Welcome to the Travel Again podcast Spotlight series, covering the people, products, and technologies that bring the business of travel to life. Please welcome our co-hosts, Mike McCormack and Ed Silver.
Hello Mike, nice to see you as always. Good to see you, Ed. Mike, this is the launch of our new Spotlight Series where we highlight companies, people, technology products, and innovation from around the world of travel. This complements our podcast series and, as always, focuses on bringing clarity to complex issues in the business of travel.
Yeah, I’m excited. This is a really great way for us to probe a little deeper, discover some new products, gather new information, and kind of look at what’s out on the horizon. I think it gives us a nice new forum to do that and some more terrific content for our influencer network that’s out there.
I think it’s great. Today we will start our Spotlight series talking about a new company focused on helping businesses get their hands around how to use and deploy generative AI. It’s a hot topic as always, and it just so happens that the new founder is an old friend and colleague who is now a true expert in the world of AI: George Roukas. His new company is called Guyapan.
George and I first met back when we were working at Biztravel. I just have to say, I know you and I really rely on George from a standpoint of being on the forefront of what’s interesting, new, and impactful with technology products as it relates to travel. As long as I’ve known him, he’s been such a great innovator. It’s no surprise he’s jumping into AI and all that goes with it. It’s an exciting topic and one with a tremendous amount of resource going at it by some very big players. It should be a lot of fun to talk about.
Of course, the three of us also met and worked at Cendant Travelport as well as started Hudson Crossing together, so we go back. Let me introduce George. George Roukas is a senior executive with in-depth experience in product management, technology, and competitive strategy. In 2007, he co-founded Hudson Crossing, a management consulting company dedicated to the travel industry. Before that, he was the Group Vice President of Product Management for Travelport, where he led strategy development and management for all products facing Galileo’s North American corporate and leisure agency partners. He also participated in the leadership of several early e-commerce companies including Biztravel, Room12, and https://www.google.com/search?q=ClickRadio.com.
At the end of 2022, convinced that new generative AI models would have an unprecedented impact on nearly everything we do, George left Hudson Crossing to study AI specifically, its applicability to business full-time. He now advises companies on how best to adopt generative AI through his new company, Guyapan. Please welcome George Roukas to the stage.
Georgie! That was a great intro, thank you very much. You’re very welcome. It’s so good to see you on our podcast. Welcome. We’ve got some questions for you today, and Mike’s going to kick it off.
Yeah, let’s dive in. George, let’s just start with the obvious. You started this company, Guyapan. Why the name? Why did you do all this? I feel like you went off to some AI monastery somewhere and studied for a year. We could barely get ahold of you for months at a time. You had some automated voice talking to us pretending it was you, but it wasn’t you. It was amazing.
That’s a great way to think of it. First off, as with many things, it was almost entirely unplanned. I was getting into the middle of 2022 and had started thinking about how management consulting is a great business to be in, but the schedule is just super demanding. I thought I should start thinking about winding things down soon. Then I saw a demo of—I can’t remember if it was Midjourney or DALL-E—which are image models. It was the first time where you could say, “Give me a picture of such and such,” and boom, 30 seconds later you’d have example images.
When I saw that, it rocked me. I thought, “This is a sign. It’s an omen.” It was time to hang up my business consulting spurs and just play around with this stuff. In the beginning, I wasn’t really thinking of doing it as a business; I just wanted to learn as much about it as I could. I worked for Hudson Crossing, which is a terrific management consulting company for the travel industry, but it was just time for me to take a new direction.
I spent 2023 really throwing myself into it. It helped tremendously that DALL-E was released to the public in June of ’22, and in November of ’22 was the first time we got to see ChatGPT. There was the image model and then the language model, and I thought it was a good thing I decided to take this track. I really immersed myself. It was great to be able to do that because generative AI has so much potential that is still potential.
If you remember back then, there were all kinds of issues with it. It was very expensive to run, it had hallucination problems, and it was non-deterministic, so you could ask it the same question ten times and get ten different answers. That unnerved a lot of people. Understanding the ins and outs of how those things happen was key. Plus, we had a whole bunch of different models spring up, and each model seemed to have its own personality that was a reflection of how it was trained and what the developers wanted to focus on—some on code generation, others on writing, others on getting concrete answers to business questions.
All of the models have a learning cutoff date. You may be dealing with a model where—as you experienced the other day, Ed—you asked one of the very large and well-used models a question about something in the future and it gave you a bizarre answer. That’s because today’s models have different abilities to answer things that are current. You have to pick the right model. Anyhow, it was complicated enough that I thought it was great fun.
