Technology in focus: How AI is shifting the technology and distribution landscape for aviation and travel

Transcript

Ian Kershaw:Good afternoon, everyone. Welcome to the post-lunch afternoon session, Technology in Focus: How AI is Shifting the Technology and Distribution Landscape in Aviation and Travel. With me today is Sean Sutherland, Head of Product, Air and Hotel at Flight Centre. Next to him is Peter Phillips, the Chief Technology Officer at Cover Genius. Then we have Evert Meyer, General Manager of Strategy and Transformation at Virgin Australia. No, sorry, I said Sean earlier.

Sean Sutherland:Yeah, sorry, I'm just doubling up.

Ian Kershaw:All right, try again. Sean Sutherland, Head of Product Air and Hotel at Flight Centre. And on the far side is Justin Warby, VP of Distribution at Philippine Airlines. So we've got a nice mix of airline, agent, technology. We're going to talk about ancillaries and a few other things today. This is a 2-part session, right? Distribution and AI. It's been merged together. And if you're wondering, we will be speaking about more than just distribution, particularly when it comes to AI and the improvements that are being delivered, promised by that new technology. So we're going to get into it. We're going to start with distribution. And Justin, a heads up, it's coming to you first. The context for this is the global distribution landscape has become much more complex, fragmented. One stat I found is that travelers today have an explosion of choices. So the industry in 2010 went from offering about 500 air travel products and fares. In 2024, that was about 10,000.

Evert Meyer:Right.

Ian Kershaw:This also has a lot of impacts. Cost of distribution, probably headaches for you and others in the industry. If you could kick us off with the trends, the key trends that are significantly reshaping distribution in aviation.

Justin Warby:Yeah, thank you, Ian. So from my perspective, I think the biggest change occurred during the pandemic. For me, when we came out of the pandemic, there was new actors, There was new technology, and it was though people were sitting around at home during the pandemic thinking up evil plans about what they could do. Say, let's extract value out of the distribution land space. So whilst everyone knows about NDC and what that can entail, I think there's been a large increase in terms of redistributors of content, particularly in the B2B space. And for me, that's the biggest change that I've experienced and the team at Philippine Airlines experienced. So the space now, like you mentioned, has a lot more content. It has lots more things you can sell. But it has a lot more players in the space. So the technology enables low entry barriers for these players to come in. And I'll give you some specifics that keeps me awake at night. So we have a number of content redistributors, particularly they redistribute their content to OTAs. Not saying anything bad's wrong with that, but it's where they get their content from which is the problem. So a lot of these redistributors of content get their content by scraping content from airline.com, so in this case from our own website. So we spend a lot of time either inhibiting or trying to limit these actors who come and sometimes do nefarious things on our website.

Ian Kershaw:So what happens when they do those nefarious things?

Justin Warby:So one of the nefarious practices that we see at the moment is a practice called seat spinning. I'm not sure if people in the audience are aware of what seat spinning is, but they'll particularly look for flights that are near full and might only have 2, 3, 4 seats left. They'll then use your own ticketing time limits or or they'll use your own payment methods such as book and hold that are close into departure, and they'll actually come and hold that inventory. And then they'll resell that content to a third-party provider to resell, and that resell can be at prices 50%, double, 3 times the normal price. We don't hold the inventory anymore because it's being held by this third party, being sold by another party, And as soon as the ticketing time limit or some mechanism kicks in and it gets released, they'll buy it again. So this will constantly be happening in a couple of days before departure where the flight is nearly full.

Ian Kershaw:So that's a revenue impact.

Justin Warby:That's a huge revenue impact because we're not selling the seat, right? Because it's being held at that particular time. It might spoil—

Ian Kershaw:It's also a reputation impact.

Justin Warby:Sorry?

Ian Kershaw:It's also a reputation and brand impact.

Justin Warby:Reputational brand. Consumer looks at us and goes, well, Philippine Airlines, you're way too expensive. I'm not going to buy from you again. So yeah, we're spending a lot of time trying to prevent. We've actually gone out and launched some bot defensive tools again this year. One, I'm not going to give company secrets away, but there's one tool that we launched recently that looks at the computing power of the session that's coming into your website. And based on the computing power, we will then interrupt that session, or actually probably more likely pause that session, which then creates the bot to create another session. And effectively, we just overwhelm the computing power of the bot.

