Building Trust in AI Powered Business Travel With Amex GBT

[00:00:03] What happens when AI can rebook your cancelled flight, prepare an expense claim and even predict disruption before your journey even begins? Well, anyone who has raced through an airport, wrestled with a travel policy or discovered a forgotten receipt at the bottom of a bag, you will know that business travel contains plenty of avoidable friction.

[00:00:29] Well, today I'm joined by the Chief Product and Strategy Officer at Amex GBT. I want to learn more about research from Ipsos found that 4 in 10 Gen Z business travelers consider arranging work trips just being too difficult. So I want to discuss how AI can maybe simplify searching, booking, disruption support and expenses

[00:00:55] while also preserving the traveler choice and access to experienced human advisors when we need them the most. So my guest will explain why trust will determine how much authority people will give an AI agent and where corporate travel has an advantage over consumer travel and why early agentic booking experiences, if done incorrectly, can take longer than the more familiar tools.

[00:01:21] So if you're a business leader assessing AI, this conversation will offer a very practical reminder that saving time only matters when the technology also earns confidence. But enough from me. Let me introduce you to my guest right now. So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do?

[00:01:45] Sure. My name is Steve Conner. I am Chief Product and Strategy Officer here at American Express Global Business Travel. And we are one of the leading B2B travel companies in the world. We prepare, manage and help our customers and their travelers travel the world for business. Well, there's so much I want to talk with you about today, especially the changing habits in travel too.

[00:02:10] So, I mean, for example, four in 10 Gen Z travelers have said arranging work trips is just too difficult. So I've got to ask, from everything you're seeing and hearing here, which parts of the experience are creating the most friction? And where can maybe AI make an immediate difference rather than just simply adding another interface on top? Yeah, well, searching for travel has definitely gotten more complicated.

[00:02:35] And that's because our airline and hotel partners have wanted to really merchandise and retail a different kind of experience. So it used to be you wanted a flight from New York to Chicago and you pick the best schedule and you pick the flight. And now you have not only a choice of schedule and airlines, of course, you also have a choice of different fare cabins and multiple cabins in that plane and multiple fare types in that plane and multiple seats. It's for each of those fare types and different amenities and what we call inclusions, which are things that come with that.

[00:03:02] And then each of those, you have to think about what is in my policy? What am I allowed to buy in the corporate environment? You may think about what's important to you. You may think about what loyalty you have to an airline and benefits you get, what credit card you have and the benefits that you get with that credit card. Right. And it goes on and on and on. And so the complexity of buying a ticket today is actually way higher than it was 10 years ago, 20 years ago. And so absolutely, I can sense that, you know, Gen Z and many others, frankly, are seeing that when they search for travel.

[00:03:31] So our job is to make sense of all that for travelers, for their companies, help set the right policy and the right guardrails, and then help travelers get on with their day as quickly as possible, because we would like travelers to focus on the business trip and the experience and what they need to be successful and less time focusing on logistics. That's our job. And I'm curious, if I was to take you away from your job, just on a personal trip somewhere, are you finding you're using AI more? Because I know I do.

[00:03:58] If I could be landing at an airport somewhere, it will be AI that I'll say, what's the quickest way of getting to my hotel from here rather than a Google search? Are your habits changing as a user too? I mean, I play with AI a ton, right? But part of my job is to play with AI. So I play with it on the go. I also play with it, you know, just from my office for fun to see what it gets. And, you know, I think I'm experiencing, I think, what a lot of people are experiencing, which is, you know, part of the time, it's just unbelievable how amazing, you know,

[00:04:27] the results are and how fast and how accurate and how, you know, how much value they deliver for your experience. And I love it. And then some percentage of the time, it's just a complete failure. Yeah. And the problem in our business, you know, in the business of travel, we don't believe in complete failure. Like, that's just not something that we can do. We cannot put somebody on a train that doesn't stop at the station to get off on. We cannot send them to a location that's going to be closed.

