Welcome to one of our first episodes of the thought leadership podcast from our team here at Interactions, where we share conversations between our CMO, Peter Mullen, and some of the incredible people who are key to shaping the future of Conversational AI.
Today, Peter is joined by Dr. Lisa Michaud, Sr. Product Manager at Interactions, who is an incredible leader in this space and was named to the Class of 2023 “Women Leaders of Conversational AI.”
In this episode, you’ll hear about Lisa’s unique story that led her into this field, her insight into why conversational ability is so crucial to the effectiveness of AI technology, and practical ways of how AI technology can support your company’s goals. You'll also get an inside look at our processes at Interactions when we start working with new clients, providing you with a deeper understanding of our approach.
Lisa made a great point by diving into the reality that there are many different approaches to Conversational AI today. The most important question at the end of the day is simple: Does it work?
What are the tasks customers want to achieve? How can we create an experience that supports those needs? Beyond that, by implementing Conversational AI, companies can get rid of the friction that is created when a customer needs to call within a certain timeframe, or experiences a long wait just to talk to a customer service representative.
This is something we are pursuing week after week here at Interactions, and is truly a fascinating part of our work that Lisa discussed. Effective AI technology needs to be able to respond to fragments of thoughts, references made to something said earlier, etc. Lisa gave some great examples during the episode.
Hello everyone! I’m Peter Mullen, CMO of Interactions and I’m here today with another episode of our Thought Leadership podcast with Dr. Lisa Michaud. Hey, so welcome to the podcast.
Dr. Lisa Michaud
I am thrilled that we’re getting a chance to talk, and I get to learn about all the cool stuff that you do and all the ways that you are helping CX not only with Interactions but in the overall ecosystem.
Let’s dive right in. I want you to tell me a little bit about your story. Who are you? What’s your background? And then we’re going to get into what you do at Interactions.
Dr. Lisa Michaud
Well, you know, to start with my background it’s sort of an interesting journey in that I started out actually with the intention of becoming a professional writer. My two Graduate Degrees are in English literature, but also computer science. And not a lot of people thought those two things were connected until I found this field called Computational Linguistics in grad school, otherwise known as Natural Language Processing.
So I did graduate school in that language processing, and I actually taught computer science for my first career before coming into the contact center industry. And I taught that and did research in an LP for 13 years. And then I started the transition from academia to industry. And in the last seven years, I’ve been working in the space of product centers and in conversational AI in the contact center space.
So continuing that sort of blend of my interest in computers, the computational approaches to things and my fascination with language and it’s interesting how those two things which seem so separate, have basically joined together and form the thread through my career. And it’s nice to look at two things that I like at the same time.
Yeah, it’s fascinating to me and I’m guessing that you don’t bump into many people like you who have a hard science, who have a soft science, who have intertwined together to create something that combines the best of both, but also has a net new external or exterior purpose in the world.
Dr. Lisa Michaud
Yeah, I mean, I think that there are more of us out there than people realize. But in certain circumstances, in certain circles, people have seen it more often than others. I was sort of a little microcosm when I was an undergrad and doing two different majors because that was really, really common for people there to do something that was humanities and lab science or something in the social sciences and physics or whatever.
And it was only more after that that I started encountering people who like either one or the other and thought, why would somebody who likes this also like this other thing? But the great thing about the field of natural language processing is that it always brought together a very interesting mix of people. And in particular, when I first got involved in the field at a conference, what you would meet were computer scientists who loved language and you met linguists who loved to sing a little bit of programing code.
And so you found a whole community of people who liked to bridge those two interests.
It’s, I think, 2022 and beyond this concept of bridging is one of the most essential things that we can be doing in anything. I had a mentor years ago who would not hire you as a computer scientist unless you played a musical instrument, because it was this fusion between the two sides of the brain.
Tell me what you now do when you bring these two sides of your brain into Interactions.
Dr. Lisa Michaud
At Interactions I’m a product manager. And I have a lot of different hats that I wear as a product manager. My primary focus is in how we create an instance of our intelligent virtual assistance. So my primary focus is the environment that we use to create that experience and author that dialog and the underlying components that actually run that once that’s been created for a client.
