Navan CEO Ariel Cohen: AI Challenges and Business Travel’s Recovery — Full Video

Travel brands are eager to profit from artificial intelligence, but they’ll likely encounter several obstacles.

That’s what Navan co-founder and CEO Ariel Cohen told Skift Senior Research Analyst Seth Borko at the recent Skift Global Forum in New York City. Cohen also discussed what he’s seen in business travel’s recovery as well as Hotel Concierge by Ava, the latest update to its traveler-facing chatbot.

Watch the full video, or read a transcript of it, below.

Interview Transcript

Seth Borko: Hello. Hello, everyone. Thanks for joining us today. Thank you, Ariel, for being here. This is going to be exciting.

Ariel Cohen: It’s amazing. Thank you for having me. Great conference. Really big.

Borko: Ariel, you are the CEO of startup tech, corporate travel company. The start of this year, you were the CEO of a company called TripActions. Today, you’re the CEO of a company called Navan. Did you change companies or what happened there?

Cohen: I’ve moved. No, we actually changed our name and our entire branding roughly almost a year ago. The reason was when we started TripActions in ‘015, we only thought about this traveler, that frequent traveler that is going for a business trip. Over the years, we realized that there are so many needs.

First of all there are two sides of it, the company and the employee and the traveler, and then there is payments. We just said Visa talking about it. Payments as you go throughout the trip, and then there is expense management, and then there is meeting and events, and there is VIP travel and so many things.

We definitely didn’t think about it back then. We wanted a name that we’ll be able to say, “Hey, this is Navan Payments, this is Navan Expense, this is Navan Travel, this is Navan Pro, and so on,” and something that is very accessible. Navan, you can read it from both sides. It speaks to every type of user, and that was the idea.

Borko: I like that symmetry of the name was a consideration.

Cohen: It was definitely a consideration.

Borko: Love that. It reflects your transition from maybe some of your pure corporate travel origins to more of a FinTech broader expense management platform as well.

Cohen: I’m thinking about it as an experience for the traveler and experience for the people that are managing travel. Think about it, you go to a trip and you’re in the hotel. You book the hotel through us, and probably the flights and the car, but then you need to expense it. Everybody here in this audience did probably itemizing a receipt and it’s a nightmare.

It’s a waste of time sitting there and saying, “That was from the [inaudible 00:02:32], that’s the three types of taxes, that’s this thing and that thing,” and it doesn’t make any sense because it was digital five minutes ago, so why printing it out and doing it? That’s the entire thing. It’s the total experience all the way from the time that you book your trip, things that happens to you while you’re on the go, expensing it, the payments that you use, everything.

Borko: You guys have a full ecosystem around corporate travel, around business travel. I think Mark Hoplamazian was on earlier and he was talking about how really the travel recovery has been a leisure recovery. Business travel has lagged, you have a lot of insights into the world of business travel. Is it coming back? Is it back? What do you see happening there?

Cohen: First of all, I always need to separate Navan and our growth rate because we are still small in gaining market share. I’m putting that on the side and looking at our data team and trying to ask them on cohort space, “What do you see?” What we see, we saw last year pretty much February, March, I would say post Omicron, travel came back, business travel came back. We saw it at 300% rate year over year. Really, really with a vengeance, and I think it normalized. In the last six months, it’s normalized. The habits got back to much more C-level salespeople, business development.

Last year, you’d see everybody traveling, the engineering team, the product marketing team, everybody, so it’s normalized. I also think that what we see right now, you see recession. You definitely see recession. Yeah. You see recession mainly in tech. Across the board compared to what we would expect travel to be business travel, we’d see 20% less. Again, our growth rate is different. We’re actually going to go this year more than 50% year over year, but that’s because of our growth rate. But if I’m taking into account what we’d expect business travel to be, it’s 20% less.

Borko: If I can just double click on that comment, you said recessionary growth rates, a lot of people say that travel feels really, really healthy. I mean, maybe tell us more about that. Is that just tech-specific or are you seeing other sectors that are seeing growth?