Ian Kershaw:I think this is almost one of the themes for the session. We'll probably get into it a little bit more later, but almost the kind of battles that are being waged. And we would never see it, right? We just go to a booking website. We go to Rakto. We go to Booking.com. We go through Flight Centre. We have no idea this is going on in the background. I think we're going to come back to it in probably 15 minutes or so.

Sean Sutherland:Sean? Yes.

Ian Kershaw:From the agent perspective, kind of a similar question, challenges and opportunities presented by channel proliferation and fragmentation.

Sean Sutherland:Yeah, so probably the most obvious of the lot, the proliferation of NDC and the non-standard standard that we find ourselves in. So Obviously, from a— the challenges we face is obviously extra overhead for our agencies or even online to find the right product. It's hard to differentiate between products because, you know, there's not a— it's not standardised. So we have to, at least from our agent standpoint, try to standardise it for the customer so they understand the differences between these 2 products, whereas before it was very simple.

Justin Warby:Yeah.

Sean Sutherland:Then we obviously have the hardest piece, which is the servicing, which is a dog's breakfast depending on who you're dealing with, and it is all different levels. So even, and then from the back of that is how much we have to train our staff for the different airlines and the different products and offers that are there. That obviously is very disruptive in our group. However, I will say that whenever there's disruption, there's opportunity. So what does that mean? We obviously acquired TP Connect out of Dubai to bring us closer to NDC and so So that we can have direct connects with some of the airlines instead of just via the GDSs and their NDC connections. And hoping that we can get some strategic partnerships together around that to help improve. And as we get to '23 and whatever versions, it's going to get better and better, and we understand that. So we need to lean in, and it's going to be extremely painful. And then out of the back of that, which I'm sure we'll get on later, is how do we then use automation AI and all of the rest to help service that or bring that cost base down.

Peter Phillips:That will be—

Ian Kershaw:we're going to come back to cost base before we leave distribution. But Peter, there's revenue opportunity here, right? Embedding, embedding ancillary products. You represent a company who offers a protection product. We just saw your video before.

Evert Meyer:Yeah.

Ian Kershaw:So what are the opportunities here, particularly if we're talking about what are the trends in distribution, how the technology and the And the ways of doing business are changing.

Peter Phillips:Yeah, so I guess I'm somewhat of the outsider here sitting between our OTAs and airlines. But so from the video that played just before I started, which was the first time I've actually seen that video, it's very informative. We do work across, you know, all different lines of insurance products and warranty products, but travel is obviously a big part of us. We work with large OTAs all around the world and large airline groups as well. So I guess when I talk about challenges and opportunities, I guess from my lens, I'm really thinking about the distribution of ancillaries, either to OTAs or to airlines directly. And there's plenty of challenges, but the challenges are solvable. And if airlines and OTAs can overcome those challenges, I think there's massive opportunity to unlock revenue growth for them. So maybe I'll just quickly touch on some of those challenges. From my lens, I do see technology as being one of the biggest challenges. for a lot of our mature airlines to sort of overcome. You know, they typically are working with large legacy systems and they're quite complex, and that really is a barrier to innovation and can make it a lot harder to embed these personalised protection lines to really grow your ancillary revenues. And I guess probably the second challenge I'll call out is arguably a lot more complex than the first, which is Regulatory compliance. Your example was fascinating, by the way. I've never heard of seat spinning, but I'm going to start looking into that.

Justin Warby:Don't start it on our website.