But unofficially, I also end up poking myself, mark my nose into just about every portion of our platform, because all of it is really part of our product. I like to say that anything that any piece of code that we execute in between a sentence coming in and a sentence going out is part of our product. So I end up learning a lot and interacting a lot with a lot of the engineering staff and all that work on all the different components along the way.
But officially I focus on that authoring experience and what that means to create a virtual system for one of our clients.
And I think it’s worth pausing for a minute and me asking the very simple question about Interactions. What does Interactions do that makes us different from so many other conversational AI players in the marketplace? And how does that specifically fit into our value proposition?
Dr. Lisa Michaud
Well, I mean, I think the simplest answer to that question about what we do that makes us different is, like you said, there’s so many different approaches to conversational AI. And the bottom line of it as to what is important is does it work? Does it get the job done for the person who’s talking to this particular instance of conversational AI?
Because we’re not talking about chatbots. I’m not talking about having a pleasant conversation passing the time. We’re talking about technology that’s being specifically deployed in a situation where someone has a task they want to get done. They have something they want sold, an answer they want. They want to accomplish some task. And if the AI can understand them and get that done, then that is critically important.
And in terms of where we sit in that we have an AI capability that goes so far beyond what so many other systems do in terms of we so rarely have to answer with “I’m sorry, I did not understand. Could you please repeat that?” and that you know that because of that realization of what AI can do and the fact that we blend the intelligence of humans with the intelligence of the AI in order to close the gaps, to make certain that the AI successful, that sets us apart.
And that’s what really attracted me when I first started working with Interactions a few years ago.
It feels very much oftentimes like you’re solving one of the most important communication puzzles that are out there. So a new company comes in and starts working with Interactions. You’re greeted with a brand new puzzle, a brand new project, right? How does that start on day one and what does that begin to look like after day 60 as you’re looking at a new project and building it out from the seed of where you were?
Dr. Lisa Michaud
I would say the expert on the answer to that question would be someone who works on one of our project teams, but at the same time, it’s something that I’ve had to be very close to because again, the primary product that is my central responsibility is the environment where we craft that and that has to be the place where they go to start that journey.
So at the very beginning, it’s very much about discovery. It’s about what is the problem that they need to have solved. And some of the questions that we’re trying to answer at that time are, for example, what is it that people are talking to them about when they’re reaching out in that customer service space?
What are those tasks that the customers want to achieve? And that’s just by itself a difficult thing to discover because a lot of contact centers don’t have a lot of visibility into the relative importance of some of those different things because there’s very different difficulties around reporting and around getting the disposition code set right to give them that kind of visibility.
So the first thing we need to do is help them discover what are the most common things that people are reaching out to the contact center about, and then which of these are the ones that AI can be empowered to solve? So that’s both understanding the request and having all the tools at its hands to get that done.
And if you think like fast forward what that looks like a few months down the line, once we figure that out and we decide, okay, these are the things we’re going to first build the system and then it’s about going through and making certain that we create an experience that allows somebody to get that done effortlessly.
And, you know, I love that you often introduce yourself as a communicator or even a great communicator, because I don’t think most people think about this type of engagement, calling in for customer service as a communication. But it really is, right? You’re having a dialog back and forth with people. That’s how you frame it, right?
Dr. Lisa Michaud
Well, absolutely. I mean, it’s … I think in many ways I make the joke that one of the things that I’ve learned most about this field of NLP and specifically about dialog is it’s a very difficult task and that’s great because then I have job security. But the reality is if you go all the way back to the beginning of artificial intelligence, if you go back 70 years and you look at the writings of Alan Turing, he literally set the bar as that.
If you want to define when have we achieved artificial intelligence, it’s as when a machine can have a conversation. Right. That’s what he defined as that’s the metric for how we have achieved AI, is when a conversation can happen. And there’s a lot of reasons why he set that bar. But it’s something that we keep coming back to. It’s a difficult task.
And there’s shortcuts you can take and there’s hacks you can do. But what we really want to achieve is that it’s as easy as talking to a human being. We don’t necessarily need for our conversational AI to become sentient, but we need for that conversation to be as fluent and effortless as if they were talking to human beings because they’re being understood, because the conversation turns feel natural, and they don’t have to have that, you know, that frustration of iterating over what they said, fixing the mistakes and whatnot.
