Podcast | How low-code and pre-built tools are accelerating service automation

Jean
Shin

Director, Strategy & Content | Podcast Host of Mobile Interactions Now

27 min podcast

In this Part 2 of our discussion with Dan Miller, we delve deeper into the deployment side of conversational AI. Exploring various modalities and channels connected to vertical solutions and enterprise systems, we touch on the low-code, no-code and pre-built tools that would help enterprises simplify and speed up deployments. (Plus the reason why we set a high bar for chatbots :-)

Podcast Transcript:

Jean:

In our previous episode, we touched on the history and how different technologies are coming together to make self service work no matter what the company and business it happens to be. And now I would love to do a little deeper dive into what we are beginning to call conversational commerce. I don't know if it's your standard term already, but it still raises eyebrows a little for some, as in what are you really talking about? So I would love to have some basic understanding and then quickly get a little deeper into what it really takes, looking at the guts as well as if possible. So help me with the definition first. What is this conversational commerce?

 

Dan:

Sure. And then please understand that there's two schools of thought on this. And one sort of traces back to what I at Opus was doing back in 2011 when we held the first conversational commerce conference to bring together MarTech and sales tech is what I guess you would call it now. But this idea that we wanted to look at technologies that improve conversations that can be person to person, they can be person to machine, and that's sort of where we focus where speech-enabling IVR at the time was the big deal, but somehow bringing natural language, understanding, and machine learning to improve people's ability to use their own words to take over the resources that were either helping them buy stuff or helping them get support for stuff. Then the other thing were conversations that were machine to machine because they would get rid of the latencies in those automated conversations. They would also simplify, bringing dynamic information in for the automated bot, if you will, to do stuff.

Right around 2016, Chris Messina started writing about conversational commerce specifically about the power of messaging platforms and the messaging model for person-to-person communication and person-to-machine communication just as a more efficient conversation. And what it brought in was here's the user interface that people just totally get and it's asynchronous. It lets them use their own language. The expectation level is I can leave a message on a machine that could get answered by a bot. It could get answered by a person. It was primarily in text, but, I talked to Chris about this, that wasn't how... Everybody saw that the web world was so much more than a messaging platform. For some people, it's the first thing you look at when you wake up and everything lives in there. Your payments are associated with it. So the idea of having this conversational platform, a messaging platform, be soup to nuts where you get your information about the products, where you consult with friends about them, where you can consummate a deal. There's nothing more powerful than that. And I think our two visions they just dovetail really nicely, but I think that the machine learning and all the stuff that improves a machine's ability to understand what we say is vital, and I think that the idea that people are getting accustomed to using "conversational platforms" to accomplish their goals is evolving nicely together.

 

Jean:

Not as much these days because I think we have some global scale examples. But I remember a few years ago when we were having this conversation about things that is happening on messaging platform, I was often told to look at Asia, for example. Look at what people are doing in China on WeChat, look at how Korea is doing with KaKaoTalk and things in that nature. Now, do you see more globalization or are they still way ahead of that curve? What do you see there in country, rather continent to continent comparison?

 

Dan:

So with great disappointment, the Eastern model has not permeated here in North America, and Western Europe still remains something of a mystery to me. But I don't know whether it's in months or years we'll move to something that looks more like a messaging platform model. But I mean, I have so many... We're not going to be on the air, but I have so many apps on my phone. As a sample of one, it's going to take a long time for me to get to a state where I'm treating a single platform as my only source of information. So there is still a lot of fragmentation here. I think there should be a lot more backlash against it. I mean, I wrote a thing about what we're learning even in the smart speaker world where there there's developers around Alexa, there's developers around Google Assistant, and now there's Samsung each having different platforms for building essentially the same thing, which means that the world or whatever wanted... We either have to think about, "Oh, how do I develop my app once and have it rendered many times" which is the iOS versus Android thing all over again.

So I've given a really long-winded answer to say that what we have right now can't hold and some element of bringing it into one or two or three is what we're closer to. And the ideal of one, the one true conversational platform, is highly unlikely to go global, but there will be whether... I mean, there's a lot of planning going into what Facebook is doing around Facebook Messenger that they would ideally have you spend your whole time in [inaudible 00:08:04] just as WeChat, WhatsApp which is them, too. There is a world where you do spend a lot of time in one platform, but then there's the world we live in.

 

Jean:

And you're sort of justifying some of the hefty amount of money being paid to some of the messaging platforms and I think that is just interesting to see.

