AI: The future of travel?

Everyone is talking about artificial intelligence (AI) these days, but what does it mean to the world of business travel? In this episode, Chad and Miriam connect with Yannis Karmis, Sr. Vice President, Product Planning & Development, to talk all things AI. Yannis illustrates how using AI is nothing new in business travel, how we're using AI on both the front and back ends, and what's on BCD's AI roadmap. 

Learn more by visiting https://www.bcdtravel.com or https://www.linkedin.com/company/bcd-travel/

Transcript

Miriam Moscovici:

Actually, good question, though. Just wondering, Yannis, do you have any cosmetic or skin care products on your desk? You're never at your desk.

Chad Lemon:

He's always at his... I've never not seen him at his desk.

Miriam Moscovici:

Well, he's travels a lot, is my point.

Chad Lemon:

I... Well, yeah, yeah.

Yannis Karmis:

I travel a lot, but no, I don't keep any-

Miriam Moscovici:

Like your lip balm?

Yannis Karmis:

I've got my vitamins. My vitamins are over there, but other than that...

Miriam Moscovici:

Okay, alright.

Chad Lemon:

Alright.

Miriam Moscovici:

We had a long discussion about all the things you have on your desk and the morning routines because we work at home. I can do my morning routine at my desk, right? Anyway. Nobody cares.

Intro:

Welcome to connections with BCD Travel, an ongoing conversation about the modern day travel program, the impact of technology, and how travel buyers can take control, and drive change. What are we waiting for? Let's start connecting.

Chad Lemon:

Hey everyone, welcome back to Connections. I'm Chad Lemon.

Miriam Moscovici:

And I'm Miriam Moscovici.

Chad Lemon:

So Miriam, remember a few episodes ago, when we had my boss Heather Wright, on to talk about various digital products and solutions that are available for use? And I kept joking about the pressure. I was under recording to make sure the experience was good for my boss. You remember that, right?

Miriam Moscovici:

Yeah.

Chad Lemon:

Yeah. Okay.

Miriam Moscovici:

Yeah, of course.

Chad Lemon:

Well, today's guest is your boss, and truthfully, he's my boss's boss. So I guess, in a way, one of my bosses too. Anywho, his name is Yannis Karmis, Senior Vice President, Product Planning and Development at BCD. We're so thankful to finally get you on the podcast. Tell us a little bit about yourself.

Yannis Karmis:

Yeah, good afternoon. It's great to be on the podcast. I'm not sure about boss being the right word. It feels like my principal job is to make sure that you guys have all the funding and the resources that you need to be successful, and then to stay mostly out of the way. But no, it's a pleasure joining you today. And I guess I'll tell everybody a little bit about myself. I've been at BCD a little over seven years, and I run our Product Planning and Development team. So I work with everybody on the BCD side who's building applications, everything from TripSource, from our travel engagement perspective, DecisionSource from an analytics perspective. We kind of oversee our BCD marketplace and third party integrations and BCD Pay, BCD Invite. That was on the podcast a couple weeks ago. So we got a lot of exciting things going on, and I think it's a great time to be at the intersection of travel and technology.

Miriam Moscovici:

Yannis, today we want to talk all things AI, artificial intelligence. It can mean a variety of different things in business travel, and we know you're the guy to tell us about it. So you're ready?

Yannis Karmis:

Yeah, let's go.

Chad Lemon:

Okay. So Yannis, you and I were both just at GBTA, and I know you did an interview about BCD's use of AI. So talk to us a little bit about our historical record in this space.

Yannis Karmis:

Well, I like the analogy of historical record, so let's run with that a little bit and think a little bit about the history of technology in corporate travel. I think maybe, as I reflect back on it, in many ways corporate travel really has been at the forefront of adopting technology. I know sometimes it doesn't feel like that, the grass is always greener on the other side, and you feel like other industries are moving more quickly. But I think if you really look back quite a distance back in corporate travel, you really see that this has been one of the industries that has adopted to technology pretty well. So maybe rewind back to the 1970s. Call it the Stone Age of travel, right?