Peter Phillips:So regulatory compliance, I mean, for an airline to offer protection products, be it insurance or non-insurance products, to their global footprint means that airline needs to go through a maze of different regulations, compliance frameworks, and licensing across pretty much all the countries they operate in. And it's usually that operational and legal burden that means that they end up offering, say, a standalone, one-size-fits-all type travel ancillary product. And, you know, these— both these challenges are easily overcome by working with, you know, a good partner. And if you can solve them, then the opportunities really are massive. So for me, I guess the one thing I call out is If you can overcome these, you can really drive a greater customer relationship. So the airlines are really trying to drive a deeper understanding of their customer, and they want to be involved in all that customer's touchpoints throughout the whole journey. And by providing personalised ancillaries embedded in the flow, you can actually unlock that. Now, probably most, some people here, a lot of people probably might see ancillaries as a burden or a hindrance to actually maybe the core transaction or the core flight purchase. And I'm happy to say that we've actually got case studies from our partners that show that the opposite can be true as well. So one of our partners is Scandinavian Airlines, and they've actually told us that they've found that their net promoter score, so their NPS score, for customers that are buying truly personalised embedded ancillary services is actually higher than those that aren't buying. ancillary services. And, you know, everyone knows NPS scores are impacted by a whole range of different things. But, you know, they believe the answer is quite simple about why this is the case. And it's really to do with being able to offer a truly relevant ancillary service at the right point of time and really trying to enrich and enhance that customer's trip. And by doing that, you can really drive enhanced customer loyalty, And really just has flow-on effects through revenue.

Ian Kershaw:Thank you. I think maybe let's pierce the wall for a second on AI, because you're talking about personalisation. And so what's the potential? What's the role? Or how are you starting to use AI to improve the offering?

Peter Phillips:Yeah, sure. So I think when we talk about ancillary products, the use of AI in ancillary products is really changing how we need to think about travel insurance or other ancillary products. Historically, travel insurance in particular has always been a sort of a static, one-size-fits-all sort of approach. So by and large, regardless of the fare type or the destination or the consumer profile, the benefits of that travel policy has largely been static. And through the use of AI now, we're really able to accelerate the adoption and the change from that model to one that's a lot more dynamic and a lot more personalised in real time. And that then creates lots of opportunities for us to increase revenue, increase NPS, increase customer loyalty and brand recognition.

Sean Sutherland:Can I just add to that?

Evert Meyer:Yeah, please.

Sean Sutherland:Just in regard to that, obviously the benefits are, you know, if you've got a customer who's going to Las Vegas, you don't need to include ski protection in that insurance. So that drops the cost of the product for the customer, and there may be something else you want to put in. So it's fully customised onto what you would do, and that's why their Net Promoter Score's up.

Ian Kershaw:Did you want to jump in, Evert?

Evert Meyer:Not on that topic.

Sean Sutherland:No, then we can move on.

Ian Kershaw:Let's go back to cost, and then we can spend probably most of the balance of the time on AI topics. Cost stack, if you will, and distribution has changed a little bit, and we mentioned before kind of the The running battles in the background that we don't see. I think, you know, you've got a good example, Justin, from your own experience, but please, anybody else jump in what they're seeing in the background. You want to start with yours?

Justin Warby:Yeah, I'll give another one. Like, you know, we work with our OTA partners. They're, you know, a good source of volume from a revenue perspective. You know, they've got the scale. But the way that some of the OTAs now are using AI, they're looking at a number of factors, whether it be differences between availability on a point of commencement versus point of sale, variances in taxes, variances in forex, different deals that they have with volume incentives with the GDSs. So we'll find the cross-border selling in OTAs moves around the world. Based on the AI, the machine that's doing all that heavy lifting from the OTA perspective. Now that's fine from the OTA, but from us that can actually have huge cost burdens because not all of our— we don't pay our sales equally based on where the sale was commenced or started to be booked. So we've done a really extensive program this year to try and identify those arbitrage tactics that are going on from a, from a particularly an OTA perspective, but not always. And just by cleaning up some of our fare filing, changing some of the way we reward, changing some terms and conditions in our fares, we've actually been able to see the bookings move to places that we wanted them to occur. And you think, oh, that's not much. But just that activity in 6 months has reduced our GDS cost by over $10 million US.

Ian Kershaw:Straight to the bottom line.

Justin Warby:Straight to the bottom line. So it's not any more sales or any less sales. It's the same sales. It's just where they've been transacted in the world. So yeah, it's a constant battle.

Ian Kershaw:Is there anything that's changed in the cost? From the perspective of Flight Centre?

