Thomas Gregorson: Good morning, all. Good to see you all. It's great to be here. We're going to share with you some exclusive research we did on AI. We've been hearing a lot about AI over the last couple of days, and it's really an amazing event. I want to calibrate first with the audience. Who here uses AI in your personal, daily life? Is anyone not raising their hand? I don't think so. Okay, now the second question. Who's used AI to help book or find travel? A little bit less. I do use AI to help book travel, and it's really the fundamental most important question that I have been trying to share for the last 25 years.

The question I asked AI is, "Why travel to Fargo?" People that know me know that I've always been trying to preach why we should travel to Fargo, and AI gets me. I love the answer. It said, "Traveling to Fargo may not be the most obvious choice," but there's a kind of a point to that. It's one of those places that surprises people once you actually go. And it talks all about the things you can see and do in Fargo. So you won't be surprised about this AI evolution and AI in travel and how it's going to be.

We're going to talk a little bit about what's happening in it. You can't go through a day where you don't see a LinkedIn post or some kind of announcement that on the travel search side, AI is a topic that's evolving incredibly fast. That's because the technology is developing. It's not changing just how people search, but it's actually changing who's doing the searching. It's going from a person instituting a search to an agent actually performing the search. And the most important piece of it is not only that, it's how the decisions are going to be made. That's going to disrupt or change the way that we actually have to think about flight shopping in the future.

We're already seeing this shift, and I'm going to try to put a little context around it. We've talked a lot over the last couple of days. AI is answering questions, but now moving to taking actions. In the typical travel search, you go in and put in your flights, your dates, your destination, all of that. But with agentic AI search, it's going to be different. It's going to interpret your intent. Why do I want to go there? It's going to search across multiple channels trying to find out what and how to satisfy that intent. It's going to compare and optimize the different options you'll have as a consumer, and it's going to retry if needed. Then over time, it's going to initiate the booking. Basically, it's not going to be a human-driven activity. AI is going to propose opportunities for you in order to find things, and it's actually going to reinvent how the top of the funnel, or how the search, will happen.

So what happens when AI, not people, is shopping for flights? What we found is three things out of this research. One, it's real enough to matter. It's happening now. It's not going to go away. You can't put a board over agentic AI and say let's stop it. It's going to happen, it's going to be there, and it's real enough and happening already. Two, it's early enough to help shape. We're all trying to figure out how to make this work as a new way to sell and inspire travel, and it's early enough to shape. It's not set in stone yet, so we can get involved. And three, it's risky enough if we get it wrong. There are risks and challenges with agentic AI that if we get wrong, we're going to have some serious problems. Kristine is going to go through the exclusive research we did to demystify what agentic AI is, what the status of it is, and then how we move it forward. Kristine.

Kristine Jenkins: Thank you, Tom. Ultimately, the shift towards agentic AI and flight shopping will be driven by consumers, by flight shoppers. That's the way they want to search for travel. We saw when Tom did the show of hands that even in this room we're already seeing demand signals, and research from PWC shows that 68 percent of flight shoppers expect to use AI to compare flights, and 57 percent expect to use AI to book travel. I would imagine, based on the pace of change, those numbers have already probably gone up since this survey was done.

So use of AI introduces a new layer in the shopping journey, with AI agents acting on behalf of travelers, and that raises some structural questions. Who ensures that airline offers are represented correctly? Who absorbs the cost of automated shopping, and how does it change existing commercial models? Who owns the customer relationship, and who governs the system? Does anyone? Ultimately, this will create structural shifts in airline distribution.

A big challenge in having conversations about AI today and understanding the space is clarity. Agentic AI as a term is used loosely. Many of the solutions that are in the market today are actually conversational search, assisted planning, or workflow automation, and these definitely create improvements in customer experience and increased efficiency, but they aren't really the structural change I was mentioning. True agentic AI involves autonomous multi-step decision-making, dynamic adaptation, and minimal human intervention. While most are not there yet, pilots are increasing across the industry, announcements are accelerating just like Tom showed earlier, and experimentation is widespread. So it's coming, and faster than we could have believed just a few years ago, and honestly, just a few months ago. The real question we are looking at here is not can AI shop for flights. It's can it do so in a way that the industry can trust and that can scale.