Chad Lemon:

I love that.

Yannis Karmis:

Almost every industry processed things really manually. Everybody did things manually. You called in orders, you punched tickets. And it was the airlines in the GDSs that brought mainframes to travel and started the automation. I think the travel industry was very much one of those industries that started with technology. Fast forward a decade to the '80s, and maybe that's the Bronze Age in travel. And what happened in the Bronze Age? We started adopting things like scripts. Let's look and see what an agent does. Oh, they need to do 25, 30 entries to issue a ticket. Well, let's write a robotic, a script, that does those keystrokes for them, and let's assign it to keys on a keyboard and let's try, and figure out how to make the processing go a little bit more quickly, and those proliferation of all these scripts and processes. A decade later, and now we're in the '90s, and I don't know what came after the Bronze Age, the Iron Age.

The Iron Age of travel was about starting to wire up databases and starting to get some graphical user-

Miriam Moscovici:

That's right.

Yannis Karmis:

... bases in front of people. And again, I think the travel industry, if you go back and look at the '80s, actually was pretty innovative in the types of technology that we were bringing online. And now we're dealing with maybe more online booking tools and the web era of travel, and goes on to the 2000s in terms of new technology that's coming online. And now, you're getting to things that are touching the consumer, the traveler. You're starting to get to things that are kind of changing the interfaces. And again, I think the travel industry was one of the first online digital marketplaces that came out.

Miriam Moscovici:

That's right.

Yannis Karmis:

By the time we got to the 2010s, everybody's working on mobile. You look around, and almost every sizable travel company today has a mobile app.

Again, I think the travel industry was one of the first to get mobile applications out. And so, we're sitting here today in the early 2020s, and I think we've got the next evolution of technology, and I think it is pretty much predicated on where artificial intelligence is going to go and how artificial intelligence is going to change the shape of our industry for the next decade. And I think that's kind of exciting. Of course, every new technology brings a ton of questions, a ton of unknowns. There'll be some companies that move very quickly. There'll be processes that are better apt to move and adopt artificial intelligence earlier than others. But I think it is potentially the driving force for the next decade or so of innovation in an industry, and I think we should be excited about that.

Chad Lemon:

So you're saying using various aspects of artificial intelligence, it's nothing new for the travel industry or BCD in general, right?

Yannis Karmis:

No, I think, well, first of all, we've been evolving for decades on technology, and I think even on the BCD side, we've actually been using certain capabilities in artificial intelligence now for a couple of years.

Miriam Moscovici:

Now, I think it's important for our listeners to note that today many people are using the term AI, but really, that could mean any one of a number of things. And when we use the term AI in business travel, what does that mean?

Yannis Karmis:

I think, in its simplest form, you'd say AI is really the science of making machines, making software think like humans. Now, that's a bit esoteric. I think, in the practical sense, it's around leveraging technology to try and better understand the way humans process knowledge and how we synthesize information in different ways. So historically, when you wrote a software program, you had to code for every possible eventuality. Think of writing a piece of code, a piece of software that analyzed emails. So you would have to understand all the possible words that you would see in an email, and what the meanings were, and how different words created different phrases. And that's kind of a daunting task, right? There's way too much vocabulary and syntax in the human languages for somebody to go word by word and build that model. So it was very difficult to use semantic information to create software.

So with artificial intelligence, a software developer can now take a dictionary of terms, take the syntax, and, if you will, the neural network that is being developed more broadly by companies like ChatGPT and others, and we can start to train it or teach it how to leverage that dictionary to be able to write code. So you don't need to train a bot on everything to do with understanding an email or understanding a text message. What you need to do is help understand how to build the technology in a way where you're leveraging those neural networks.

Miriam Moscovici:

Yannis, would you say there's a difference in using AI on the front end in business travel, as in with travelers, versus the back end, as in with agents, and is one more important than the other?