Sean Sutherland:Obviously we now have to shop twice, whereas we wouldn't have had to. So there's obviously the computing power that goes with that, and then all the deduping and what have you, which obviously over time will disappear. And we have bad actors as well. The same, Australia's rife for fraud on domestic routes because we don't check your licence or anything to get on a flight. As long as you've got an e-pass, you can get on. Unlike North America and a lot of Europe. So we see a lot of fraud happening where there's— especially the day of travel. So we're always fighting that. And obviously we get a chargeback on that every time. So that's ongoing battles. We have the type of thing, try to be sophisticated tools out the front, but they're always— then we're not talking about what's been mentioned earlier today. We're not talking about someone in a garage. We're talking about big actors that are making a fortune and they've got a lot of resource behind them. And it's a business for them. It's not just a side hustle.

Ian Kershaw:Which is coming out of your margins, right?

Sean Sutherland:Yeah.

Ian Kershaw:Stepping into the middle. Why don't we spend some time on AI to make sure we have enough time for it? And it's been a lot of advances in the last 12 to 24 months. I'm sure we're all using it personally in an enterprise capacity. And so today we're going to talk about the applications in particularly the airline industry. We don't have a hotel up here, unfortunately, but that would be interesting too. Evert, you shared that you have a longer history than most might expect with artificial intelligence. Can you briefly take us back to your PhD and early days with American Airlines? And in particular, you were looking to solve some of the same problems that you're looking to solve today, but now you have the, uh, the tools to do it.

Evert Meyer:The tools and the computing power. Yes, that's right. Yeah, in the '90s, I found myself as a South African exchange student in in the US at the University of Texas, and I was looking for funding as you do, and American Airlines and Dallas-Fort Worth Airport had this problem where, you know, storms, tornadoes would come through, shut the hub down, and how do you restart it? And they had big data centres or mainframe computers back then running in Tulsa, Oklahoma using AI— no, not AI, integer programming and traditional mathematical techniques to recover from that. And it was just the Impossible task. I don't think it ever amounted to much. And I had this professor who was a mathematician from MIT that suggested, hey, let's look at fuzzy logic, neural networks. Eventually we settled on genetic algorithms with a lot of neural networks sort of prepping. So very sort of bleeding-edge PhD stuff. Back then, I'd never imagined I would ever get to use it, right? Because—

Justin Warby:Yeah.

Evert Meyer:That was my entry into the airline business, and I had a career in strategy and network, and now I do fleet as well and other long-range planning at Virgin. And here we are, and suddenly it's with us and it's everywhere. And I think I just wanted to touch on the things that may not be immediately apparent to most of you out there. We all are now familiar with ChatGPT.

Peter Phillips:Yeah.

Evert Meyer:in uni and high school. They are masters at it. It's amazing. Half of my team that does advanced analytics and operations research now use ChatGPT to write probably 80% of their code, right? So all the software we develop, you know, people say— we went to Agaforce and did a presentation on a piece of work we did there. And how hard is your team? This is an incredible piece of work. Oh, well, one guy and ChatGPT. Develop this application, right? The ability to sort of just build a website, do a portal, you know, do your own IP. It's incredible how powerful those tools are becoming, and they're going to just change the way we do business on an IT delivery perspective.

Ian Kershaw:So leave this off here. What are some of the ways you think near term and then longer term?

Evert Meyer:Right now we use it extensively, probably in 2 areas, right?

Sean Sutherland:We—

Evert Meyer:there's a lot of vendor products out there that we're testing and evaluating and starting to— so, and some of them are from the major vendors like Sabre and Amadeus. But the AI capability needs to be trained. You need huge datasets, you need data to train it, specifically if it's your own data, your own customers, your own—

Justin Warby:Right.

Evert Meyer:Sort of operations. And the data just in handling that, you use another AI bot to do that, right? So you use Databricks and that makes that a lot easier. You don't need teams of SQL experts to do that anymore. So that's been happening in the background. I think most airlines are doing that. And at Virgin, because as Dave mentioned this morning, we have so much catch-up to do, we have no choice but to sort of really do it in a very innovative, Yeah. sort of fast way. You know, we cannot afford to have too many big projects with vendors that take years, right? We have to be innovative. So that's, that's been going on and it's been quite successful for us. But there's another area that I find particularly sort of exciting and it's sort of really starting to show dividends and that's using AI and various AI techniques, back to my PhD, to really model scenarios in our operations and test network scenarios. So in airlines, traditionally, you have various teams that hand over complex pieces of work to the next team, to the next team, and it all goes downstream. And often the intent of the schedule or the hopes you had for it, how it would perform, sort of deteriorates. And by the time, especially as we saw after COVID, what actually happened was completely different and not what you suspected. So that ability to sort of simulate and use AI to make simulations of things that are still going to happen in the future possible is really where we've been playing around and finding some great success. And the next level we're now looking at is in Australia domestic, you're not on your own. We're a small country, 26 million people on a huge continent. We face a large competitor who has airline offerings on either side of us. So we have to think through the competitive response and the secondary and the tertiary effects of all the decisions we make.