[00:04:55] We cannot send put them on a roadway that is closed for construction. And so we have to make sure in our environment that we drive trust. Trust in AI, that's the ballgame in my view. And so in the B2B environment, our view is to take AI, take the capabilities and the data that's available to us in our private store, but also the public store, and curate it and make it available timely, proactive, seamless, right, to the traveler so that it helps them

[00:05:23] on the journey in a way that establishes trust. And then as you establish trust, you get the opportunity to do more and more and more. But if you mess up one time in a material way for a traveler, you're toast. Everybody knows the person who had a bad experience and they swear off that airline forever. You know, maybe they go back if it becomes the only nonstop later down the road, but they swear that off. And us as a travel agency is providing travel service software and services. We have to make sure that we generate trust in AI. So that's really our focus.

[00:05:52] It's such a great point you've highlighted there. And I'm glad we've covered that. And this year, I think I've been to 20 different tech conferences around the world, a lot in the US. And of course, this year, all anyone's talking about is agentic AI and how that can work in industries and transform experiences. And agentic AI can also move beyond suggesting an itinerary to completing part of the booking process. So what can agentic AI do on a traveler's half today?

[00:06:21] And what actions still require clear consent or human approval? Tell me a little bit more about this. That's a great question. I love that question because we obsess over this all the time. So a couple of things. So one, right now, travel, because of all the reasons I just said, it's a human in the loop process. And that's not a technology limitation. That's a personal choice limitation.

[00:06:45] Very few travelers today are wanting to outsource complete authority over their travel to an agentic capability. And then part of that is trust. To be perfectly honest with you, I'm not sure for some people if the trust will ever exist. I mean, I'm a travel geek. I want to know exactly what seat I'm in. I want to know exactly what's happening on that plan. I want to know exactly what my options are. So, but part of it is I love the thrill of that process, right? And there are probably people like me who want that.

[00:07:14] And then there are others who want to sort of just make it happen. And so there are others that will be, you know, maybe the first to go in that. But we're not there yet, right? It's just too important. It's not a digital activity. It's a digital activity that leads to an in-person physical world activity. And so the stakes are much, much higher. And so right now, agentic is, we're working on agentic capabilities and we've launched some that help you with that booking experience. But they're going to ask you to say, this is what I want. Confirm that booking.

[00:07:42] Now, in the business travel world, we can then use agentic workflows to take care of the rest for you once you've confirmed that's what you want. Because we have what I call trusted transaction authority. We know who you are, your profile. We know your company and the policy. We know the payment vehicle you want to pay for. We know that you're allowed to make a trip that your company is going to pay for. And so we can actually have agentic workflows that do that for you because we have all that established. In the leisure space, you have none of that, right?

[00:08:09] So in the corporate space, actually, B2B, we have an advantage in agentic use cases and we're going to move a bit faster. Now, the next step is to say what you just said, which is, okay, that's human in the loop. What can we do without human in the loop? So I think the first use case is going to be, and we're starting to play with disruption management. So when your flight's canceled, you're a bit more willing to let an agentic agent do something for you if it helps you out, as long as it could be unwound, right? You don't mind if they go grab that flight and book it for you.

[00:08:37] And then if you can unwind it, if you don't like it or change it later, maybe that's fine. But you'd rather have that new seat than not have that new seat. So I think that's where we're going to see the first use cases. We're still human in the loop today. But that is where we're going to push the limits on use cases where time is of the essence. I think as we go, there will be some trips, types of trips that we can have agentic booking. So today we have companies that are planning big events and they want to move a thousand people to Las Vegas for an event.

[00:09:05] And it's an incentive event, let's say. So everybody's having fun on the trip and they're comfortable telling people what flight they're on. And companies, absolutely, some companies in some circumstances are comfortable doing that. And so in that case, today we have workflows to book those trips for those colleagues. And now we're putting in agentic workflows to book those trips for those colleagues, which will make it faster, more seamless, fewer humans in the loop. But again, that's sort of an evolution of a workflow that already exists in the way. It's not really a changed user experience.

[00:09:35] So I think there's opportunities to dabble into this. I think corporate will lead is my view. But I think we're just at the beginning. And a delayed flight or worse, still missing your flight is the stuff of nightmares. Everyone's worst nightmare there. So the problem is, though, that disruption is the moment when any travel platform will face its toughest test. And so how can AI further help when that flight is cancelled or that connection is missed?