I remember seeing one of your reports, I think it's called A Decision-Maker's Guide to Enterprise Intelligent Assistance, and I remember seeing some of the verticals that we talked about in our previous session that there are definitely some of the industries that have been playing with this longer. But I just wonder when you look at across sectors, are there just common use cases or kind of a user scenario where without even looking at the first-party data, how much are we spending on this and how repetitive is the task for us, without really getting the data…believe it or not, I don't know about you, but a lot of customers that we have on the enterprise side, sometimes they have a problem with the siloed data. So some of these data is just not given. So if you were to kind of look at this from some of the horizontal cases where it would really make sense…to look into utilizing some of these technologies. Any thoughts on that?

 

Dan:

Well, I have tons of thought. So we're obviously in a point of transition. I used to write reports and say, "Oh, X is a silo buster", usually something around customer care or something in the contact center. I think more accurately what we're finding with the implementation of chat bots, intelligent assistants, they start in the context center, meaning that what we initially looked at in our Decision-Makers Guide was like a short list of technology providers that were bringing intelligent assistants into customer support in the contact center. And to your point about they tend to do appeal to that short list of industries, that usual suspect, and they did have and have built a fairly large a list of common activities because the first step with bringing in a chat bot platform was to discover when to ingest past conversations, discover what the most common calls were, match them to the right answers as we mentioned before, and guess what? The statistical model, you know what 80% of the calls are.

I remember actually I went to a vendor meeting with their largest utility customer and utilities as opposed to airlines or banks tend to be territorial in nature, so they were very much sharing their experience because they didn't regard each other as competitors. So they would say things like, "I can handle 80% of my calls within five categories: Somebody has moved and needs to initiate service. If there's an outage and I need to handle that. Somebody has a billing problem." I'm not going to go through the five; I don't want to waste our time. And those were common across every single utility and they all knew it, and then they would talk about, "Oh, what did we do to handle that and how did we answer it?"

And the same thing happens in, well, obviously, telecom is a form of utility and they have the same sorts of things. So to answer your question, yes, there is a narrow band of horizontal problems, categories of calls, and there is best practices and identifying the best way to answer them and we've had many years to compile those, so there are solution providers that Opus has studied that have built libraries of sort of right answers that they can bring. And bringing those to bear means that the nth company that comes along without sort of stealing data or anything from their competitors can benefit from the fact that we've all learned more over time and we can get up and running and get good accurate answers pretty quickly.

 

Jean:

That kind of makes me wonder about the data part of it again, because it really is about sharing information of it. I mean, you come from a tech background, you're used to seeing mar tech, sales tech, support tech. There're different stack…but at the same time as an end user, me on the other side, the quality of the experience that whether I'm satisfied or not oftentimes it really depends on the data that was available. I mean, in our previous episode you talked about somehow it was able to authenticate you. How did it do that? Yeah, after some coding it’s the data that came. So for me, what I also started seeing is that there's the data I have as an enterprise, the first-party data, and then there is data that could be third-party data that just makes these things simpler, whether there's authentication happening on another place. Now, we all live in API-enabled ecosystem in a way and we have a way to put it together. So this is part of, I think, a bigger conversation, because any time data is discussed there are many nuances and many things to talk about.

So let's talk about it, let's say that the codes have been written. Now, in terms of the data in order for us to really deliver a good end user experience…because a person like myself, I want this experience without knowing what it will take, right? The data that goes into, how do we think about…because the customer service scenario that you're describing is basically happening with what I have within my walls…I have those data. I know the customer’s previous call-in incidents and things like that. But in order for us to actually start going beyond that, talking about conversational commerce, being able to provide a way to say it comes in new trend colors, or something like that. In order for us to contextualize all these things, you're going to need data that is not necessarily sitting on your CRM, for example. So how do we even think about this as a whole…what is required in terms of data?

 

Dan:

Yeah. Well, the other way to look at it is some of it's in the CRM, but it has to be combined with all the other stuff in order to recognize you, match your intent with what you want to know. So we also coined a term called conversational service automation, and we chose that very carefully because we're seeing trends in robotic process automation, in sort of data discovery and data management, that are trying to answer that very specific question that you asked like, "Where does this...?", and this is from time immemorial that somewhere the companies have within their walls the "360-degree view" of their customers and they do have connectors or APIs through their back office. So theoretically, you could have this data lake that some rendering of information or some connect to that information or pointers live so that in real time or near real time it can be used to answer a question or inform an agent to answer a question because some fairly high percentage of queries into the back office system are done by employees, and a lot of the automation that we talk about helping self service helps employees help themselves as well.