Justin Warby:Yeah.

Evert Meyer:And we're just finding that those models, now turbocharged by AI, are just opening up new opportunities in planning and putting out schedules that are back to where they should be in terms of on-time performance and operationally while running at incredibly high load factors. And, you know, the schedule density we're putting out there is quite remarkable at the moment. It's more like a low-cost carrier because we are fleet constrained essentially, right? We, we have to do that. So, so it's, it's really worked for us in the operational optimization.

Ian Kershaw:That's great. That's really good to hear about the operational side. I think often you start with something in discussions like this that's important, and let's talk about it quickly now, which is the customer side, the customer-facing. That's kind of where it starts in a way, a chatbot, right? The chatbot gets better, it's an AI agent and so on. And I think, Sean, you said that Flight Centre is looking at this, the kind of front-end customer interaction.

Sean Sutherland:Yeah, so I think for us there's 3 When we're talking short-term, obviously customer service, and obviously we'll take that easy and lean in. We don't want to be Air Canada with their challenges with just a simple chatbot, not even AI. The next one is predictive analytics, which touched on how do we price automatically and how do we present personalised offers even to ourselves, to our customers. And then the third one is in all of the automation in the backend. How do we get rid of all of that overhead? They're the 3. And obviously we see the future as agentic AI. Our luxury brands would say that they call it— what they're striving for is humanised digital. So how do we blend the two together? We believe that there is going to be a need for travel agencies or whatever you like to call them, depending on what brand you operate in our group. And of course there'll be customers like my children who will not want to talk to a person. They don't talk to me, so— and trying to get them to talk on the phone is impossible. You get agents. But even they won't— they'll be more than happy to deal with an AI bot, and there'll be other customers. And we've definitely got an older demographic, especially under the Flight Centre brand, and they're probably still going to want a bit of handholding and whatever. So we have to pull them both together and present it to the customer so they get to make a choice where it is, with obviously bringing in operational efficiencies in the back end.

Ian Kershaw:You're opening the door to a really interesting topic. I'm hoping everyone can jump in as much as they'd like. If we're entering the AI agent era, then traditionally what happens is all of us, we go on and we look, we search, we're inspired, we research, we book, and so on. And so we might be interacting with technology that is working to personalise offers for us and so on. But if we are in the future using a an AI agent, a personal travel agent, then you have systems talking to systems. And the one I'm using to book my travel will know my preferences, my schedule. I will hopefully have prompted it correctly. I don't know about everybody here. I'm still kind of woeful at prompting. But then the systems are interacting, and that has implications for distribution, for marketing, for ancillaries. I'm not sure who wants to jump in first.

Justin Warby:Probably whoever just We're literally the new AI kid on the block. So look, when you're looking at problems to solve, you can do it with people or you can do it with the computing power. Luckily, we tend to have a lower cost base on the people side. So at the moment, we're taking smaller steps in the AI space.

Peter Phillips:I guess maybe I'll let the experts here tackle that one. I'll just touch on one part, then I'll pass it over to you. About the agents talking to agents, you know, we are working— some of our large OTA clients internationally, they are, you know, they're preparing themselves to sort of be disrupted in how their customers are booking flights or hotels or whatever it may be. And they're preparing to be more conversational in nature, which, you know, you would expect. And that's all fine. They're working through those issues. For the ancillary sales, they're worried— or not worried— they want to ensure that they're still able to get their revenue as this distribution changes. So we're working closely with them so that we're able to be that backend agent that can actually start to potentially answer questions about, say, the insurance policies, because the view is that as the travel industry the original transaction moves to more of a conversational nature, naturally the people may start asking questions about the protection offerings themselves.

Justin Warby:Specifically.

