Episode 3

The Human Touch: How Human-Assisted Understanding Changes the Game

Welcome back as we share conversations with our CMO, Peter Mullen, and some of the incredible people that are shaping the future of conversational AI.

In this episode, Peter is joined by Mike Pell, the Director of CX Design at Interactions. Mike probably doesn’t do what you think he does, but you’ll have to listen in to find out what we mean. He has worked in and around call centers since 1991, and is particularly passionate about the industry because of how challenging and ever-evolving it is.

Today, you’ll hear Peter and Mike discuss how his role impacts millions of people each year, why humans are ultimately the best at interpreting language (and the secret sauce to a successful experience), how to craft effective customer experiences, and much more.

Short on time? Here are some quick takeaways:

  1. Designing customer experiences is all about finding solutions that require the least amount of effort for customers.

    At the heart of what we do is working to find the most painless path to lead customers along that solves problems and drives value. When you do that, you end up saving money and finding real solutions. Plus, your customers walk away happy!

  2. Human-Assisted Understanding (HAU) makes all the difference.

    Have you ever heard “I didn’t get that” during a conversation with an AI Virtual Assistant? It can be frustrating, right? This is a situation where Human-Assisted Understanding (HAU) can make all the difference. This is something that gives Interactions a leg up on the competition. If the AI doesn’t comprehend what is being said, it gets played for a human in real time to provide a frustration-free experience for your customers.

  3. We always need to keep improving.

    This goes to everything from grammar to your specific industry’s lingo to how language evolves. Things are always changing, which requires effective AI systems to constantly evolve and update as well.

Michael Pell

Michael Pell

Michael Pell has worked in Call Centers since 1991. He’s answered phones, trained agents, written scripts, recorded vocal performances, supervised designers, and managed IVR development teams for all kinds of businesses and industries. For the past decade, Mike has served in a variety of roles at Interactions, designing highly conversational speech solutions for enterprise customers.
Read the transcript

Peter Mullen
Hello. My name is Peter Mullen, CMO of Interactions. And today we’re continuing our new Interactions video series where I get to interview the incredible people that make Interactions the most successful conversational AI platform and solution in the entire world. I am thrilled today to be joined by Mike Powell, Our Director of CX Design. CX Design is focused on differentiating our company and giving incredible experiences to millions and millions of users that are engaged with the Interactions platform.

Mike probably doesn’t do what you think he does, and we’re going to dive right in and we’re going to learn how Mike positively impacts these millions, hundreds of millions of people each year. So let’s get right into it. Mike, welcome to the podcast.

Mike Pell
Hey, P.H. Thanks.

Peter Mullen
Let’s start with a really simple set of definitions. CX Design at Interactions, which you oversee. What is it in general? What is it at Interactions?

Mike Pell
So CX Design is essentially our design team. These are the 18 professionals, most of whom have been in the industry for 15 plus years. We’ve got a very seasoned bench of great smart designers who interface with our customers. They work with the customers. They try to find out exactly what the pain points are for the customers and for their customers, more importantly, really, and work through solutions.

That’s really what design is all about, is trying to find the most painless, least effort on the caller’s part or the chatter’s path through the various business rules to try to solve problems. And I mean, it seems silly, everybody says it, but to drive value. To make it so that our clients are seeing both a financial value and then a customer service value.

I mean, one of the hallmarks I think of an Interactions design system is that it is conversational and it is easy and people who go through those systems, our systems, typically are very happy when they come out the other end. And therefore we’ve had a lot of surveys, set of benchmark before we install one of our systems and then after and we see customer service go up along with, the other things that our clients want to see go up like, self-service levels and better routing and all the things that you can offer with an intelligent virtual assistant.

Peter Mullen
There’s so many places to dive in on this. And so I’m going to start at the 30,000 foot level and then get right into it because there’s a ton that we get to unpack over the next handful of minutes, Mike. So Interactions provides conversational AI solutions to the world’s best brands. We know this. What that means is that in a customer support engagement, we provide better experiences for the millions of customers, in the hundreds of millions of customers, using our technology, where we talk about it as being the best front door experience in the world for customer support because of our CX design.