I have a driving thesis that over the next 15 years, customer satisfaction will be the number one predictor of repeatable revenue for companies. And buried inside that grand statement is this concept that today 80% of the people will express a desire to leave a brand with one customer support experience. And multiple studies have repeated these types of staggering numbers.
As we move forward over the next 15 years, companies need to recognize that their customers can go somewhere else. And so we have got to remove all these friction points that you’re talking about, that AI is delivering a better, seamless, transparent processing that gets people to their resolution faster. And I think that this will become, in the years ahead, the strongest push in corporate levels for us to figure out how to make customer satisfaction even better than it is today.
And I think that you’re at the vanguard of where we’re going with all of this. You and people are recognizing the work that you have been doing. So I know you just were recognized as a class of 2023 member of the Women’s Leaders of Conversational AI. Congratulations on that. Tell me a little bit about what that means.
And I’d also like you to talk to me a little bit about women leaders in this particular field because I think that’s an important element of this.
Dr. Lisa Michaud
I mean, I think it’s interesting because there are certain levels to it because of course, the fields of computer science and artificial intelligence have sort of ebbed and flowed in terms of representation of women in general. Conversational technologies have been relatively, evenly, traditionally, somewhat evenly represented in terms of men and women, unlike other fields of computer science.
But that’s actually gone down in recent years rather than going up. So I think it’s more than ever important to make certain that we do take a moment to recognize folks and the list of people that were part of this, women and leaders Group for 2023 involves some absolutely fantastic people, people who have been involved in this technology since the earliest speech technologies that were being first brought on board and are as simple compared to what we have today as, you know, doing computation as an abacus or something to women who like a woman who founded a women in voice community, which started off as a sort of an informal forum and is now this
massive flourishing community across voice technologies. So it’s really fantastic to see my name in that group. And I’m really hoping that a lot of them will come to the summit associated with it. So I get the chance to hang out with them. They’re all really fantastic people. So that’s very exciting. And it’s interesting and it’s so important to form those communities because the reality is that, you know, like I said, the participation of women in technological fields has actually been going down rather than up.
And so more than ever, you need to form mentorship, relationships and networking and whatnot to make certain that particularly the people who are just entering, don’t feel alone, you know, that they feel like they’re supported and they feel like they’re going to be, you know, part of a community and not just the only woman on the call, for example.
You know, and I think even in this post-COVID world, this is more important than ever. We’re now majority remote. The work that we do can often be a slog and this area of computational linguistics that you dive into, it can be lonely at times as well. I can’t agree with you more. We have got to make sure that we’re lifting up people all around us, regardless of what company they work for, regardless of what’s due to the career they’re in. There is just so much happening at such an incredible velocity that collectively we can continue to push the entire field forward.
But we just got to watch each other’s back and support each other going forward. I want to touch again on something that you touched on a moment ago, and it was about the direction of conversational AI. When you think about the ambitions of conversational AI and what it is capable of achieving. I want to go about sentients versus not. What do you see happening from a category standpoint over the next five years in this area?
Dr. Lisa Michaud
Well, I mean, I think that one of the interesting things, one of the interesting viewpoints that I have, given my background, is I can see both what we have out there in enterprise, in enterprise technology, where we’re actually using it and relying on it versus the research thoughts which are in some respects science experiments, right? They are the research work, which is about publishing a paper and not necessarily something that we rely on.
And there can actually be a rather large gap in some areas. For example, there are these systems that I see sometimes being presented as research papers where when they’re good, they’re actually fairly good. But when they go off the rails, they go utterly off the rails. And so they’re not, you know, that technology is not quite there.
The things that I’m most interested in seeing coming out of the experimental phase it into sort of enterprise reliability and something that we can see in product often revolve around things like evolving beyond seeing every sentence in its isolation. So a lot of systems that are out there today will work on understanding the sentence you made just now, but just that sentence, not in the context of the entire conversation, which human beings don’t do, right.