Peter Phillips:Where historically very few people click through for the policy wording, very few people go into the detail about what they're actually covered for. So then that shifts a little bit because we need to be able to service that and still be in a compliant fashion. Obviously, they are regulated products in a lot of jurisdictions. So we're working through those issues. I'm really excited to see what the future holds.

Evert Meyer:I think it's a mistake to just see AI as sort of a one-for-one replacement of humans, right? So if you, if you think about an AI bot is and you personify them, and I know a lot of people in my team even give their AI tools names, right? They develop personalities. And I think that's fun and it helps us visualise it. But I think it's a mistake because you're going to ultimately not only run into sort of team morale, you know, all the industrial issues that comes with that if you go down that path, but you're selling AI short and you're not using it to its full capability. So I think the hard thing is for organisations, we need to take a step back. You need to look at the overall function. You need to ask yourself, what is this technology doing and what can it do that you're not even doing today. Redesign the whole process. Redesign the interactions and the handoffs and how data is interchanged because that's often the hardest bit, right? And then use the technology for that. And I think when you do that, you'll find, yes, you may have displaced some humans doing traditional functions, and obviously it sucks for them, but what you've probably created is new skills, new roles, new—

Justin Warby:New careers.

Evert Meyer:You know, I think the young people today that are going to university and studying these things, I think they face an incredibly exciting future. And I tell this to my daughters, is like, just focus on that stuff because that will be the world you live in. We'll just have to do things completely different.

Ian Kershaw:Anything, Sean?

Sean Sutherland:I think the big challenge for all of us is that we don't know what's possible yet. So you— and the more I've just started reading a book and it was released in March and the first 3 chapters are about the limitations of AI and have already been fixed. So this is how fast it's moving. You cannot, like, you couldn't unless it was your full-time job to read all day every day, you wouldn't be able to keep up with what's happening in the marketplace. So, and by jumping too far ahead, you're setting yourself up for risk anyway, because if it's not designed properly, in the first place with the customer in mind or the user in mind, then it's very likely you'll break something or cause some unforeseen issue for yourselves. So I think there's a lot we're going to envisage in the future. I agree, humans are key to be involved. There will be other jobs coming out of it. I mean, you know, there's conversation engineers now because how do you want your AI to talk to your customers? That has to be programmed specifically with your tone of voice and whatever. A lot of those types of roles didn't exist. So your CX teams are probably trying to train themselves up on that. And that's just one of many. And we already know that the prompts will disappear probably in the next 12 months. So you won't have to prompt. It'll remember a lot of stuff. There'll be, especially generative AI, short-term memory. It's like a goldfish. It forgets what you said yesterday. That's going to change over time. It'll start to remember the conversation you had and your preferences and whatever, especially if it's digging into your personal database at Virgin or Flight Centre. So it is exciting where it's going, but I'd say it's small steps, understand where you're heading, and just remember, great is the enemy of best. You just want to get better and better. You don't want to go for the great. piece right off the bat, because you're going to be doomed for failure. And by the time you get there, everything would have changed.

Ian Kershaw:Well, you raise a good point around, um, the customer experience. And there's been ear chatter perspectives that there's, um, a risk of lower customer experience or even reputational risk through AI. What happens if it hallucinates? What happens if it— um, now notwithstanding that in 3 months this might all be fixed, There's the flip side that's also potentially true, that people may actually much prefer it, that it may very soon give you a far better answer than the call center person that you get connected to after 15 minutes, and so on. And so I guess to the group, the balance of perspectives on likelihood of improving the customer experience or decreasing it.

Sean Sutherland:I think that's— The key is you don't want to try to look at AI as replacing all your humans. If you do that, I think then you're going to set up for failure. You need to be supplementing, learning, seeing what happens in the market with how customers interact with that, and then start to move forward with your customer base. If you try to, you know, I'm going to replace all my call centre with AI, then I think that's highly risky from my point of view at least.