[00:10:02] And when should the traveler be transferred immediately to an experienced human advisor that can sense that there's some frustration or anger going on there? So that's the other part of the magic, right? Is when you just give up on the agentic workflow and send it to a human. I mean, we have an amazing cohort of colleagues that are professionals in travel that work with our customers every day on the phone or through chat to make sure that they get on their way. So agentic capabilities build really good workflows really fast.

[00:10:32] So a flight's cancelled or delayed, you get a notification. We're also working on saying we can identify risk spots where we think there's a higher likelihood that a flight might get delayed or cancelled. And so we can queue up different alternatives even before something happens and either be what I call like fast reactive, like ready to act. Or in some cases, you know, if we see a storm coming, we can say, listen, would you rather just change your flight to another time, another day, another airport, another connecting city? So you don't avoid it. So you avoid any risk at all of what's happening over here.

[00:11:01] And so we can do that too. So AI is great to scale those use cases on helping people avoid what might be an issue and helping people fast react to an issue. And then we have to, but we have to have it every turn. We have to have an easy opportunity to get to a human. And so in our agentic workflows for disruption, for example, that we're launching, you can simply click a button or ask for a human, a travel counselor, an agent, however you want to do it. And we'll send you over to one of our humans who can help you.

[00:11:30] And they'll get the context of that conversation, the agentic, the AI conversation to help go fast. So you can either go straight to a human or you can see what we can provide for you in a digital self-service fashion. That might be quicker. If there's something right there and you want to grab it, go ahead and grab it, right? And then if you're not satisfied with what you see or you have concerns, you can go talk to a human. And so putting a human service professional in the loop is absolutely critical. I mean, our business is helping people travel around the world, as I mentioned.

[00:11:59] And humans are going to be part of that service experience. I think forever, either by necessity or frankly, for some people by choice. And that's perfectly fine. And one of the things that surprised me in your research was expense management appears to be affecting participation in business travel with some Gen Z employees, even avoiding trips or abandoning legitimate claims. How can AI and integration such as Concur expense, how can that help reduce the burden, but without

[00:12:28] weakening accuracy, policy compliance or employee control? I would imagine it's quite a balancing act. Well, I think it's a great question. I think AI is the answer to the expense. Like, I think the notion of expensing something is going to go away. Like, it won't be in our vocabulary in 10 years. Like, it'll just be gone. We won't have to do it because the workflows behind the scenes will take care of it.

[00:12:54] And AI will create the way to do it that reduces risk and fraud for the business. And that's the issue, right? The friction of expenses, you go through a manual process to reduce risk and fraud for the business. And so there's always a seesaw, right, of you can make it easier for people, but there might be a higher risk of fraud or you have tax implications, or you can make it harder on people and you can cross every T and dot every I. And most businesses try and find that balance, right, of how they want to do that. AI means you don't have to trade off.

[00:13:23] We can make it easy for the user because we can automate everything. And then we can only send things to people for manual workflows when there is a high risk, when there is something that pops out as an exception. And we can use all the data in our repository and Concur is using all the data in the repository together to make sure that then those exceptions, those high risk moments are really well informed because they have so much data. So they know when there's an outlier. They can tell you when something is likely a red flag that needs to be addressed.

[00:13:51] And so using all that data, they can create workflows that really make the manual labor of an expense report reduced to de minimis. And as we take all receipt data, trip data, and we flow it through to the systems, we can eliminate the need for expense reports. And we're doing that already in some cool cases. So, for example, if you're using an integration, if you're on complete our joint solution with SAP

[00:14:18] Concur, or you've integrated our agency solution with SAP Concur, the receipts flow through into Concur, pre-populate as an expense, and get flown through. Now, you still have to go in and click submit a report. Okay, that's not that laborious, right? Just that encapsulates that as part of the expense workflow. But you don't have to populate all that information. You don't have to give reason codes. You don't have to give any information. It all just goes through.

[00:14:44] And another example for non-profile travel or guest travel, which is a huge issue, somebody who's a recruit or somebody who's a contractor or an invitee to a meeting, we want to make sure that they can pay for things on their trip and not have to file an expense report because it's really hard when you're not an employee to file your expenses. Sometimes it takes weeks and months and tons of paperwork. And the companies historically have had no incentive to make that easy. They're not their employees. They're edge cases, right? But some of these are really important people.