And that provides greater efficiency. It provides a faster speed for the customer to get satisfaction. It also provides tons of efficiency in the back end for the business that is using Interactions. And there’s a bunch of channels that we’re going to talk about and unpack, as I said, on this. But let’s talk about the customer first, and I’m going to play out a scenario for you, and I want you to walk me through it as if I’m a new customer of yours.

So in this scenario, I’ve just hired you, a Fortune 50 big box retailer. I have millions of customers that come to my big box and shop online every year. I’ve just engaged you in Interactions to improve my customer support hotlines. We’re on day one. We’re meeting. And what does that start to look like from your end? To build a better solution for me?

Mike Pell
Yeah, absolutely. So I mean, I think the best case scenario and where we’ve had the most success is when on that day one, you know, we literally lock ourselves in a room with the client and start just talking. Right? And from that conversation or from those many conversations over the course of one or two days is going to come sort of a list of things that we know we can do and the things that we want to know.

What do they perceive as their pain points? That’s where we start. Where did the client see room for improvement? And then we’re going to go and talk to the agents who are actually taking the calls on the phone. And because of our approach and our human understanding that we employ as part of our solutions, we’re able to do that.

Peter Mullen
You just said something really important. What is human understanding? Let’s define that for folks.

Mike Pell
Yeah, absolutely. So at Interactions, we employ ASR and all the same tricks of the trade that all of our competitors use. However, we do have one more sort of ace up our sleeve that no one else has, and that’s HAU or human assistant understanding. So human assist understanding is essentially the ability to insert human intelligence into any conversation.

And so, I mean, everyone’s had the experience of calling some system, some IVR and having it say, I didn’t get that or can you repeat that or I think you said and then repeat back something that may or may not be what you said. And then the caller has to then autocorrect or say yes or say no and, you know, whatever in the conversation.

Well at Interactions we skipped all those steps. We don’t need to do that because if the ASR and NLU combination is not confident that what it heard is correct, we literally play that to a human in real time. So there is a human being who’s not listening to the conversation, but who suddenly hears, Oh, this caller said, I want to pay my entire balance with a Visa card.

So then they can let the system know, Oh, this is what was said. And then the system will follow all the business rules and do what it would have been. Just as if the ASR and NLU had understood these things. So the idea basically being that humans understand language, that is what we do, That is our secret trick.

That’s how we’ve conquered the world, right? By being able to speak and communicate. And we do it better than machines. I started in this business about 20 years ago and 20 years ago, they were like, Well, in five years machines are going to catch up. And it’s been five years all the way to today.

And machines haven’t caught up. Humans understand English and any language that they speak natively better than any machine has been able to duplicate. So we use those humans to make our systems better.

Peter Mullen
And I think what’s also very distinct about this and this is a really excellent point of clarity, is that during the course of a conversation, there is a system that is deciding whether it’s going to a robot or it’s going to a human. It’s not like a human is listening to the entire conversation, right?

Mike Pell
Not at all, No. We have basically a recognizer that says, is this thing that a person said, which we refer to as an utterance, you know, is this utterance something that can be handled by ASR and NLU, and if so, we try that. And the vast majority of the time, quite frankly, it can be.

But then roughly 20% of the time, depending on the type of utterance, it might be up or less. Right. It decides, you know what? This is too complex and it goes to a human and the humans make the determination of what was spoken. And then again, the business engine kicks in and we do what we should based on what was said.

Peter Mullen
Let’s get back to the CX design. But I will point out that independent studies have shown that our solution, compared to others, tends to have an average of about nine utterances for someone to complete a call. And that’s an enormous number, meaning there is efficiency, there is satisfaction, and we are resolving the customer situation quickly that provides happiness to her and business ROI to the client that we’re working with.

Let’s get back to the scenario. There’s a big box retailer that has millions of calls coming in and they now have decided that we are the right solution for them because of what we just talked about with HAU and all of this really dynamic processing. You were describing how you will go in for several days and roll up your sleeves, start whiteboarding with all of the different stakeholders in the ecosystem. Move forward from that. Take me where you now go with this.

Mike Pell
Yeah, absolutely. So just what you said, we sit down and we work out the problems that need to be solved, then honestly, quite frankly, we look for more problems or more opportunity I think is a better way to look at it. So we’re going to talk to the agents, talk to the trainers, talk to the managers, the leads, you know, the people who are in the call center who really, really, really know this is what is driving our volume.