Peter Phillips:I think there's an interesting point that I think combining both these The data that you're training on is obviously a key part, and I agree, you don't want to replace all your humans with AI or with any sort of automation. What we found actually, which was interesting, the call centres and support people, we were auditing the responses, etc. As we were training our chatbots, it was picking up what it was classifying as incorrect responses, which turned out to actually be the correct response, but our humans were were, you know, incorrectly providing the support, as it were. So, you know, there's opportunities on both sides. There was one point here around— I think it was from you, Sean— you know, for our claims handling side of the business, you know, we believe that speed to claim resolution is a key benefit for people. We want to be doing that really quickly. Conversely, though, we do want a human in the loop, especially when, for any declined claims, you know, human in the loop those. But if we're going to approve a claim and we're going to pay out money, we want to be doing that as quickly as we can. And we see AI to really be able to speed that up dramatically.

Evert Meyer:Maybe just moving beyond the sales distribution angle a bit. The one thing we've noticed when we look at vendor-provided products with AI features, which you have to be a bit skeptical about some of Right. It's a bit like running around with a hammer trying to look for nails to whack, right? But a lot of the effort, because you have to train this and you have to provide so much data, and then you get it, is the test and learn on the back end. And we have one project going at the moment where we can't quite figure out, is this really working or is it not working? And then there's this big debate about, well, is the conditions of the test and learn right? So now we're running certain products on alternate days, the human, the AI, and then you look at the data on the backend and are we seeing the differences? What we do know is they drive very different strategies. The bots after a while go off in all sorts of weird corners of the solution space. And some of it is incredibly educational. You learn, why is it doing that? And what is it looking at?

Ian Kershaw:Yeah.

Evert Meyer:But half the time it wanders off for no good reason, and you definitely don't want that. So you have to stay on top of it. And so the ability of humans to do traditional statistical and just large dataset scientific checks and balances is going to be even more important than ever because it's such a black box, right? If you think about it, what is it? What is AI? It's a big neural network. It's a digital array of synthetic neurons and synapses that mimic the brain, and you just show it lots of data, and it sees all sorts of vague patterns that we don't, as humans, quite understand why it's seeing it, and it produces stuff, right? So I'm a big fan of just in parallel with that, running your traditional statistics, your traditional big data science, basic mathematical models, and just always Trying to sort of see how the one performs relative to the other. I think that's going to be the solution going forward.

Ian Kershaw:I want to follow up on that, where we're kind of talking about vendors inherently kind of in-housing, and there's a spectrum. I'll just describe kind of roughly 2 ends. One is you in-house everything, and that has an implication for the skills, capabilities, team, etc., that you need. The other end, you outsource it. Which then implies the skills and capabilities you need are much more around contract management, people who know how to use it, to really understand the technology and what you're getting and how it's going to be used, but nothing in-house other than that contract management knowledge of it. People will probably find different places on the spectrum where they want to play. But I mean, Evert, maybe starting with you, and then everybody jump in. And I think we'll hear from COVID Genius.

Peter Phillips:Yeah.

Ian Kershaw:On the vendor relationship. But Evert, what do you think about operating on this spectrum? Do you think that there's a place where most airlines, or even hotel groups, or other travel providers will kind of land on in-housing versus outsourcing the skills they need then inside the company?

Evert Meyer:Look, it's going to become so pervasive that if you're not right now sitting there thinking and strategising about how you're going to do it, You're going to be left behind. We're at the beginning of an S-curve, and the S-curve is just going to accelerate. Sitting back for another year or 2 and just waiting to see what happens, at least in the airline industry for sure, is a dangerous strategy. On the other hand, you don't want to be on the bleeding edge. You know, back in the '90s, we were clearly still on the bleeding edge, but the computing power, the just horsepower of data, is now there, right? It's made it available. You can go online and find thousands of university and academic articles that have incredible technology written back in the '80s and the '90s that have just been non-usable, that suddenly are actually achievable quite easily. So you have to. And if you just rely on vendors, then at least you need experts to really interpret the vendors and who's selling, you know, You know, ice to Eskimos, and who isn't, right? So you need the expertise, you're going to need people, and the young generation that's coming through university, the young graduates that we see come out of the unis these days are incredible. Everybody complains about Gen Z and they're lazy, but man, they're smart and efficient. And we're just going to have to embrace that and work with that as organisations going forward, right?