So with all of this in mind, we set out to understand industry status, risks, and opportunities when it comes to agentic AI and flight shopping. We conducted research across four dimensions. First, airlines. We talked to 18 airlines globally. Next, we looked at systems, channels, and ecosystem partners, which included some discussions and market research as well. We assessed the standards environment, and then we also looked at technical maturity and capabilities. The primary research was done in December and January, but we've continued to conduct ongoing secondary market research and technical research since then.

Of the 18 airlines that we talked to, 67 percent expect major industry impact. Of those airlines, 39 percent confirmed that they have active internal AI programs, but only 11 confirmed that they were actively working with external AI agents for shopping. These findings are consistent with other research we've been seeing, like Bain & Company's independent research, which showed that airlines are actively exploring AI but remain cautious in committing to agentic use in flight shopping. The key takeaway from our conversations is that airlines see the opportunity but do not fully trust or understand the full operating model in terms of how it's all going to work yet.

Looking at segment trends within the airlines, unsurprisingly the large carriers were leading experimentation. The mid-sized carriers had some selective exploration, but it wasn't super widespread. And the smaller regional carriers were mostly observing at this point and waiting to see how the industry evolves.

We also saw three themes that were consistent across everyone we talked to. First, as Tom mentioned, agentic AI and flight shopping is inevitable. There wasn't anyone who thought the cat could go back in the bag, so to say. Second, the distribution impact is uncertain. There are a lot of questions and concerns about the impacts on the ecosystem, which I'll go into more in a moment. And then lastly, industry frameworks to ensure scalability and trusted results don't exist yet.

Next, looking at systems and channels, our discussions confirmed that ecosystem players are definitely moving faster. We are already seeing use cases like embedding conversational AI, testing autonomous trip planning, and building agent frameworks and protocols behind the scenes to start building capability within the industry. But overall, these approaches are inconsistent and the standards are not yet aligned. Innovation is moving faster than governance.

Agentic AI in shopping represents a real opportunity for the industry, and I think we've heard this from our panelists throughout. If done right, it helps enable smarter personalization and higher conversion, really improving the customer experience. It can help with servicing and rebooking, both from an efficiency perspective and from the customer perspective. The ones that do it right will see improved loyalty and customer experience through their implementation. And then finally, operational efficiency through the automation that's enabled by agentic AI.

But at the same time, there are risks. Skyscanner talked about the look-to-book pressure that could come via agent-driven shopping. The retry loops could put a lot of load on systems, and that's a very real concern many airlines have, which we've heard reiterated throughout the sessions. Another risk is data misinterpretation and inconsistent offers. Do agents understand how to interpret airline data? Do they know where to go to get authoritative sources? We also have concerns with security and trust in agent behavior. Are agents acting in a way that's going to get the right results? Are there cybersecurity concerns based on the new technology? We also have regulatory exposure from opaque decisions. If it's not clear how the black box came to a decision, is that going to open up liability? And then finally, value chain disruption, which has been discussed a lot. Who owns the customer? Who pays? Is there going to be channel shift related to all of this? That came up a lot in our conversations. The bottom line here is there is really high upside, but there is also high structural risk and uncertainty.

Standards are one way to start to address those risks. Standards and common protocols for agentic commerce are emerging, but not yet specifically for aviation. Some examples include MCP from Anthropic, which is a protocol for AI models to interface with external tools and context. Moving to commerce specifically, there's UCP and AP2 from Google. OpenAI also has ACP, which starts to enable AI to shop and pay on a user's behalf in a standardized way. And then finally, agent frameworks such as A2A for agent-to-agent and agent skills, which are standards for managing agent behavior and interaction. All of these are associated with tech companies. They are cross-industry and not airline specific.