Yannis Karmis:

Yeah, I don't know that one is more important than the other, but I do think there is a difference between thinking about AI in the front end and the back end. So at BCD, we've actually been using some AI in the back end now for a couple of years. We built some intelligence around our hotel platform, where we're doing room and bed type mapping. We built some intelligence into our BCD Pay platform in terms of matching transactions, and that's using artificial intelligence and some of this latest technology. But that's all sitting, if you will, on the BCD side. It's not engaging with a traveler directly. And we're excited about the possibilities. I think we're just scratching the surface there. But what some of the new advances in AI, things like generative AI in large language models that are really focused on developing this syntax and this language-based kind of understanding of AI.

What this allows us to do now is kind of bring that from the back office, if you will, to the front office, moving it from processing of data to now moving it to a place where it can engage travelers. If you haven't played with ChatGPT, I really urge you... Mark an hour in your day, sometime over the next couple of weeks, create an account, go in, and just play with it, and ask it some questions, and see how intelligent it is. And it's remarkable, right? It really is very capable of understanding what you are asking in terms of a question - and ask the question with different syntax - and then going out and finding answers. And I think for the first time we're now able to think about how to take this technology and move it into a place where it can engage with a traveler. It can answer questions that travelers have. I think that's going to open up a whole new sphere of innovation as we think about new ways of meeting travelers in different applications, in the different places that they're working.

Miriam Moscovici:

Now, let's get more specific. So we're clear on where BCD Travel has been, but where are we going in the future as a leader of our technology, right? What's on the roadmap, so to speak?

Yannis Karmis:

Yeah, so I think, in terms of front end, we've actually been working on a couple of things. So we've been prototyping an email bot. So about a year ago, maybe even nine months ago, we built our first email bot. It's application that can comb through emails and categorize them, and in theory, it can take some action based on different use cases that it sees in email. At the same time, we're also out talking to a number of third parties. I think the technology in this space is moving very quickly, and it's great to go from demo to demo. I think everybody's technology that you meet who's focused on AI in the corporate travel industry they've got fantastic demos, they've got great things that they're piloting and bringing to market. And I think we are very much focused on AI is going to be one of our top five, maybe even top three initiatives for 2024.

But now, it's trying to understand what do we want to build, how do we want to partner, how do we bring this to market, and what are the processes and what are the ways that we want to deploy AI? So we're very much working through that synthesis right now.

Chad Lemon:

Now, I want to be clear here, you're not suggesting that AI is going to be making random travel decisions unsupervised, right? Humans still need to manage things like cost and policy compliance, and privacy, right?

Yannis Karmis:

No, absolutely not. I don't think there's a scenario where we would ever put an unmanaged AI bot in front of our travelers. So I think there's maybe two ways to think about this. The first one is, almost all of our customers today have a well-developed, highly honed travel program. They have already set policy, they've got content decisions that they've made, and low fare algorithms that they've deployed. They know what type of reportable information they need to help complete a transaction, and what billable project codes need to be associated with a trip. Those are well-defined structures of their travel program. So I think for us, as we're thinking about deploying AI, we want to leverage those settings, those decisions that have already been made, those configurations that are already set up, so that in principle, if an AI bot is helping to make a reservation for you, it's using exactly the same policy with exactly the same content as our TripSource and AgentSource experiences, or ideally the exact same policy that's showing up in your OBT.

We'd like to do it in a way where we're not proliferating one more place to set up the same settings, configure it in your OBT, and then BCD configures it in our applications, and now you need to go configure it in an AI bot. And we've got triplicate versions of the settings going on. So I think the more we can kind of hone that and drive it into one place where we're able to say to a client, "Look, it doesn't matter which way your traveler chooses to engage with us." Go to an OBT, go to TripSource, call an agent, go to a chat or email interface, "you're going to get the same policy parameters deployed in any one of those scenarios." I think that's really, really important. The other thing I think I would stress is with any technology, but I think in particular with artificial intelligence, process matters, and I think process and process design is going to matter a lot.