Ian Kershaw:I think you were nodding your head, Sean, about the capabilities required, or—

Sean Sutherland:I agree. It's just going to be all of our life, and it's not just going to be in our businesses. It's going to be at home, you know, the way we work. If we're not leaning in and learning as a group, then you are going to get left behind. So I think we're going to have to have the skills in-house. We're going to have to rely on vendors. I do agree. A lot of the vendors just went to their marketing department and said, all of that marketing collateral, do a find and replace on machine learning, and Insert artificial intelligence. That's pretty well a lot of what we see in vendors. So you do have to have a bit of a dig around and see what they're offering. But there are some obviously big players out there that are obvious, and that's probably where I'd start. And then the next one's going to be— because the hardest bit's going to be integration and getting your systems up to speed. Because you can have great tech, but if you can't integrate and you've got bad data, then it's irrelevant.

Ian Kershaw:That's actually— you've given me the perfect segue. Perfect tee-up point to come back to Peter. And what does a successful partnership look like around AI?

Peter Phillips:So, I mean, I guess I'm the vendor in this relationship here, but, you know, I have my own vendors that I deal with. So if I, you know, abstract myself away, you know, I think you have to— just coming back to the first point— I think you have to sit in the middle there. I don't think you can just rely on a vendor to do everything. You need to have some of those skills in-house, otherwise you're definitely going to be left behind. To your point, Sean, like everyone is just saying they've got AI and potentially there is zero to little value, very little value if you can't do that integration. So I think a successful AI partnership, you know, the end goal is really going to be, can that partner really be a co-driver of innovation with you for your business? That's what you want. That means that partner needs to be deeply integrated. You're sharing data between yourselves, you're co-creating solutions. And you're really being able to deeply embed that technology, be it a protection insurance or whatever, or just one of your other vendors. You really need that deep synergy of what you're actually trying to achieve. If you are just using an off-the-box, out-of-the-box solution, a vendor comes to you with just a piece of software or a pre-built solution, you're probably not going to get the outcomes that you really want from it. So, So I think, you know, just look for vendors that you can really co-create solutions with.

Ian Kershaw:We're under a minute. I'm not sure if I've seen any questions come through from the audience. If there haven't, I'll ask one more and then we'll thank the, thank the panel. It's gonna be a quick one in a minute. Rapid fire. Um, really rapid fire. Any perspectives across airlines, agents, anywhere else on how it's gonna, how AI may change the competitive landscape? I know we need more than 30 seconds, but—

Sean Sutherland:I think unfortunately some of the people with the deepest pockets are going to have a lead because it is not necessarily going to be cheap. So that's going to be a challenge.

Justin Warby:Yeah, I'm worried about if you start getting rid of people because you are starting to use more AI, will the cost of AI be the same in 5 years' time? And then you haven't got the skills to reverse yourself out of it. So I think the blended approach of leveraging AI but keeping your skilled people is important.

Ian Kershaw:Final words?

Peter Phillips:I mean, I agree with this sentiment, but also coming back to the rapidly evolving technology, you know, with the newer models that are computationally less dense, I think maybe the cost isn't going to be an issue. We don't know. That's the problem. I share these concerns as well though, but I see the future as being— I'm an optimist, so I see the future as being bright.

Evert Meyer:I think AI is not a substitution for labour and cost. It is truly— it's going to have the same impact on our industry that computing had in the '80s. I remember my first job as a young consultant was SAP system implementation. I did the first implementation at an insurance company, right? So you did those big mainframe transition stuff. And back then the dream was that we'll take all these people's jobs away and make this cheaper. Never happened. It created a whole new industry that was so much more complex and expensive than the one it replaced. But what improved was the capability, the power multiplier. And I think this is going to be the same. AI will just make our lives harder, more complex, more Challenging, but the productivity gains, the, the, the, the, the, what we give to our customers, the flexibility, and they will take it for granted. They will get used to it. They already are. If you're not playing that game, they will just say you're an ancient dinosaur of a company. You're being left behind. You have to keep up because that's what they're growing up with now and they'll expect that.

Ian Kershaw:Wonderful.

Sean Sutherland:Wonderful.

Ian Kershaw:Thank you so much to the panellists.

Copyright policy: All transcripts on this site are the copyright of CAPA - Centre for Aviation. Our reproduction policy is as follows: you may quote up to 400 words of any transcript on the condition that you attribute the transcript to CAPA - Centre for Aviation and link to the original video page. All other use is prohibited. While we aim for 100% accuracy in the transcript, there may be some minor transcribing errors.