So we're trying to crack this code. There is a lot of progress. I wouldn't characterize it as silo busting as much as recognizing that the silos themselves are poorest and there is a lot of points of ingress and egress to existing data in a company's knowledge management system, in the documentation for both. That could be employee manuals, it could be product catalogs, all that sort of thing. More and more of that is machine readable and you have technologies from something like an IBM Watson or sales force calls like Einstein that there's a lot of investment being made and get this, conversational AI. And conversational AI is applying sort of deep neural networking and other things in ways that will look at both structured and unstructured data regardless of where it lives and use it to inform bots. It's as simple as that, and it is taking place.

 

Jean:

So we are basically now into conversational intelligence.

 

Dan:

Yes. Conversational intelligence and conversational service automation, and they'll go hand in hand.

 

Jean:

This is a high bar because sometimes I cannot get that from a human being, either.

 

Dan:

Yeah. That's a really good point, and we'll make it a topic of some future discussion is do we have higher expectations for our robotic helpers than we do for people? And the answer is initially yeah, the expectation was a lot higher because people weren't going to tolerate having to repeat themselves so that speech-enabled IVR could understand what they were saying, and yet they would do it for people. So our tastes are changing and the conversational models are changing in parallel with that.

 

Jean:

Yes. I can't wait to see the whole relationship shift between the human-to-human and human-to-machine and see where we reveal ourselves more, and that will be another podcast session.

 

Dan:

Yes, exactly.

 

Jean:

It's an interesting time. Now, as soon as you get into this whole intelligent interactions that we're talking about, whether we have enough technical capabilities or data, but I think we have to say this before we go any further, do we want this? And how to, because there is a philosophical question now and then when we start talking about technology. I don't want to get into a deep debate over this. We'll save that for another day, but what will be the clearest way to frame this when we do face this question?

 

Dan:

So you're asking a profound question about whether... And hypothetically someone like me would... Under what conditions do I want to share information about myself so that I will get better service from some automated helper, and we haven't framed it that way yet.

So I'll give you a couple of examples. As an analyst, I go to a lot of vendor events and one of the vendors gave a group of 40 analysts Echos back three years ago. And then the following year... I mean, I took mine and set it up and happily tried using different things, trying to do different stuff. The next year I was asking the same group of analysts have they done anything interesting with their Echo and more than 50% said, "No, I would never activate one of those things in my house. Amazon is listening in on everything we say, and I just don't want them to have that." So there is awareness and lack of trust between people and technology providers, and so the barrier is how is the technology provider going to earn the trust of the individuals that they want to use their technologies? And the answer is that you have to be ethical from the beginning and you have to prove that you're ethical, and that's a really high bar. You mentioned high bars. Because not everybody is hoovering up... Well, I said many companies and it's sort of the Google syndrome. I use Google every day, but I am told that I'm a product for Google and I'm one of those eyeballs or whatever that Google delivers to advertisers and they're advertiser driven and that sort of thing.

So we just have to figure it out. I don't think that dev world is going to solve this problem, but I do want to counsel everybody developing a self-service chat bot or getting into digital self service to keep ethics in mind and to have if not a high bar, high standards, for how personal information is going to be used and we'll do the best we can.

 

Jean:

I can take that as a wrap. I think it's being just honest about the reality we face. I think sometimes that’s a great starting point.

So just before I let you go, I have a simple question that I ask all my guests at the end of the show, and here it is. Are you ready for it? What are the three things you use the most on your phone?

 

Dan:

Okay. To be honest, I'm a radio guy. So number one time wise is listening to our local PBS station through its app. And there you go. And then I'm a news junkie so I have three different newsreaders, but my latest one that I like is Flipboard, which you may or may not have heard of, and there's a voice services element to this because Mike McCue, who is the founder of Flipboard, was also the CEO of Tellme, which was one of the original voice browsing companies back in the '90s. So that's that connection. And then I mentioned it already. Third has to be Google because it's intermittent, but it's an extension of my brain and I'll be involved in conversations with people and they'll say, "Oh, I wonder whatever happened to X" and I won't remember, either, so part of the collective memory is captured somehow in Google searches. So those the three. I hope it's not boring.

 

Jean:

That is fun. Actually, you are my first guest whose top three apps are not communications per se. It's about information. This is very interesting. It's not about email and it's not about any of this. I absolutely love it.

 

Dan:

Oh, cool. Well, it may be discounting how much I use Zoom because we're spending a lot of time on Zoom these days, aren't we? But I wouldn't call it an app.

 

Jean:

There you go. Thank you again, Dan. That was an absolute pleasure. I thank you very much.

 

Dan:

Thank you.