Cohen: Most of the area that we see decline is tech. You do see some industries, but you actually also see industries that are traveling even more than what they used to travel before the pandemic.

Borko: What’s an industry that’s traveling more you think this year?

Cohen: I think entertainment. I’ll not go to specific company names, but entertainment and media are actually traveling more right now, which is interesting. I think the writer strikes and all of this impacted the trend for a certain period of time, and now you can actually see it coming back and pretty fast. That’s an area that we see a lot of travel.

Borko: I mean, the economy is not tech. There’s so much more to it.

Cohen: Exactly.

Borko: One of these conversations we’ve been having a lot is around this idea of blended travel. People who are traveling for business, use cases, but also tacking on or adding on, leisure use cases. If I’m not mistaken, you guys have launched or are launching a new feature around that. What’s the thought there?

Cohen: People that are using our platform, so we are (business-to-business), but we always allowed the employees of a company that is using us to use us for their consumer travel. It’s a different functionality mostly geared into vacation and the family is coming with you, and so on. Five months ago, we introduced a rewards program where we pay your employees to save you money. If you’re basically deciding to get a better rate for hotel and stuff like that, we’re going to reward you. We give you rewards. You then can use these rewards for your personal trips on our platform.

We see this entire bleisure thing to your point of like, I’m adding two days on a business trip. This is a major, major area of growth for us, but also I’m taking my family for a trip on our platform, because first of all, we have really good inventory, good prices. Also, you can use this rewards program when you book your personal trip.

It’s something that companies really like because it’s rewards for their employees, it creates a lot of engagement. But it also meets this thing that if I’m going all the way and, I don’t know, I’m in London and the weekend is coming, can I stay for an extra two days? The functionality that we’ve introduced is the ability to separate. You book the five days trip and we know to say, “The three days are paid by your company,” and they have visibility to it, but these two days are paid by you and they don’t have visibility to it. That’s the interesting thing that we did there.

Borko: Have you seen a lot of uptake of that?

Cohen: That’s one of our fastest growing areas right now. I would say that the two, and it’s interesting because we talked about the economy, the two interesting areas that we really go these days are bleisure and also non-managed. Non-managed, I’ve never managed travel in my life. I’m maybe an office admin. I want some organization there and I’m going to introduce business travel management.

Borko: For the folks in the audience, managed travel is when a large Fortune 500 company, they’ve got a very prescribed policy and plan, you have to book within the plan. Unmanaged is maybe smaller enterprises or just companies with less of a structure around their travel plans.

Cohen: 100%. I would even say, and maybe it’s a, I don’t know, corporate travel dirty secret, even if you are a big corporation and you are managing travel, what we see is that half of the employees are not going to use the platform. They’re going to go rogue. That’s the industry term. We see it when we are converting. Customer that used to use, I would say, more traditional travel management company and goes to us, and you do see the 50% adoption there. It’s your users, your road warriors, your travelers are telling you something, “We don’t want to use this thing.”

We consider this as non-managed as well. We know to convert that to using our product, but we also know to go to companies that, for the first time, wants to manage travel. I’m always joking and I’m saying, “They don’t want to go to the dentist in the process, because if you look at the tools that are out there, it’s hard to bring it to a company that never managed to travel.” That’s an area of growth for us.

Borko: I wanted to ask you, let’s talk about technology. You are a tech-forward company. You’ve built your platform from the ground up around technology. The big buzzword we are required to mention in every session, it’s artificial intelligence. You’ve done a lot with AI. Tell us about that.

Cohen: I tell you, I think we talked about it there that I saw this meme that this guy is coming to the stage, maybe it’s a joke about me, and it is coming and the only thing that he says is, “AI,” and he leaves the stage. Maybe that’s what you’re going to do.

Borko: It’s a round of applause. Well, you’re an AI guy. You’ve got a lot of AI.

