Erickson: On that note, Andy, of data getting personal: Maybe if we step back and go a little bit higher, maybe to the 1,000-foot level. Not 10,000-foot, but maybe back to 1,000 foot. Data privacy has been kind of a hot topic right now, right?
Peart: Most certainly, yes.
Erickson: GDPR in Europe, threatening 4% royalties for companies that don't comply with privacy regulations. Here in the States, of course, Facebook's been highly scrutinized for sharing information with others. This is a very important political question, I guess, for the industry that you play in: What is the impact of regulations on how you're innovating this AI field?
Peart: Absolutely fundamental, and it's going to get more and more complicated as new legislation comes out, as well.
It's a key area for us. In fact, we were the first conversational platform to come out and say, "Look, we can support our clients in ensuring that they are fully GDPR compliant from a conversational data point of view." And how you do that? Well, it's not just one thing. It's a range of different things. For example with Teneo, the conversational data is stored in one place, which streamlines the querying and interpretation of that data. It's easy to identify any personal data using Teneo and then be able to delete it, if that's what you choose to do as an enterprise, or maybe to pseudonymonize it, so that you can still get insight into what your users are talking about, but it's not linking that data directly with an individual. So that you can still get value from it, you can still provide degrees of personalization, but you're not breaching GDPR policies.
Earlier on in our conversation, I talked about the ability to self-host. Again, this is crucial for enterprises, because it gives them the security and the confidence that they have the control of data. It's not going off to the cloud where, who knows? So there's many layered approaches that we've taken to this very, very thorny issue of GDPR compliance.
Erickson: Sure, absolutely. Don't let the tail wag the dog. Let the technology innovate and then comply with regulations, rather than completely doing things the other way around.
Erickson: One last question, Andy. Our audience is individual investors. [They] might not be experts in the field of artificial intelligence, but certainly interested in this field. What are a couple things that, as investors, we should be keeping an eye on in the AI world?
Peart: Well, really commenting on the bit that I'm very familiar with, and that's around conversational AI, and there is no doubt that this is a red-hot topic area. Within the enterprise, I would put my neck on the line and say this is a technology that, over the next five years, is going to become as important to enterprises as their websites are today. That's because people's expectations are changing, and they're changing fast. We nowadays expect to be able to speak to devices, to applications, to services, and for those applications to understand us, and enterprises are going to have to embrace this kind of technology. So, it's on a massive growth curve, and it's a key area to look at.
I think the second area is the ability to be able to provide differentiation, as we talked about earlier, in a certain sector. There are the large tech giants who are providing kind of generic platforms, that can do a vast range of different things, but they're not specialists in certain areas. I think that leaves an opportunity for a range of start-ups to be able to pick their chosen battleground, and win very convincingly in that area.
Thirdly, this sort of technology is going to be ripe for M&A activity. You know, it's the sort of thing that ... the large enterprise software providers, who are maybe late to the game, are going to want to get involved with. And how do they do that? Well, you could argue they go off and develop themselves. But, they're starting from a very, very late starting line, and there are many barriers to doing that. So, they're going to be looking at how they can quickly get into this area, and the obvious way is through M&A. So I think from an investor point of view, this is a red-hot and really exciting market to play in.
Erickson: Great, so conversational AI, differentiation in specific sectors, and watch for mergers and acquisitions. Those are three things for investors to keep an eye on. My guest this morning is Andy Peart, again, the chief marketing officer of Artificial Solutions. You can learn more about Artificial Solutions at their website, www.artificial-solutions.com. Andy, thank you very much for the time this morning.
Peart: Simon, it's been great. Thank you.
Erickson: Thanks for listening. Until next time.