So let me give you example. A minute ago, I mentioned we had actually built our own email bot a few months ago. So we developed it, we came up with a handful of use cases that we wanted that email bot to be able to understand and be able to synthesize. And we tested it against 12,000 emails, so 12,000 random emails. And then we ran our bot against it to figure out how good it was. And actually, it was pretty good, certainly, as a first attempt, we thought it was actually very good. So one of the first scenarios we looked at was a scenario where a traveler is emailing BCD to have us resend their invoice. We get a fair number of requests from travelers that just got back from their trip. They're trying to do their expense report, and they want something out of the invoice.

And they've emailed us asking for that. So we built a bot to be able to do that. And surprisingly, out of the box, it was 90% accurate.

Chad Lemon:

Wow.

Yannis Karmis:

So we thought that's actually really impressive. But like all technology, it wasn't 100%, which means we did have some scenarios where the bot got it wrong, and about 5% of the time it thought the traveler was asking us to resend the invoice, but actually the traveler or the email was about something else. And so we got it wrong. Now, if we had automated that end-to-end, for the vast majority of travelers, we would've given them exactly what they wanted, but about 5%, we would've automated a response back to them with, "Here's your invoice." And they would've been looking at the email response back from us, scratching their head, "Well, that has nothing to do with what I asked for," right?

Chad Lemon:

Right.

Yannis Karmis:

Now, I would characterize that as the parking ticket violation. It's annoying. No one wants to get it, no one wants an irrelevant email sent back to them, right? But we didn't really break anything. But what do you do for that 5%? I'll give you a different scenario. We also had the email bot looking at canceled trip requests, right? So we obviously get a lot more requests for people asking us to cancel a trip. Well, not surprisingly, that bot was also a little over 90% accurate. Again, we felt really good about that. And wouldn't it be great if we could just add the automation at the other end and say, great, we know that the email's asking to cancel a trip. Let's just automate the cancel trip process and call it a day. Don't even bother sending it to an agent. But once again, we had 4 to 5% errors and, we can't just accidentally cancel 4 to 5% of people's reservations by sending an email that's not a parking-

Miriam Moscovici:

Don't laugh, Chad.

Chad Lemon:

No. Can you imagine?

Miriam Moscovici:

That's a felony.

Yannis Karmis:

That's a felony. There's going to be an accounting of that error. So I think that that's where process comes into place. That's where design comes into place to say, hey, wait a minute, before we take this action, does a human need to review it, or before this action, do we need to confirm it back to the traveler? "Hey, it looks like you want to cancel your trip to Dallas. Are you sure that's the step you want to take?" There's got to have to be some process that we think about validating what the bot is looking for. So it's great when you see these proofs of concepts. The technology's really, really powerful.

Now, the question is how do you build it into our travel management processes? And I think that that's where, at least at BCD, we have the benefit of the great operations teams and great process developers to sit down and say, "Now, how do we take this wonderful new technology? But how do we use it in a flow chart of processes that actually make sense in corporate travel to protect the integrity of the travel program and to protect the integrity of the service request or whatever the traveler is asking us for?"

Miriam Moscovici:

Isn't it also that the artificial intelligence capabilities are just extending the same solutions to new channels, which could be email, or messaging, or chat of some kind, right? And whether your policy and your process, and the great service you get is coming through TripSource, or through AgentSource and served by an agent on the phone, or through one of the other channels like chat or email via artificial intelligence, it's still all in the same mothership, and it's still all providing the same access, same content, same service, right? I mean...

Yannis Karmis:

I think that's absolutely right. And I mean, I think, if you think about it from a traveler's perspective, they still generally need all the same things. I need to make a reservation, I need to cancel or change the reservation, I want information about my reservation. The things that they're trying to accomplish are the same things. We have some tools today that are actually pretty well advanced and do many great things for a traveler. Many of them are available through web and graphical user interfaces. How can we leverage that infrastructure now and make that available through chat interfaces and email interfaces, and do it in a way where we're meeting the traveler where they like to work? Most of our travelers are working in Microsoft Teams, and Slack, and these other places. Let's make it easy for them to engage with us where they're working.