Cohen: Exactly, but let’s talk about it seriously besides, I would say, just the trend. We, as a company, from day one, we said that people will want to look for personalization. That’s why they are calling to a travel agent and they want the travel agent to know them. From day one, since we started eight years ago, we had a pretty big machine learning team that was doing the Google thing, which is more of, “We’ll optimize the search results based on your past behaviors. You are likely to book the things that we’ll show you first,” so saving a lot of time.

Obviously, generative AI, which is more new, is allowing you to really understanding the user. That’s really what it allows you to do. You can understand it on a language level. I can understand what you want from me, but I can take it further and anticipate what do you want to book, what you want to book there. It has the ability to create this personalization that you’re getting usually on a conversation with a person. That’s what we’ve released today and we can talk about it more.

Borko: We’ve got a new release. If you’ve actually got a video clip we can put up of your… You have a chatbot called Ava, but today it’s going to become I guess more than a chatbot. You’re launching some really new concierge features.

Cohen: Exactly. Ava was always trying to understand you. The obvious thing was to go to support. “You need to change your trip, you need to get something from us.” For quite some time, we’ve been using Eva for that. What we’ve released today is Eva, that you’re basically telling Eva, “Hey, I want to go to London, so where should I go?” Based on my past behavior, it’ll tell me, “Hey, these are the three hotels that you are likely to want to book.” There is something here in the demo that she’s telling you, “Do you want me to book this thing for you as usual because you care about restaurants, and the gym, and so on?” It’s real. It’s not like we are looking if the hotel has a restaurant or a gym. It’s not interesting. It’s about the type of gym, the type of restaurant that you usually like to go to.

We are scanning all of your past chats with us. We are scanning all of the reviews. We are actually not relying on structured data there. We’re making some sense of it and we are telling you, “Here are the three hotels that we think that probably you want to book. Do you want us to create some comparison between them and book it for you?” To do that, from a technology perspective, you need to actually, in this very simple functionality, there are three LLMs that are involved there, three completely different LLMs. There is our data, there is our API to eventually book a trip. It’s actually fairly simple functionality but a lot of complexity behind the scenes.

Borko: LLL is a large language model.

Cohen: LLM, yeah.

Borko: That’s the kind of AI that we’re talking about here. I want to ask you a bit about AI innovation here. AI is the big thing of hype. I’ve heard people say, “Oh, we’ve been using AI for years.” ChatGPT is only six months old. Can you differentiate maybe between LLM AI and the old school AI, and how you guys are moving quickly to integrate new AI?

Cohen: I don’t think it’s old school versus new school, it’s just different branches on the same thing. Just the concept of “I can predict from a large language model what you’re going to say next or what do you want” was introduced, I think, in ‘018, and it took some time to… By the way, interestingly enough, in Google, but they didn’t do anything with it. Now it’ll be interesting.

Borko: It’ll be interesting to see.

Cohen: But I think it was introduced back then and it took some time for a real application to emerge. It’s also very expensive to process the model, so again, it took some time, but you can see it’s real. I can really understand what you’re going to say or what do you want. If you do that and then you reuse it, can I understand what you’re going to book? Yes. Can I understand what type of support do I need to give you? Yes, and in really, really high percentages.

I’ll tell you an interesting story about our technology. In our context, we could only understand you in of roughly 89%. We are having this discussion, and if I will not understand 10% of what you’re telling me, you’ll not be able to speak with me. 10% means a lot. We said, “Yeah,” but in this discussion, if I will not understand you, you’ll ask me a follow-up question. You’ll try to narrow it down.

Can we do that? Can we actually run the model again? Instead of asking you, “What the hell do you want?” already assume what would be the question that will narrow it down. It took us to 99.9%. That means that I really get you. You look at maybe the model is lying, and it will. No, it’ll definitely will. That’s more of the risks around AI. How do you capture that? I can start to tell you some… A model can tell you some bullshit things. We’ve put another model that actually supervising the first model.