This is what is causing us to have to do work. And we try to figure out, well, what can we automate? A lot of times the key to that is data. So part of that day or two day adventure that we have with our new clients is going to be spent with the people who own the data, right?

It’s going to be spent with the IT folks who have built the APIs, who have built the soap interface, who have done, who know, sort of what data we have available. And then we’re going to try to figure out how we can leverage this data to the biggest impact for their client base, for their customer to, again, drive that value and allow callers to quickly complete the tasks that they called to complete.

Because when a person calls, they’re really always calling for a specific purpose. And if an automated system is both smooth and excellent at completing that task, the vast majority of people are very happy to work with the system. It’s only when the system fails or doesn’t understand or is incompetent that people get frustrated and then demand to speak to a human.

So we really try to just make sure that the experience is good enough that people will choose to automate.

Peter Mullen
And what you just said is so basic, obvious and natural. I find myself nodding my head, but we really stand out. Your team is our secret weapon. We do it differently. We do it with more depth and perhaps with more attention than many of our other competitors in the market. Is a fair statement?

Mike Pell
I think it is. I mean, there’s a lot of excellent designers across many, many companies. I think we’ve got a great deep bench and a lot of talented people, but we also do and we always have. I mean, when this company was founded, it was founded with the ideal of conversational approach. And if you listen to an Interactions system as opposed to most any other system out there, we try very hard to be conversational, to the nth degree where everything we use is a more casual language.

I mean, one of our mantras, try to speak like you would to a trusted friend or a relative and not being formal. And of course, sometimes there’s business needs that dictate a certain language you have to use. But whenever possible, if you are conversational, people will speak to you. And one of the, again, the hallmarks of our systems are because we can understand complex utterances and complex detailed speech from people.

It gives us a big leg up in terms of the ability to start and stay conversational throughout a transaction or throughout a call.

Peter Mullen
We have, as we know roughly with the industry benchmarks, that AI is so remarkable that it’s able to manage about 80% of these utterances today. But for most companies it stops there. And everything you’re talking about right now is the fact that we can add an extra nearly 20% to understanding. Right? Our numbers typically show we have about a 97% understanding rate, but we’re not managed services.

I want to emphasize this, that we have a constant interplay between the digital and between the human. As you were talking about a second ago. When people hear that, though, they start to scratch their heads and say, gosh, this must be a whole army of people, human beings behind the curtain that are taking care of all this. But talk to me a little bit more about that challenge question because it’s not like that, is it?

Mike Pell
Yeah, it’s not really. Again, as you correctly point out, the AI handles about 80% of what’s done and in reality the humans only come in when they’re needed. And if you think about the length of a call, you said that there are roughly nine utterances per call and in a typical Interactions call. I think of those, if you just kind of apply the percentages, humans are going to come in for one of those.

And a typical human utterance is between one and a half and 4 seconds. So you think about that. The call itself may last 3 minutes, right? A three minute phone call and you completed accepting a payment for some overdue bill. Right. So there was a balance thing, there was a discussion about a late fee maybe, and then we processed the payment and that took about 3 minutes.

However, the human aspect of that was 4 seconds long, so a human had to engage for 4 seconds. So when you think about that across your entire call base, if you’re thinking about the math of how the humans, we call them IAs, interactive assistants, what percentage of their time as compared to one of your phone agents. A phone agent is engaged with one caller for the full three, three and a half minutes of that phone call.

Whereas one of our IAs is engaged for that call for 4 seconds, and yet the caller has the same task completed. And, obviously there’s a lot of money to be saved in that differential between 4 seconds and three and a half minutes.

Peter Mullen
Absolutely. And over that, what you’re describing is in, let’s say, the first quarter of the first half year of a relationship. Over time, pretty rapidly, our intelligence gets even smarter. Because we’re capturing all the instances that move back and forth between human and between the digital side of things. And we can continually refine the machine for a better experience, right?

Mike Pell
Yeah. I mean, now you’re definitely talking about our secret sauce. So when we engage humans, there is a slight pause, right? Because the caller has already started speaking and we do stream in real time so the humans can pick that up. But there is a slight pause in there. So we have to fill that with during the phone call to the caller, they hear a little bit of filler or whatever, and if it’s a chat application, obviously there’s nothing missed because that sort of chat is not the same expectation of immediate response that speech does.