Chad Lemon:

Yannis, we both know you, and you're always on the road, and you never quite frankly sit down. So we know you have some stories here. So during this Quick Connect, I'm going to ask a question or two and just say the first thing that comes to your mind. What is the one thing or task you wish you could remove from your daily life?

Yannis Karmis:

Other than expense reports, right?

Chad Lemon:

Yeah. That's a given for anyone.

Yannis Karmis:

Yeah. I think, I'd say I'm very wary of team and Zoom meetings where I look at the attendee list and it exceeds 10 people. There's a lot of steering committee and project updates, and I generally find that the level of value I bring to those meetings relative to the amount of time they take is probably disproportionately not good. And so I try and push back a little bit on getting too involved in some of those larger meetings.

Miriam Moscovici:

Mine would be filling the water in my humidifier every day.

Chad Lemon:

Okay. I swear to everyone listening, we did not plan this, but mine too deals with water. I hate that I have to drink so much water every day. I wish I didn't have to drink water. Weird, but just annoying to me. It's annoying.

Miriam Moscovici:

It's daunting, you're daunting.

Chad Lemon:

It's a daunting task, yeah. It's a lot. Oh, man. Do either of you have-

Yannis Karmis:

If you want, I've got a water one too, because I've gotten hooked on carbonated water. My wife got me one of those carbonated water.

Miriam Moscovici:

SodaStream.

Chad Lemon:

Oh, yeah, yeah.

Yannis Karmis:

Yeah, exactly.

Miriam Moscovici:

SodaStream, not a sponsor.

Chad Lemon:

Not a sponsor, but call us.

Yannis Karmis:

And I will say that for the first three weeks or so, I was over carbonating, and I didn't realize why I was having all these chest pains. So I dialed it in now, but I find that I am filling up the water jugs probably at least once a day.

Chad Lemon:

So for anyone listening-

Miriam Moscovici:

That's funny.

Chad Lemon:

... please, think about your water story. Okay, last question, and this has kind of become a little tradition, but when you travel, window or aisle seat?

Yannis Karmis:

So I would say, I go by the rule of sevens, which is, if I have a flight leaving before 7:00 AM or landing after 7:00 PM, I want the window, if not, the aisle.

Chad Lemon:

Oh, I am flabbergasted right now. There's no other word for this.

Miriam Moscovici:

I'm impressed.

Chad Lemon:

What?

Miriam Moscovici:

I actually think I might incorporate that.

Yannis Karmis:

If it's early or late, I prefer to be against the window. I don't want to get up for anybody. I can generally sleep okay on flights, so I just want to be left alone. But if it's the middle of the day, I'd rather have the space and the room on the aisle.

Chad Lemon:

That is the most intelligent response I have ever heard of to that question, that's amazing.

Well, this has been one of the most interesting episodes I think we've had on the podcast series, so thank you so much for agreeing to come on and talk AI with us.

Miriam Moscovici:

Yeah, thanks, Yannis, for joining us, and I'm super excited to see where we head in 2024. There's a lot to do and so much opportunity out there.

Yannis Karmis:

Perfect. And it's a pleasure joining you, and look forward to all the great podcasts still to come.

Miriam Moscovici:

If you're a travel buyer, frequent business traveler, or just someone who likes hearing about artificial intelligence, be sure to download and review the Connections with BCD Travel Podcast to stay up to date with new releases and listen to your favorite episodes.

Outro:

Thank you for connecting with us. BCD Travel helps companies travel smart and achieve more. We drive program adoption, cost savings, and talent retention through digital experiences that's simplified business travel. Learn more about the topics you heard on this episode by visiting bcdtravel.com/podcast.