[00:15:11] Recruits, special guests, contractors are really important constituents. And so now we have workflows with virtual card systems where you can get the virtual card on your phone deployed, all deployed by the person managing the event. You get everything paid for with virtual cards, your hotel, your air, potentially your ground through our system. And then as long as you're spending in normative ways, you're buying under a certain budget at Starbucks for your coffee. You're using regular taxi companies. It's a hotel expense, air expense.

[00:15:40] You know, the stuff that you would expect somebody to spend on travel. It all goes through. It gets audited, you know, automatically and it's done. You don't have to do anything. It all gets paid on the card. And they're only going to call you if, you know, something looks strange. But we have controls at the point of sale. So if it's strange, we won't let you do it, right? So if you try and go into Best Buy and purchase a TV, it's going to reject it at the point of sale. It's going to say, you're not buying a TV. And actually, if you go into Best Buy even, and you try and buy even a candy bar at Best

[00:16:08] Buy, right, which you think otherwise might be a covered expense, it may actually also reject you because it's saying, listen, this guest traveler should not be buying stuff at Best Buy, right? You know, so on the edge. But what that does is when you put that control in there, it means when you go to Starbucks and they receive and it goes through, you're done. You don't have to. Why do you have to write? I got a coffee. You spend four bucks at Starbucks. You got a coffee. It's totally cool, right? And so how do we make these processes seamless and smooth for travelers any circumstance?

[00:16:38] And it's really exciting what we can do now. Your agentic AI might not be secure even with real-time data and proper guardrails. But Denodo makes sure your business has every avenue covered. By placing all your data platforms under one AI data layer, your business can't be can reach semantic consistency safely and securely. So get your agents on the same page by visiting denodo.com.

[00:17:08] And you can learn more about how to start trusting your agents to make business decisions. But now, back to today's guest. Love it. And I was also reading before you joined me today that agents here is bringing disruption support into tools such as Microsoft Teams, which I think is a great move too. But for people listening, hearing about that for the first time, what is gained by placing travel assistants inside their existing workflows?

[00:17:34] And how do you, again, stop convenience from creating maybe fragmented records or uncertainty over where the responsibility sits? Yeah. So your comment to open this up, which I think, you know, based on some of the research that we've done, you know, yeah, 24% of travelers say their experience is seamless and low effort. I mean, come on, we want to we want to push that way up. So how do you make things seamless and low effort? That's across all generations, right? And as you pointed out, it might the friction is even higher among some of the younger generations.

[00:18:03] The way that we do that is, is we use technology to make that experience easier, as I mentioned before. But then we put that experience where you want it. And when we put it where you want it, we do two things for you. One, you get to take care of your business travel in the same workflows that you take care of lots of other business. That makes life easier. So if you're in teams talking to your team all day, dealing with things, dealing with other admin issues, you know, within your work life, and now you can take care of your trip in teams, you know, your life is easier. You don't have to go into another tool. You don't have to find it.

[00:18:33] You have to remember where it is. Some of our travelers are traveling once a year, twice a year. This is, you know, and some are road warriors, any use case. And so we want to be where our travelers want to be. And so it's not that we're forcing you to use teams. If you prefer the full travel booking experience on the web or on mobile or any channel or talking to an agent, that's great. But we want to be where you want to be. The second thing that that unlocks for us, and we also launched our Gencia integration

[00:18:56] into Cloud for B2B, for folks at B2B Cloud, Enterprise Cloud, is that these tools, teams, for example, Microsoft, Anthropic, they have access in an enterprise environment to other data about you and your company that may be useful for booking travel that we don't today. So for example, if you have a Microsoft Copilot and Copilot knows you're all about your calendar and they know all about things going on in your life based on your calendar that is very

[00:19:24] relevant for travel, that's now a more informed system that we can interact with to make that booking works more thoughtful. So we can say, well, listen, we see they have an important meeting at one o'clock. So we know that they have to arrive in the city at 11 a.m. And they can't come in after 11 a.m. So why even provide them search results on a flight that gets in too late to hit their meeting? Now, we can do calendar integration some ways today.

