Voice bots and IVR Systems –
The Future of On-Call Support

Jordi Torras:
Hello everybody. And welcome to the Future of Customer Service podcast, by Inbenta. We are really excited today to have with us Livio Pugliese. He’s the country manager of Interactive Media, an amazing company that specializes in making chatbots and artificial intelligence talk through phones and traditional lines. So, that’s a very interesting topic that I want to discuss. So Livio, how are you today?

Livio Pugliese:
Very good. Thank you Jordi.

Jordi Torras:
Absolutely.

Livio Pugliese:
Thanks for having me.

Jordi Torras:
Oh, absolutely. It’s an honor having you, such a knowledgeable person, who is the country manager for Interactive Media in the US. It’s a real pleasure having you here. Could you introduce yourself, what you and your company do?

Livio Pugliese:
Sure. So, I’m Livio Pugliese. I have had a long experience in telecommunications and customer service. I was with several companies in the industry. And right now I work for Interactive Media, which is based in Italy, with operations also here in the United States and in Brazil. And we’ve been in the telecom voice industry for a very long time, about 25 years at this point. And we have a lot of experience in this.

Jordi Torras:
Absolutely. So if you, listeners of this podcast, take some time to look at the Inbenta website, we have this integration hub and there’s a number of partners that we work with, that support our chatbot product to work over lines. And one of the things that you will see is this integration with a product called PhoneMyBot. So Livio, could you tell us what is PhoneMyBot, please?

Livio Pugliese:
Yeah. So Interactive Media, a certain point looked at the market and so that in conversational AI, there’s a lot of text-based bots out there. It’s been really an explosion in the past few years, but not so much for voice bots. So clearly they exist, but they’re much less common. And, in general, it takes a lot of effort to develop voice bots. And so we decided, it was a good niche… There’s definitely a need to expand what has been done in the chat sector and give it a voice. So really PhoneMyBot is a solution that gives chatbots the gift of a voice. In a sense.

Jordi Torras:
Could you tell us, in a nutshell, what are the main modules of a voice bot? How does that work? How does it interact with text bots, which are the main components of voice solutions such as yours?

Livio Pugliese:
Yeah. So, of course, you have the intelligence. The conversational AI is in the backend, in the chatbot, and the chatbot has a knowledge base developed and integrations with all the systems in the backend of the customer service system of the company. And they’re receiving text interactions in essence. Now, what does it mean to have voice, especially a telephone voice that interacts with the chatbot? Well, first we need to connect with the telephone network. And, obviously, we do that, we’ve been doing that for forever, and then you need to take the speech and convert it into text. And there are many commercial offers for that. So, there is a module that manages the speech-to-text engine, and we use more than one.

Livio Pugliese:
At that point, you have the words of the user, which have been translated to text. You send it to the chatbot. The chatbot does its processing, comes back with an answer or another question, which means more text, and then we need to convert that text into speech. Again, there are many commercial offers for that. There are different voices, different languages, and we usually manage that through our media server. Once the speech comes back, we pipe it into the telephone connection again.

Jordi Torras:
All right. That’s amazing. I remember we had in this very podcast, we invited once, Fabien Reinkemier from a company called Elaboratum. They help customers build and understand conversational AI. And one of the things that I remember is, well, if you want to build a good conversational assistant, you have to imagine that the conversation would happen over the phone. So when you are writing your answers, imagine whether it will work over the phone or not. So my question to you is how different is building a conversational AI, let’s say over a pure digital channel, like the web or WhatsApp from envisioning it for a voice conversation. Is it different at all?

Livio Pugliese:
Well, let’s say that over the phone, you have obviously a more stringent real-time requirement. So people, when they use a chat application, they are used to getting answers in a few seconds, maybe 20 seconds. Something like that. Well, obviously on the phone you need an instant answer, in like one to two seconds of maximum delay, right? The other thing is, well, there are different challenges. So obviously speech-to-text is not perfect, we are trying to do our best to optimize it, of course. For instance, using different special tech engines for different use cases, different languages, and so on and so forth. On one hand, something might come back wrong, or maybe with not enough confidence to send it over. On the other hand, you will never have problems like misspelling or things like that because the speech engine takes care of it.

Jordi Torras:
It generates perfect written language. And we have seen in the industry of call centers, the idea of IVR, right? Which means…

Livio Pugliese:
Interactive Voice Response.

Jordi Torras:
Exactly. Interactive Voice Response, which is a glorified way to say a voice menu. So press one, if you want to do A, press two if you want to do B. And I don’t know if that happens to other people listening here, but me, when I get to number nine, I don’t remember what was number one. So a natural question here is, what are the advantages of using voice bots or call bots, for calls versus IVR? What is the advantage for customers in this transition?