Together, they’re actually realizing to what we’re calling the truth. There is a lot of complexity there, but there is a lot of innovation there that is happening …

Borko: It’s not as simple as just taking AI out of the box and plug it in. I’ve spoken with your Chief Technology Officer, Elon, he’s a bit of a mad scientist. He’s tinkering around with all that stuff. What are the lessons then for the travel industry? If someone wants to launch a generative AI-powered tool, can they just do it? How did you do it?

Cohen: Firstly, you need to have the team. When I was saying that we started AI or machine learning in 2015, I meant it. Today we have a pretty big team that can do a lot of different things around AI, and it’s not easy to have that team. I think that sometimes people are confusing IT with engineering.

The fact that you have an IT team that can integrate some stuff or can develop some stuff does not mean that you can hire an AI engineer. That’s actually probably… I can tell you something, it’s something interesting. The AI engineers in our company, some of them are making way more money than what I’m making. I can tell you how expensive it is.

Borko: We talk about a labor crisis in hotels or airlines, but hiring machine learning or AI engineers is also a bit of a labor crisis.

Cohen: It is really, really, really, really hard. They will not come to work for you just for the money. They will need to see that what you are doing is cutting edge. It’s super interesting and there has to be research as part of it. If you will not offer that, there is no way that you’re going to hire the people that will actually build the stuff that reaches-

Borko: Hiring AI engineers becomes a cultural problem for a company then?

Cohen: I’m fairly old. I’ve been around in ’99 in a startup and it reminds me of the same thing. Worldwide web was introduced and a lot of ideas came, and a lot of them were probably right back then, but there was this hype and then it went down. It’s the environment that we see right now, which all of us are obviously using the internet. I think it’s the same thing, but bigger and faster, which means that definitely you’d have resources shortage. That’s what we see.

Borko: How do you think about the human piece of it? I think you’re talking about people, they want to talk to travel agents, they have complex problems. It’s a complex industry. You have a lot of partners and things that can go wrong. I mean, are you worried that you’re taking too much of the human element out of it?

Cohen: I think that any kind of technology or revolution, and that’s definitely a big one, change the work environment. This one, for sure, will change the work environment. I have no doubt about it. It’s scary, because a lot of jobs will not be relevant, but I think a lot of jobs will be created. First of all, what will happen in the future actually? I don’t know and I think nobody knows, but is that the big enough of a change to change what people do? I think that the answer is yes. I’ll take it to an area we’ve just talked about, the engineers. Engineering today, where you have a good engineering team, is not done today in the same way that it was done a year ago. I’ll give an example.

I’ve talked about this engine that we’ve developed to go between models and APIs and so on. Let’s say that I want to add… You asked me what’s the weather in New York, and let’s say that I don’t have any functionality that we’ve developed about the weather in New York. Our model will tell us, “Go to this API. This is the command that you need to implement. Implement it, run it.” That’s engineering that happened on the fly in real time without an engineer to actually give you an answer for a question that we didn’t anticipate. That’s changes engineering.

Borko: Let me ask you, you’re in a tough space. You’re in business travel. Let’s bring it back to business travel. You’ve got a lot of incumbents, a lot of competitors. Are they going to be able to do this kind of stuff or you think this is a real differentiator for you?

Cohen: I think the startups will be able to do it. There is a lot of startups across. I think some of the players in the consumer market. I know that you had Glenn Fogel in the morning.

Borko: Yeah. Glenn Fogel.

Cohen: I’m certain that Booking can do that. I think that it comes to engineering capabilities. I think that the likelihood that the traditional corporate travel players will be able to do it, it’s probably zero. No, I mean it. I think it’s zero. Will they integrate to maybe solutions that others will build? Probably yes, but it’ll never give them the edge because these solutions will be available to everybody. You cannot take a non-tech company and make it a tech company overnight because you’re afraid of dying. That’s an issue. That’s what will happen. I think that that’s something that will probably accelerate the change into technology and business travel.

Borko: Technology and business travel. You really consider yourself a tech company first, right?

Cohen: Yes.

Borko: All right. I can’t wait to see what technology and business travel are going to bring us in the future. Thank you, Ariel.

Cohen: Thank you. Thank you.

Borko: Thank you.