We always understand what is being said right now in the context of what we’ve been talking about for the past few minutes. And while you can talk to your virtual assistant in your home and say something like, what’s the weather today? Well, how about tomorrow? And you can say, hey, look, it understood that sense at that second query, even though that second query didn’t have all the information and it said, Oh, she must be repeating the previous query with this little change actually that it’s preprogrammed is not right now a very intelligent interpretation of a question, which is knock out all the data.
They just had to anticipate that people may ask a follow up question of this kind and have that ready to go. And what we really need to do is to continue the work towards understanding when someone is asking a follow up question that is not fully formed, that makes a reference to what we just talked about in that context.
Without having to anticipate all the possibilities ahead of time. And this is a question that has been worked on. I mean, my PhD advisor was working on this in her PhD in the eighties. So this is a question and a problem that we’ve been working on for a very long time. But I think this is where we are starting to build that structure and build that capability.
And that’s going to be something that we’re seeing coming up. And from a personal perspective, one of the things I would love to get more into is looking at how conversational AI can be particularly robust to the fact that not everybody speaks with the same accent, that everybody is a native speaker. And, you know, if we’re talking about effortless, if we’re talking about somebody being able to talk as naturally with the AI as they would with another person, a large number of people will flip languages as they talk as a natural part of how they express themselves.
There’s not a lot going on right now to be completely robust to that, to that code switching. People mushing two languages together in the same utterance. I think we have all the tools ready for that. I think we’re on the cusp of being able to do those kinds of technologies. It’s not that far out. And as I did a lot of work and non-native users of a language in my own Ph.D. research, it would be like coming back to a familiar territory for me after 20 years to see what we could do with that.
With the technology we have today as opposed to 20 years ago.
Code switching, it is fascinating to start pulling back the layers on this. When we were in conversational AI, are there different accents that are harder or easier for the machine to understand?
Dr. Lisa Michaud
It all depends on whether it was trained on. So, you know, one of the threads that’s going on in AI in general is there’s a lot of conversation about how we need to have ethics in AI research that, among other things, becomes critically aware of the fact that over and over again we have AI projects that have shown that the lack of diversity in the input leads to output that only works for certain people.
So everything from people training facial recognition software on faces that are all white and what that means for how that software deals with, you know, when it has a more diverse group of faces to categorize, to dealing with accents. A lot of my friends who are not native speakers of American English get incredibly frustrated when they’re talking to the interfaces on cars because they said the car never understands me.
A lot of interfaces are primarily trained on men’s voices, so they don’t deal as well with women’s voices so recognizing that you need to have diverse training data, a diverse input is critical to making a system which is robust to that. Now we have a lot of ways of dealing with this, you know, at Interactions.
And one of those ways is that we do a lot of training on the population of speakers that actually use the system. So that system evolves to be really good at understanding that population, but you’re still going to have outliers for that population and that’s going to pose a challenge. But to be aware of the fact that you want to take care of those outliers, you don’t want to have a system that just can do people who have a certain kind of voice, a certain kind of accent and a certain kind of linguistic structure is, you know, is absolutely critical that, again, we make certain that that as an equitable access and it has that effortless interaction so that people don’t have to feel like this is something that they are, you know, they’re struggling to communicate, basically.
I think it’s a huge issue from both an ethics/morality standpoint, but also from a business standpoint. We have the opportunity to de-marginalize everyone and we’ve got to put in the systems and build the systems to ensure that there is equitable access and equitable engagement across the entire spectrum. I know that sounds very obvious what I just said aloud, but that takes a lot of effort and a lot of work and I actually believe it’s one of the obligations of conversational AI and AI itself to be thoughtful of that at all times.
And I know that at Interactions we spend a lot of time thinking about this level of nuance. And even using the word nuance is not correct, it’s more essential to this. As far as I’m concerned, it’s the foundation of how we should be building everything.
Dr. Lisa Michaud
And I think that one of the other things that we need to keep in mind is not just like we’ve been talking about accents, so that talks about the voice channel, but if we’re talking about equitable access, we have to talk about the idea of the fact that the Interactions is an opti-channel IVA, it’s not just the voice channel, and by opening up those other channels, what you also do is you make certain that you are providing equitable access for people who don’t use voice, who don’t you know, who maybe can’t use voice, who are going to go to those text channels.