At the beginning of 2024, I started to have a couple of former colleagues ask me about generative AI. We had some conversations, they introduced me to some other folks, and we started talking about whether they were doing it right. That was a lot of fun too, but I didn’t want to get back into full-time consulting, so I decided to just hang back a little bit. As for the name Guyapan, it really has a couple of origins. One is G-A-I, obviously for generative artificial intelligence. Pan is a suffix that means very broad, so it’s sort of a shorthand for generative AI everywhere. And I’m a big Clavell fan, so it was also a riff on Tai-Pan. Some people have asked if I named it after Muggsy Spanier, the Chinese dish from back in the ’60s. Younger people ask if it’s that online game that has a city called Guyapan. No problem with trademark issues there.
I feel like you have a Godfather III moment where they just keep pulling you back in. Like you said, there’s the allure and the potential, but then there are the realities of how an average company or our entire industry has to figure this out. While tools are being developed, there’s that old expression about trying to change the engines while mid-flight. This is just wild west time with all kinds of implications. You developed a framework to help companies initially start to get their heads together about what to do and how to do it. Tell me a little bit about that framework and how you are approaching it from a business perspective.
My background involves a lot of years of technology experience from the perspective of a product person. When you’re in this kind of business and delivering digital products, you have to understand the technology behind them. In order to manage properly, you must understand from wall to wall—from consumer all the way back to supplier—what’s happening in between. I came at it with that technology background and about sixteen and a half years of management consulting, which was really about understanding. I dealt with easily over a hundred different clients during my time at Hudson Crossing. That means a hundred plus different managements you have to talk to.
It’s really a lot about not just finding a solution for them, but understanding the motivations, possibilities, and aptitudes of the companies and their leadership, then matching it to what you think a good solution might be. You might come up with a solution that’s a home run but understand that this particular group can’t execute on it. Other times there are companies that can execute on the big idea but are not quite sure, so you have to help them along. That understanding of what’s possible and practical was one of the hallmarks of Hudson Crossing. That is the second way I look at it.
The third way is having that year off to really focus on generative AI and get into all the warts of it. I wanted to understand where people were succeeding, where they were failing, and what they were doing that might be sub-optimizing. There are untold examples of companies that have put together an individual use case and applied generative AI to it and come out with a positive result. But what then? I don’t think they had a real good way of prioritizing them, and they certainly didn’t have a good way of thinking about where they were ultimately going with this or how to provide for a future where all of this fits into a cohesive whole.
You don’t have to do it on the first project, but you should at least have a vision for where you want to get to before you start deploying stuff. We saw what happens when you didn’t do that with databases, with cloud computing, or with agile development. If you don’t know where you’re going, any path will take you there, and most of them are bad paths. The idea with the framework was to give people a vision of the future. The other thing that was almost universal is they had no way to prioritize. If you don’t know how to prioritize work in an impactful technology like this, you may wind up with a couple of great examples, but you’re never going to get towards an optimal situation.
I encourage them all to think about it holistically and come together with a prioritization scheme. Then we start to take a look at how companies can actually deploy a thoughtful process for adopting generative AI. I put together the framework which includes a deployment matrix. It asks where you want to go and what is the appetite within the organization. If you are a manager in a business function and you want to do something that’s going to change the whole competitive strategy of the organization, it’s not likely to happen. What’s your domain? What’s your sphere of influence? How far can you take this? We start from there. It’s a six-week process I use with clients. It takes six weeks to review those possibilities and then put them in a structure that will serve as a three-year or four-year vision, giving you a good chance at a path of deployment.
That’s great. As a part of this, you’ve also been doing some writing. You put together a series of articles. The first one focused on Google and their impact, another focused on Apple Intelligence, and then moved on to loyalty. We’re going to help publish and get these out to our network because it’s really insightful commentary. What inspired you and what will readers find in them?
One of the things I noticed as companies were looking at generative AI is that they were running a race staring at their feet. They weren’t looking up enough at what generative AI could do for them. The first step, which I agree with wholeheartedly, is just bringing a model into the organization and sanctioning it, saying, “Here is a model and here’s how you can use it to help you do your everyday tasks.” That’s great.
The second level is how we deploy these things at a business function level. The best example of this is customer service chatbots. We know you can create a chatbot these days in a day that is somewhat trained on your specific organization and can answer customer service questions. Some companies have taken this to the ultimate level and done amazing things with it.
The third level is: how can we make our products and services more competitive or more valuable? Can we improve the business value proposition? Can we reduce prices? The last level is competitive strategy. How might generative AI, as the big wave comes in, impact our company’s competitive strategy? The articles came about as I was talking to early customers trying to get them to think about how it might impact their strategy.