Livio Pugliese:
The advantages are enormous. As you’ve said, IVRs, or touchstone IVRs, which are still probably the most common ones are hard to navigate. Sometimes, people forget the options or they have to wait until the end of the menu to see if their option actually comes up and all these kinds of things. It takes a long time and a lot of effort. However, if you have a conversational IVR, the intelligence behind it understands the intent immediately. So instead of having to navigate menus, you can get pretty quickly to that particular intent. It has to be flexible, because sometimes people don’t express everything at the same time of course, but you can usually narrow it to the reason for the call, after a first question, or after they express what they’re calling for. Then, you can follow up with different questions and very quickly understand what they want. In essence, it flattens out the menu tree into a single line, if you want.

Jordi Torras:
I’m assuming that once you have this call bot in place, it’s not only about static answers, right? It’s not ‘you ask a question and you get what you have to do’. It’s not a static answer, because the system is using text-to-speech technology, so it can deliver personalized information. It can implement transactions in the backend, right? So my question is, how is that done? According to your experience, how is security involved in that? I am calling, talking to a machine and the machine is going to give information about me. How is the machine sure, that I’m me and not some other person, right? How would that work?

Livio Pugliese:
There are several ways. First of all, when you call in the system or it receives your telephone number, there is usually a first hint of who you are from the coding number. Then at that point, you can, for instance, ask for a PIN, if there is a PIN, or there are several ways now, like a biometric-

Jordi Torras:
PIN is the personal identification number, right?

Livio Pugliese:
That’s right, yes.

Jordi Torras:
It is the equivalent of the password in the world of phones and it has just a few numbers.

Livio Pugliese:
Absolutely. But we also find many instances in which we integrate with biometric authentication services. So depending on how the company system is organized, these are services with voice springs, for instance, in which the customer has previously recorded, his or her voice within the system, and then it’s recognized by the system.

Jordi Torras:
That’s super interesting. And you can integrate with that right? Essentially I call and the system recognizes my voice across so many, multiple, maybe hundreds or thousands of different voices that are prerecorded.

Livio Pugliese:
Absolutely, that works really well. One thing that I will say is that there are two ways of doing this. One is passive in which the system does record a snippet of the voice, and then authenticates it without telling them we are saying anything to the caller.

Jordi Torras:
Okay, we are secretly recognizing the voice, not telling, Hey, Livio, I can hear that it’s you.

Livio Pugliese:
That’s right, yeah. I think it would be pretty disconcerting and not a very good customer experience. So I really prefer if the system asked for a sample, say for instance, my voice is my password. And then you recognize that. So the user at least knows that he’s being authenticated and he’s not sort of suspicious.

Jordi Torras:
So, we are learning a lot today on all this world. Right? So an important component here is what we call speech-to-text, right? Or ASR it’s called, right?

Livio Pugliese:
ASR, Automatic Speech Recognition, yes.

Jordi Torras:
Oh, amazing. Amazing. Thank you, Livio. Then, essentially that’s transformed into text and there is the whole thing about sentiment detection, right? There are chatbots, some of them come equipped with that. For instance, Inbenta comes with a sentiment detection module as well, which detects what certain kinds of words and language imply (negative or positive). But of course, that’s what you say, how you say it it’s a whole other story, and that is lost in the text. And that’s why sometimes human communication is difficult. I send an email and the way the email is interpreted from sentiment is led to the reader. As a writer, I probably that’s why the emoticon has been so popular, because we put the little emoticon and say, well, I’m saying this, but, you could interpret that negatively, but no I’m saying it positively. But when we are speaking, there are no emoticons. We are the emoticons. So we smile. And the way we speak and the tone of the voice, is that also processed in the system? Are there ways to use that?

Livio Pugliese:
Well it can be, because as Interactive media, we don’t have our own ASR. We use ASR from commercial ASRs and, the technology is continuously changing, of course. So it’s been commoditized. There are more options. Some of them may have sentiment detection, so yes, absolutely.

Jordi Torras:
Okay. That’s fascinating. All right. So all that is about artificial intelligence, right? So what is your vision and how artificial intelligence is going to bring to different approaches to customer interaction in the future?