And if you look at your experiences as a customer today, when you’re reaching out to many companies, a lot of brands have very siloed approaches to the different channels, and those siloed approaches mean that they have one person who’s providing a chatbot that’s sitting on their web chat, and there’s one person who’s doing their phone.
You know, their phone, IVR, or whatever, and another provider for Facebook messenger. And those are very unequal experiences. And I am somebody who actually almost always reaches out in text first. And I will find over and over again that the options available in the text channels do not allow me to get the work done that I want to get done.
And they will keep saying, I’m sorry to do that. You have to go to the voice, you have to call. And I didn’t want to call, but you know, and I have the option at least of calling. Right. But a lot of people don’t have that, you know, that equality of access across the different channels.
And when you achieve that and when you bring all those together under one umbrella, you also make it a universal access so people can reach out in the channels that work for them.
Yeah, one of the lines of question that I think about is how AI can ultimately empower the customers and this is exactly what you’re touching on right now and how we need to overcome that. Could you talk a little bit more, perhaps not about the challenges that we face, but also what empowerment really begins to look like when AI is working the right way?
But what’s a practical benefit for the millions of users who, this hour, are engaging with a conversational AI system?
Dr. Lisa Michaud
Beyond what I’ve already talked about, I think it’s critically important to recognize that it comes back to my first point of you succeed when you get the job done and when you get that job done without pain. So what constitutes pain, right? There’s the waiting. Right. So if you have only humans that are staffing your, you know, your contact center and people are trying to get something done and they have to wait in a queue for a long time to get a human being, that is not a positive experience.
It is a frustrating and painful experience. When you push automation into that space, what you do is you alleviate that and say instead of waiting, you get this done right away through this, you know this other venue. And also if you do not have your contact center staffed all the time. I’m always fighting in frustration when someone says, Well, you have to contact us within working hours.
Well I’m not working through working hours. I’m often in a meeting from 8:00 until 5:00 without a break. When am I supposed to contact them if I’m working at the same time they’re working? If there were automation there that allowing me to get things done in the schedule when I want. So it’s like, you know, in the time period that you want at the time zone that you want basically and in the channel that you’re looking for.
So that, you know, either that meets your preference or your needs. And that we saw in particular some great data that we have in one of our recent surveys around how people’s desire for AI automation greatly increased after the onset of the pandemic as the contact centers had even more staffing problems and the wait times got even worse.
You know, people started saying, you just talked about brand loyalty and how important CX is. People said that the availability of technology that they can reach out to and get their tasks done when they need it at that time, without waiting for 2 hours to get a human being, was critical to their decision to stay loyal to a brand.
So, you know, all of, you’re empowering your customer to get something done without it being a chore, without it being something that they dread. Nobody wants to call customer service ever, right? They want to get something done. They want that problem resolved, but they don’t want to spend time on it and they don’t want to get frustrated over trying to get themselves understood.
And 99.9% know exactly what they want to accomplish as they engage, and we’ve got to get that to them. One of my other theses and this is not a eureka moment, is that COVID accelerated by a decade or more, everything that was happening in customer experience. It just pushed forward what we need to do and what you just described is exactly like that.
Then you combine it in the macro environment with labor shortages, then you combine it with now recessionary possibilities inside the economy and now you’re having a headwind for having an easy customer experience. Yeah, you’re up until 9 to 5. They’re also not open on Saturday and Sunday, which is when traditionally I like to call and solve. So conversational AI and just a smooth, effortless process is all part of this automatic solution that I think we are all fast becoming accustomed to and are growing to expect. And then now I’m into a flywheel and a virtual cycle of how this all wraps around for a great customer experience to give us this customer satisfaction to keep revenue predictable and grow.
Dr. Lisa Michaud
Right. So then the contact center doesn’t become a cost center. It’s not just something that is an asset and necessity that takes money, but it’s also part and parcel of how you take care of your customer and ensure that loyalty and you become part of that customer journey. So yeah, it’s an evolution. And I think like you said, I think that the pandemic pushed that further.
It was already tilting that direction, but it’s accelerating that growth in that direction.
Lisa, thank you for joining us today. And until next time, you’ve given us a lot to think about. Have a great afternoon.
Dr. Lisa Michaud
Okay. Thank you. You too.