But we do have to be concerned with response. Well, one of the great things about ASR is it responds immediately. And one of the great things about Interactions is we use those human decisions. So the ASR engine throws up his hands at an utterance and says this is too complex for me. And we kick it to HAU. The humans says, Oh, well, the caller asks for, you know, whatever.

And they categorize that intent and they do that through an interface that we create for our humans that basically anticipates what is likely being said. It makes it very, very easy and fast for our agents to decide what is said and to let the computer know this is what was spoken. And then after the fact, our system looks at every utterance that was basically analyzed by a human and then feeds that back into our grammars.

So unlike a traditional system where literally hundreds of hours and many cycles and months of time can be spent going back and tuning grammars and and really digging deep, and there’s, you know, with most of our competitors, there’s literally big teams of people who are managing these grammars and doing grammar updates and that kind of thing. And it’s expensive.

It’s expensive for their clients, right? But our clients don’t have that expense, nor do they have that hassle. And better yet, our grammars autotune with real utterances in real time. So it’s super powerful and a big timesaver and it makes our ASR get insanely good. We have a client, a very large client that has how can I help you opening where they know every single color is created.

How can I help you? This particular client has dozens of products across a broad spectrum. And inside of some of those products, there’s another 40 or 50 features that can be purchased. Right. And at that, how can I help you opening at the beginning, we have around a 95% ASR success rate. That’s ASR. There’s no humans.

And the reason why is because this has been a long standing client and over time, tuned and tuned and tuned and now the ASR’s insanely good. So at opening you can say pretty much whatever you want, instant response. And of course, for the 5% of callers that are saying something odd or doing something unusual or whatever, we still pick that up with HAU and get it up to our normal percent.

But over time ASR becomes fantastic in our systems.

Peter Mullen
I have a real world story about this client that we are talking about. Mike I was at an event in Brooklyn several weeks ago and I was manning a booth and I got tired of explaining in marketing speak what our solution offers and why it’s so great. So when several people came up, I said, Let’s do an experiment.

And I called this client of ours and I said, I’m in Brooklyn right now. It’s raining. What do I do? Should I go to one of your stores? You and I both know that that breaks every IVR ASR experience in the entire world except Interactions.

Well, I solved that in under two and a half minutes while people were watching me. Thanks to our interplay between the human and between the machine, it figured out that I was looking to go to a store. It found its stores nearby, and it did not tell me whether it was raining or not. But I figured that that’s okay In a year or two, it will. That’s your job Mike.

Huge success. And everybody said, I get it. It was that moment of time where their eyes lit up and they said, Aha. And when you’re a product market or when you’re a CX designer, that’s what you just live for, right?

Mike Pell

Peter Mullen
You said something a second ago, it was so clever. I wrote it down because nowhere else can we have a conversation where someone says we are doing grammar things. I absolutely love that. I want to come back to that in a second. But before we do, this institution that we have created that is being deployed in multiple instances at the world’s best brands is something that we have so much confidence in that we have success based pricing. And what that essentially means at the highest level is if we do not deliver, we do not bill the client.

And this has been the hallmark of Interactions for nearly forever. It’s what sets us apart because of the confidence that we have in all of these engagements, from utterances to color resolution to the improvements that happen on the back end. And Mike, you and your team get a tremendous amount of credit for the success based pricing model that it can even exist.

And it’s because of these refinements that you do. And I just want to call that out and actually just express my gratitude on how cool it is that as a marketer, I get to talk about success based pricing in an enterprise environment. But nowI want to connect that to the doing grammar things that you just said. To continue that success and that retail example that you just used about that refinement is incredible.

Once we are up and running and we are having a positive engagement, how often are you tweaking and tuning the machine from a CX design standpoint?

Mike Pell
That is a great question and I mean, to be really honest, it depends somewhat on the client. And there’s different types of tweaking and tuning, right? We have a different large retail client that does product releases on a regular basis. And so we’re constantly adding products and refining it that way.

The improving of the grammar honestly never stops. As time goes by and clients change marketing campaigns or they introduce new products or they whatever. The things that people say at the how can I help you can just change because our client has changed, right? And the things that are out there in the world that drive people to call have changed.