So what's missing? As the people in this room know better than most, aviation is complicated. When you think about airline shopping requirements, ensuring consistency, accuracy, and trust in the process, making sure all the data is coming from the right places to produce the correct offers, when we think about servicing and the order life cycle across a whole bunch of processes and partners and ecosystem players, and then finally governance and compliance and all the regulatory requirements on the industry, right now those standards have not been created by anyone with specific knowledge in those areas. Standards are being developed, but not by the industry. And the implication is that without industry input, these external standards might define airline retailing. The industry needs to think about whether that's acceptable or whether to take a more proactive role at this point.

Thomas Gregorson: Okay.

Kristine: Onto the review of technology and capabilities. We conducted discussions with some of our internal technical folks and identified five foundational capabilities that we believe are required for agentic commerce to work at scale for airlines. The first is authoritative machine-readable data, which would be consistent interpretations of fares, rules, products, and merchandising data for agents to use, so they're all using a source of data that has been validated and is known to be correct. The second is standardized governed access, a way for agents to plug in to reduce scraping and uncontrolled access, which could start to potentially address some of the look-to-book concerns I mentioned earlier. The next is delegated identity and consent frameworks, making sure that agents are able to act within defined authority. This could be on behalf of a flight shopper, ensuring the agent has the authority to go through the booking process, and it could also cover agent-to-agent use cases where an agent is acting on behalf of another ecosystem partner. The fourth is durable transaction auditability, making sure there's traceability across the entire booking life cycle so that all the industry processes work and can be audited all the way through. And then finally, shared conformance standards, making sure there's interoperability and validation across all the systems. None of these things are really new, but applying them to AI is critical. Without them, the upside I mentioned earlier will become harder to achieve, and those risks begin to look a lot bigger. Now back to Tom.

Thomas Gregorson: So what does this mean for us in the future? I'll share a little bit more on the Fargo story. I was amazed that the next question the AI asked me was, "Would you like me to plan a weekend trip?" And it created a detailed agenda of where I should go. I'm a baseball fan, and I mentioned I like baseball. It pointed out the Roger Maris Museum in Fargo. It told me what restaurants I should go to, it suggested the hotel I should stay at, and it told me the flights I should take. I don't know how it came up with these, I don't know how it picked those, I don't know how it proposed it, but it was proposing to me a very good and detailed itinerary. And I think that's kind of the situation we're going to be in. Is it going to be a black box where we just say this is how my products and services are going to be surfaced to the customer, or are we going to get engaged?

We see that there are really three ways this thing will evolve. One, it could be platform dominators. You get the tech providers, the agentic AI tech providers. They will come up with the frameworks, they will come up with standards, and they will come up with methodologies in order to surface this content, and they will have an appetite to go get content anywhere they want. We believe that particular framework is going to create a black box where you will not be able to know how customers are going to be proposed solutions or not, and you will not be able to propose your best offer and put your highlights there.

The other one is fragmented adoption, and we see that right now. Certain players are starting to interact with these technologies and asking, "How do I actually expose my content to that?" As Kristine mentioned, it's more on the channel side that are getting involved, doing their plugins in order to do this, but it's going to create a fragmented experience. You're not going to have equal access to all content, it's not going to be comparable and describable, and it will create friction in the overall journey.

We believe the best way is an aligned evolution. Let's share the wisdom of all of us that are trying things in our experimentation: what works, what doesn't work, what are the capabilities we need to address, and then how do we best address them? We think that is the only way the industry will really get a good benefit from taking advantage of the new technology in order to create the next level of shopping.

So to be perfectly clear, ATPCO's mission has not changed. We are here to help the airlines make sure they get the right offer to the right customer at the right time. Agentic AI is just a new way to surface and enable that. ATPCO is not building agents. We will not be building an agentic AI agent in order to do that. But we're here to ensure the agents work correctly in that mission of getting the right offer to the right passenger at the right time in this new world, in this new era.