Livio Pugliese:
Well, so right now, artificial intelligence in many ways is changing customer service operations. From chatbots and voice bots in front of human agents to, for instance, automatic processing, automatic process management. That means agents don’t need to put in a lot of effort in post-call analysis, post-call report, and these kinds of things. So what we see at the moment is that the role of AI is really to be able to solve a good percentage of the most common support cases. So something like, ‘I want to reset my password’, for instance, that used to need the help of a human agent, at this point it is not necessary anymore because the AI is able to understand what the problem is and do it automatically, right?

Livio Pugliese:
I mean, this is huge. On one hand, customers are more and more using self-service. Most of the time, they are happier to do that because the call times are shorter. There is no wait because obviously, the AI scales seamlessly and, only when something is complicated, they need to escalate the case to a human agent. So this is much more efficient, better for the customer, and also better for the company. And in the end, also better for the agents who, at this point, don’t have to continue to do routine tasks.

Jordi Torras:
You know, and that’s what we have been advocating for from Inbenta: is AI is not here to kind of remove jobs. What we do is we make these jobs more interesting.

Livio Pugliese:
Absolutely.

Jordi Torras:
Taking care of the most repetitive and boring activities and freeing time and energy to build better content, to build a strategy, to improve quality for services. So we are absolute believers that AI, well, first of all, is here to stay, and B, is not here to remove jobs but the other way around. It’s here to actually create new roles, new skills that are going to be needed in the future. Pretty sure about that.

Livio Pugliese:
Absolutely yeah, yes.

Jordi Torras:
So all right. We’ve been working together for a while, now, right? Would you tell us your experience working with Inbenta, this company that I happen to be the founder and CEO of.

Livio Pugliese:
Absolutely. No, it’s been great. I find that you have a very agile organization and technically you’re very strong, because you’ve been in this for a very long time. Just as we have been, right? So we definitely understand each other, but at the same time, you’re really receptive to new ideas because this is something new for you.

Jordi Torras:
Absolutely, absolutely. We’ve been in the digital space forever. We’ve come a long way from, natural language processing, search engines, conversational AI, all that. And then when you put voice technology and phone services, you feel that you are completing the circle, right? It’s like ‘Now, you and I are communicating by speaking’. And that’s the most efficient way humans communicate with each other. And probably it’s been like that for millions of years. So we are not going to change that over time. So it’s amazing to see how these apparently different technologies are starting to converge.

Livio Pugliese:
I mean, considering the speech, again, in customer service, I’ve seen research from my previous firm, confirming that obviously digital has been very much growing. It’s used by a lot of people, but paradoxically, speech and phone calls have not gone down any portion. They’ve gone up, because what happens when you can’t solve things with the digital, then you need to call and you need to have real-time feedback and messages.

Jordi Torras:
Yeah. Well, and that opens a real interesting question, right? So I call, I get a bot answering my questions. What if my question is not answered though? What if the bot cannot figure out the way to solve it? What happens then? What can the bot do?

Livio Pugliese:
Yeah. So we integrate with the contact center as well. And so the bot can, at any time transfer the call to a human agent with the advantage that the whole conversation possibly, or at least some of the data within the conversation, can be transferred to the agents’ desktop. And so they will see-

Jordi Torras:
So, they don’t have to be annoying like, okay, ‘What is your name?’ and start from the beginning. The agent can have a great deal of information on what happened, what was the question, and the answers given by the bot. Therefore, this speeds up the process of having the problem solved and simplify both the life of the agent and the life of the customers. Listen, I could be talking about this topic for a long, long time, but we have some time limit here in our podcast. So Livio, we’d like you to let us know how our audience can contact you or your company. What is the best way to find you?

Livio Pugliese:
Yeah. So I think that the best way would really be, go to our PhoneMyBot website.

Jordi Torras:
That’s PhoneMyBot.com, okay.

Livio Pugliese:
And then look around and there is a contact us button of course.

Jordi Torras:
All right. That message is going to go to you guys. Dear audience, you know PhoneMyBot.com. This is where you will find Livio and the team doing a great job regarding bringing intelligence or artificial intelligence to phone conversations. Livio, thank you so much for your time.

Livio Pugliese:
Thank you. Great talking to you.

Jordi Torras:
Absolutely. All right. Have a good day. Take care. Bye-bye.

Livio Pugliese:
Bye.

Thanks so much for tuning in. This podcast was brought to you by Inbenta. Inbenta symbolic AI implements natural language processing that requires no training data with Inbenta’s extensive lexicon and patented algorithms. Check out this robust customer interaction platform for your AI needs, from chatbots to search to knowledge centers and messenger platforms. Just go to our website to request a demo at inbenta.com. That’s I-N-B-E-N-T-A.com and if you liked what you heard today, please be sure to subscribe to this podcast and leave us a review. Thank you.