[00:19:50] But it's just an example of how when we integrate with systems that are otherwise useful in your life and in your workflows, we make everything easier for our travelers. And that makes our customers also really happy because it drives compliance in their program. They save more money for their business. They make sure that they know how they're spending their precious dollars and everybody wins. There's an old enterprise IT mantra of you can only improve what you measure.

[00:20:18] And I think that is more important than ever before. So for anybody listening in an organization currently assessing AI supported travel management today, what should they be measuring beyond speed? And how can they determine whether technology is improving traveler well-being, policy compliance, cost control, and all that stuff, and also access to human support? Any tips or advice you would offer around that? Yeah, I mean, it's a great question. And honestly, it's not an easy one to answer because the measurement is at so many different

[00:20:48] planes and there's so many different dimensions. So the first thing I'd say is for any of our customers, we have to measure success at the customer level and at the traveler level. And those are the related constituencies, obviously, but they're different, right? So at the customer level, we want to make sure that a customer is getting value for their program. So we have the most comprehensive, competitive marketplace in travel. And we want to make sure that that marketplace is showcased so that our customers are getting great value, whether they've negotiated content, whether we've negotiated content,

[00:21:17] or whether we're just bringing in all the content possible from every corners of the world to make sure they get it. So we measure marketplace. We measure the value. We measure savings and value that can be delivered to our customers. And that's really important from a customer because at the end of the day, they're procuring travel and we want to make sure they're getting the absolute best value for the travel they've decided to buy. That also means making sure their policy is really good and fit for purpose. And the way they've deployed their service and support is fit for purpose for their business.

[00:21:46] And so from a customer lens, we're definitely going to look at, are you getting good measurement of value? Do you feel like you're getting good value for your program? And is the marketplace delivering everything that you need to be confident that you're getting that value? I think, and increasingly, we're deploying AI into our data analytics environment so that our customers, the travel managers, the procurement teams can feel more confident and

[00:22:13] have more ease to get information about their program. So that's definitely an area where there's going to be a lot of innovation. And I think measurement of the program is clearly really, really important. But when it comes to traveler experience related at separate dimension, ultimately, you have to be surveying travelers and understanding how they feel about the program. So time to book is really, really important. And I don't belittle it. Time to book is actually, I think, underused as a metric. We're very proud of our time to book and our booking tools.

[00:22:43] And I think Agentic, now we're trying to get time down in Agentic as well. Agentic book flows, by the way, today take longer than going into the booking tool, right? Booking tools have been built for years for speed and efficiency and how they deliver information. And users know to use them really well. In an Agentic workflow, especially a chat interface, there's, you know, the latency is higher because we're doing more work. And the workflows are a bit more complicated. So that will change over time. But early implementations of Agentic workflows are a bit slower. But we're under three minutes to book on Agentic, for example.

[00:23:11] That's really valuable to the user because it shows that you can get in and out fast. That's pleasing. And obviously, the customer loves it because your travelers aren't spending lots of time booking travel. And then, of course, you're looking at CSAT. You're looking at what is your overall satisfaction with the booking experience, with the trip experience, and making sure that we take that in. And we look at CSAT across all channels. And we understand how does that perform if you're talking to a human? How does it perform if you're in a digital, you know, dot-com experience? How does it perform if you're in Agentic experience?

[00:23:39] And ultimately, as we view those all as just channels of the same thing, meaning you can interplay between them, right? Even in a single booking experience, it gets even harder to measure. And so that's absolutely something that, you know, we're looking at, which is how do you measure in a multi-channel environment even better so that you know how people want to navigate between those channels? Because, you know, for example, if you're in Teams, you might be talking to an Agentic tool one minute, and then you switch over to a human, and you're in that same interface, and you're going to get the same survey at the end of that piece.

[00:24:07] So we'll obviously use, we use tools to measure that, to measure that experience, to read it and feedback. And then we also, of course, are going to ask the traveler how they feel at the end. And we're also using, you know, on the phone, for example, we're also using new advanced tools AI that listen to phone calls and pull together metadata on those phone calls on whether travelers were happy or not. How quick and efficient was the service? Was the traveler engaged in a way we wanted to be engaged? Did they get the right level of service? AI can do that on scale now. Really powerful. Love it.