Kristine highlighted some of the opportunities we have as an industry in order to do this. We are going to look at these things and have ongoing discussion, dialogue, and experimentation to see how we unlock some of these industry-needed capabilities. First and foremost, it's really data standards and semantic clarity. How do you make sure that it is a fact-based evaluation that the agentic AI is bringing into its decision set, so that what's proposed to the consumer is really what is actually going to be able to be delivered and serviced? We believe that there are going to be some shopping frameworks that will have to be evolved. Not rigid standards, but frameworks to make sure that the connectivity and flow for the agentic AI workflow is able to be scaled and able to be plugged in for multiple channels and multiple airlines all at one time. We believe there's going to be something on governed connectivity for agents. Just like there is today, there will probably be good, high-value agents and ones that are not so good and actually diluting the effectiveness of the offer. There will be some kind of governance that will occur over time around when you turn agents on and off in order to make that happen.

We see that agent-ready offer data is going to be needed. This is going to change the paradigm. I don't see a world where we're going to be able to shop and do calculations the way we do today for inspiration. Thinking about and inspiring a traveler on where to go and what the best offer and content is, that's going to have to bring the data to a new way of being able to be accessed and perform that top-of-funnel search. There's going to be agent-to-agent interoperability, and these external agentic AIs will talk to other agents. How to make sure that they're talking efficiently will have to evolve. And for ATPCO specifically, we are aggressively looking at how we can apply AI in our applications to make your ability to get content or create offers much more effective in the marketplace.

ATPCO's role is to ensure that we provide shared meaning, so that people will have a clear understanding of what the products and services are and what the airlines have to offer, and that it has a shared meaning and understanding across all the agentic AI utilization of it. A shared infrastructure. We don't see it as valuable for each one of us to build these capabilities over and over again. We believe we can be a shared infrastructure that can help unlock that capability for the whole in order to make that happen. And we also believe in a shared governance, where we'll share the learnings with all and learn from you as well, and continue to evolve together in how to automate and create a great agentic AI experience.

Kristine covered a lot of the highlights from this research, but we're happy to share the research with you as well. Agentic AI will reshape flight shopping. The question is not whether, but how. Will we evolve it through framework workarounds? Will we evolve it through opaque external layers, or are we going to do it from aligned, trusted infrastructure? We believe the choice is still open, but not for long. Unless we align together and work together, we will find that we're going to have a fragmented and less perfect experience.

The research we did is just a starting point. It's a white paper that we have available for you. It has a lot more detail on the technical capabilities, the standards, and what we have done in the secondary research. Take that as your resource. Use it at home. Share it with your colleagues and say, "How do we get involved?" We would also like to make it a living document, so we want to share and understand your learnings and your experiences, and we'll continue to update and incorporate that and feed it out.

We're actively looking at how to address some of these capabilities. We have experiments going on right now at ATPCO around how we create a semantic layer on our data and how we make sure that the data is fit for purpose for agentic AI to be reading. We will share the outcomes of our findings, what worked and what didn't, with you, so you can help prepare and get your data ready as well. We're also inviting you to get involved. What I'm asking each and every one of you is this: if you are not using agentic AI for travel search yourself, do it. I think you'll open up your eyes to what the capabilities are and what the deficiencies are. Use it in your everyday life and you'll get much more experience with it.

Just like when the internet came along in the '90s and we said, "This is the way it's going to shape travel and shopping," this is also going to shape travel and shopping. Get engaged. We're going to have a design team on May 20th where we're going to start to bring and share some of our findings on agentic AI. Sit in that design team with us, we'll share the outcomes with you, get your feedback and understanding, and then also help us, challenge us, and provide your input so we can shape the future together. We encourage you to explore the findings of the white paper. We encourage you to have dialogue with us on this. Let us know how we can help shape the way for airlines to get the right offer to the right customer at the right time through the agentic AI experience. With that, I want to thank you for your time. Thank you to Kristine for all the great research. And download the white paper and please get involved.

‍