And so as that happens, we may need to actually go back into the system and design new pass through or change the types of things we understand, etc., but we never stop improving that grammar. The utterances that are interpreted by the humans continue to influence that grammar over time for all of our clients. And then, something that I would be remiss if I didn’t say is that, we’re not a huge company, but we have a pretty good installed user base.

And I’ve been involved with most of our clients in one way or another over the years. And one thing I will definitely say is that when we have an engaged partner as a client, and when we have successfully, sort of nurtured our relationship with the client as a partnership and we are working together, those are the best systems.

When our partner has a team and we have a team and those two teams are doing everything towards a common goal, we just have phenomenal systems that do incredible things. And, like you said, I mean, I think anybody who has worked here for too long has done exactly what you did at that conference, which is try to explain what the heck it is we do and then be like, you know what, Let’s just call.Call them up and improve it, you know, because then it’s like, wow, that was great, it sounded awesome.

And it’s a nice shorthand way to show really the automated systems we make are better.

Peter Mullen
Well I have a thesis that over the next 15 years, the number one predictor of recurring revenue for B2C companies is going to be customer satisfaction. And COVID has helped enforce this fact that you’ve got to treat the customer right. And this personalization and individualization that is occurring at light speed in all aspects of our life from what we are doing with our phones to our ability to switch airlines if we’re unsatisfied with the treatment that we’re receiving from one, is just unprecedented.

And when you think about the predictability of revenue being based on customer satisfaction and take that hypothesis and extend it. What that means then is that the premium for making the customer happy, giving the continued delight, which leads to enduring trust, which therefore leads to continued relationship, is, I believe, one of the biggest trends that business is going to see.

So when we get into the nuances of our business and a conversational category that by some estimates Gartner included, has over 100 people playing in the space, it becomes certifiably hard for a client to know what is the difference between Interactions and the thousand other players that are out there. But it is this CX design, right? It is this engagement and interplay that happens back and forth, and then it’s this continued refinement that goes on.

I don’t like using the word or the phrase white glove service, but in many respects, we’re trying to own the corner of the market that cares about having phenomenal customer experiences because we think that it is what drives business success over the next generation. And that’s what it comes down to with all the work that you do.

I know some of that’s kind of provocative. What are your thoughts on what I just said?

Mike Pell
Well, I totally agree. I was just mentioning, you know, our best systems are when we are working together with the client. Right. I mean, there’s really no dispute. You know, we’ve got a ton of installed systems, but the ones that really stand out and are fantastic experiences. The one where the client is very engaged with us.

And I think to an individual, those are also the ones that are very, very interested in their customers experience. Right. And we’re pushing them and they’re pushing us to excel in this area and to treat their customers better and and come up with new and interesting ways to make sure that their customers feel that.

And you really do see a tremendous uptick in both the quality of the system but also the experience that their callers and their customers get when they call our systems.

Peter Mullen
And when you are talking to somebody who is on the fence about using us versus another solution, and there are many good ones that are out there, what are the couple of things that you talk about to try to demonstrate to them how we are differentiated?

Mike Pell
Well, I mean, you mentioned white glove treatment. And I really do think because when you purchase an Interactions system, you’re really purchasing Interactions. That’s not always the case. I just had a prospective client actually last week and one of the things we learned was that they had been trialing one of our competitors products and they had spent a year, 12 months on this trial, and they had decided that it was too much, was too much work.

There were many things that surprised them that they had to do. And this is a huge company, one of the biggest companies in America, that provides a service. And they didn’t get the help they needed. They had really struggled with the grammar tuning, which is a problem for anybody who hasn’t spent a lot of time doing it.

And basically they said, this is too much. We can’t do it. When anybody purchases an Interaction system, they get me or one of the people on my team who’s in some cases more qualified than I am. They get a phenomenal QA team that does a ton of work.

And so they get all that grammar tuning. Again, with an Interactions system, while the price you pay is roughly the same as any of our competitors in terms of, you know, what it takes to get a system up and running that works well, what isn’t necessarily obvious at all times is that you also get our entire company supporting this project and trying to make sure it’s the best thing it can be.

And and we’re doing all that work that, in the case of this prospect that again, I spoke to just last week, all that was going on that they had to put together an internal team of people who this isn’t their job or wasn’t before trying to figure out how to do all this stuff and reading the manuals and, you know, contacting support.