[00:24:35] And finally, as a self-confessed travel geek on a personal level, what excites you about the role that you see tech playing in travel from research to booking, itineraries and everything in between? What excites you about where we are now and where we're heading? Well, I'll tell you what excites me is that, you know, we have this, for many years I've worked, you know, I'm an entrepreneur. I've built, you know, travel startups. I work with a lot of travel startups. And now, you know, for the last, you know, almost 12 years now, I've been working at

[00:25:05] Amex GBT driving our digital strategy. And I think, you know, what's interesting, you know, is we've been talking about the same things for so long. My first startup was predicting flight delays, you know, using data and machine learning algorithms, pre-AI, right? Using machine learning algorithms. This is 2009, and it was really hard. And the data was accurate enough to be useful, but not valuable at scale, sort of what I would say it was before.

[00:25:32] Now I see startups doing that, you know, with AI, and it's more actionable, more useful. And so, and I can give you 10 more examples of that, where the things that we've dreamed of over the last, let's say, 10 to 20 years are now things that we can build. And not just we at scale, but startups can build them. I can have a, I have a SWAT team, you know, building them. We just did a hackathon and we got hundreds of projects from our own engineering team. Each of them are building things that could be a whole startup, you know, and they did

[00:26:01] it in 24 hours. The whole, the whole hackathon end to end was 48 hours, right? They're building full products. And it used to be they'd build, you know, in a hackathon, you'd build something really kludgy. Now, with our AI-enabled software development lifecycle, they're building full products in 48 hours, right? Now, I'm not saying they're scaling to the globe right away, but like some of them are actually like, we're now starting to deploy. So, to me, what excites me is not that you have some new ideas, which, and some of those ideas are good.

[00:26:29] But actually, it's the fact that a lot of the old ideas that were really good, but technologically were too hard to deliver at scale in our incredibly complicated industry are now able to start seeing the light of day. And that's, that's amazing. That's really fun. And I think that is a thought-provoking moment to end on. I can't thank you enough for sitting down with me today. Talking about why AI expectations are accelerating for the next generation of business travelers,

[00:26:54] I will add a link to the latest Amex, GBT, and Ipsos research that we have referenced a few times today. But anywhere else you'd like me to point, everyone listening who just wants to find out more information after this conversation? Our website at amexlevelbusinesstravel.com, we have a blog, of course, our LinkedIn. We also post lots of content as well. So, there's always great, great content about all the things that we're doing and broader travel industry news. Awesome. Well, I thought I'd have loved to explore with you today the tangible ways AI is already being

[00:27:24] embedded into managed travel through Agencia's new Agencia AI capabilities. So much to talk about there. I'd love to hear feedback from people that have listened today and their experiences. But thank you for sharing your story today. Really appreciate your time. You be careful when you ask people for their travel stories, but always happy. You will get your mailbox being full, but that's good news. Thank you very much for the time today. I think my guest's point about trust being at the whole ballgame stayed with me after this

[00:27:52] interview because, yes, travel provides very little room for an impressive demo that occasionally sends somebody to the wrong train station. Yes, Agentic AI or AI or other AI solutions can prepare alternatives, but Agentic AI can prepare alternatives before disruption occurs.

[00:28:12] It can complete approved transactions, transfer booking context to a human advisor when you really need it most, and remove so much of that work that is associated with expense reports. And nobody likes doing their expenses. But travelers must understand what has been booked. They need to retain meaningful control and know that help is available when plans suddenly become much more complicated.

[00:28:39] And I think there was another useful lesson today around measurement. Yes, faster bookings matter. But organizations should also examine traveler satisfaction, policy compliance, costs, access to support, and performance across not just digital, but human channels too. So you can learn more about Amex GBT through their website.

[00:29:04] I will also include a link to their LinkedIn profile and the research that we referenced today. And even though my guests warned that asking people for travel stories would result in quickly filling up your inbox, I'm prepared to take that risk. I'm here to serve you every single day of the year. So where would you welcome AI assistants on your next business trip or family trip or solo trip? Whatever it is, tell me what works for you. What doesn't?

[00:29:33] What excites you? What doesn't? Whatever it is. TechTalksNetwork.com You can leave me an audio message over there. And you can also find out where you can meet me on the road at a tech conference near you. But that is it for today. So thank you for listening as always. Speak to you tomorrow. Bye for now. Bye for now.