When I saw the Google announcement about the enhanced trip planner for generative AI back in May, I was amazed. The demo was an individual asking in plain speech, “I want to go to this place for four days with my husband and two kids, and we like this and that,” and it generated an entire itinerary for her all at once. The itinerary was not just a list of possibilities; it was one that did two things that are really unique. First, it generated the entire itinerary with a sense of time and place. It wasn’t just, “You land at the airport at 4:30 and at 5:00 I have you at an event that is an hour and a half away.” It was aware of where things were and how long it would take to get from one to the other. It was a viable itinerary, unlike the kind you often get these days. It was also relative to the people traveling, recognizing there were children to be entertained.
The second thing it did, which I think is also equally impactful, is it really personalized the itinerary. I say that as someone who is in top-tier loyalty programs for a hotel company and an airline. Every time I go in to plan a trip, it is abundantly apparent to me that there is no personalization for me whatsoever. I log in as a top-tier member and I see the exact same stuff as anyone else. I have tested it by going in on one browser and then going on a different device with a private browser—I get the same stuff. Then I go to the hotel I’ve stayed at ten times and they always ask, “Have you stayed with us before?” I know there is no personalization going on.
This was different in that it read through the consumer’s Gmail account, looked at all the email messages, and based on interactions with different brands, it came up with a truly customized itinerary. Those two things—the logic and reasoning of the itinerary plus the radical personalization—were brought about by a new layer on top of generative AI called generative agents. When I saw agents, I thought this represents a whole new layer of the generative AI revolution.
Now we’re talking! Now we’re crossing over into something actually productive and meaningful. And that’s where the articles came from.
You can find the articles on traveladvisory.com as well as on LinkedIn. George, look, we’ve known each other a long time. You tend to call me the resident skeptic, so I’m going to ask this question. Generally, at Travel Again, we remain optimistic about the future, but we do worry that the hype here can never really match the reality. I read that companies are spending billions of dollars on generative AI right now with the return so far not even in the millions for many. Where are we in the hype cycle and how are you going to help companies find the right investment point? It may be too soon in some cases or too late in others. How do you help them through this phase where the hype may not match reality?
I read similar things, and there’s just so much garbage news on the web. It is absolutely false if you are trying to compare apples to apples. The companies that are spending billions of dollars on it are Apple, Google, and Meta. They’re spending tens or hundreds of billions creating these models and getting them out there. They are making a lot of money on it, some more than others. If you look at OpenAI, they’ve spent hundreds of billions and they’ve made at least tens of billions. They’re not making more than they’re spending at this point, but they’re in a build phase. Any technology project that is a product has an investment period where you make less than you spend.
On the flip side, the consumers of those products are making a lot more money. A great poster child for that is Klarna, which has revealed what it’s done and what it’s making. They famously deployed a customer service chatbot. They had about a thousand people in a remote call center—it was a service, not owned by them. They were able to replace 700 out of the 1,000 people they used. They determined that after one month, the run rate for savings would be $40 million a year if nothing changed. Those models get trained and get better at taking on questions. In month one, it was handling 70% of inbound questions and maintaining higher NPS than before. Customers were happy and the company was happy. I’m sure the BPO was not happy, but they probably have 100,000 people, so they just shifted some chairs.
People are making a good amount of money on it, and many are just not talking about the results yet. There are lots of people going out there trying to do isolated, siloed, de-prioritized applications, and they’re coming back without giving them full consideration or having a plan. They just want to check the box and say, “Oh yeah, we’re deploying generative AI,” and they’re not having such great luck. That’s where Guyapan comes in. I don’t scale, so I’m not trying to go out there and do this work with everyone at once. That’s why the intro only takes six weeks. I use AI heavily, and my intention is to create things that can help companies do it on their own. That’s what I’m building now, and I will hopefully have a website soon where I can start putting some of these things for people to pick up and use.
You gotta teach them to fish! Exactly right. George, thank you so much for coming and doing a Spotlight. George Roukas is the founder and CEO of Guyapan. Always nice to see you.
Great to have you, George. Thanks for having me.
Terrific stuff. As always, great perspective and insight from George. I’m really happy to have him on the forefront of this. For us, when we’re working with clients and potentially working together with him, I think we can do a lot of good and help companies make sure they’re doing the right things with the right prioritization so they can see the right kind of outcome for their customers.
He’s the right guy for it. Mike, that is our show. We hope you enjoyed our first Spotlight episode. If you are interested in spotlighting your company, your people, products, innovations, or technologies, feel free to reach out to us at traveladvisory.com. As always, great to see you. Thanks for another great episode.
Same to you, Ed.