And it got so frustrating, actually, that the company they’re working with ended up assigning them a helper to kind of give them advice and help to guide them through the process. And it wasn’t enough because, again, all the work was still falling on them. With Interactions we do all that work and we make it easy and we make sure obviously to do all the due diligence to make sure that all the businesses are being followed.

And one of the things that I think is kind of a funny side effect of how we approach these things. I’ve got a guy who was for me, Jeff Dawson, who works with some of our biggest and, sort of, most well-known names. And many times his clients will call him to ask him questions about their business, because Jeff is so ingrained in the process that nobody on their side knows it as well as he does.

He knows all the business rules. He’ll have peers that just call him up to ask him questions, not about what we do or not about these services Interactions provides, but literally about what their rules are because we’re so plugged in.

Peter Mullen
Yeah, we use the phrase internally and it probably came out of your team, “We do not beta test on our clients”, and nothing is truer than that statement after what you just said. It’s a phenomenal discriminator and differentiator of how we work with people that are engaged with us. Mike, we dove into this so quickly, which I absolutely love that I didn’t even take a breath to talk about you. How did you even get here?

Give me a little bit of background about your overall journey, your academics, and how you landed in this spot, providing all these great services.

Mike Pell
Sure. I mean, it was a circuitous path. Nolittle kid wakes up one morning and says, you know, I want to design IVAs for a living. You know, that is not how you get here. So I went to the University of Iowa and I graduated with a degree in philosophy, which, of course, is perfectly applicable to what we do.

And I basically got a job working for Apple in their call center doing support. And I was on the phones, front line person and I became a lead and then I became a manager and then I was a trainer and then I was a project manager. And then I went to another company and again, really staying in the call center world and I got hired by a company called RCN, a Cable Company.

They do major markets. I was in Chicago and I was running their call center there, and I did that for about four years. And one of the guys that I reported to left and founded this company called Interactions and he called me up and he’s actually going to hire me to work at RCN and he was our founder here.

He called me up and said, Hey, I really love for you to come work for my new company and he explained to me what Interactions was doing. And we didn’t even have a client yet. There were no clients. And I said, Well, what would I be doing for you? And he’s like, We’ll figure it out.

So I’ve been figuring it out for, you know, 18 and a half years. And early days. I think I wore all the hats as well as we all did. But then probably about 15 years ago or so, I started focusing on client relations and system design. And once we decided to make a design team, I was the very first director of that design team, and I stepped away for a while to become sort of our product architect.

But now I’ve come back doing the design lead again and I love it. I love Interactions. This company, to me more than any other company I’ve worked for, including Apple, which I think does a very good job of this, really nurtures its employees. You know, I feel like I’m a member of a family here, and I really mean that.

I think that’s a very important thing to our CEO, Mike Lacobucci. He really, really wants to help his employees and be a person who’s looked upon as an altruistic leader. And I think he is. You know, I’ve really enjoyed my time here and I look forward to many more years of doing this.

Peter Mullen
That’s fantastic. And I think that this incredible culture that the company has, and I agree with, informs everything that we do all the way down to grammar checks and updates. I think, for example, we had already started moving into our remote first culture before COVID hit. We are now a fully remote company. Just in the past two weeks we’ve had two feature articles on our culture, which has now become in the Eastern Seaboard a bit of a model for how you move into a remote world and how you continue to nurture, mentor and grow the teams around you.

And I can feel that flowing in. Mike, I did not know you had a background in philosophy, but it makes total sense from getting to know you a little bit how we think about the goodness of everything that we’re trying to do in the universe and bring that down into an ASR system. I think that there is actual intentionality behind that and it’s a great thing.

I also think that there’s a ton of moral, ethical components to everything that we’re doing with AI. We don’t have time for it in this particular conversation, but I think next time it’s something that you and I should really dive into and think about how we’re influencing so many millions of people. So that’s my hook to you to have you come back another time so we can keep having the conversation going. Sound good?

Mike Pell
Yeah, sounds great. This has been fun.

Peter Mullen
For me too. This has been very rich and rewarding. So I’m Peter Mullen, CMO of Interactions, and we’ve been talking with Mike Pell, who heads our CX design team. Thanks everyone and we’ll talk to you next time.

Mike Pell

Want to learn more? Let’s talk.