How AI is reshaping the travel experience from planning to landing

And welcome to the travel breakout. How artificial intelligence is reshaping the travel experience from planning to landing. My name is Bob Quick. I've been with AWS about 2.5 years. I'm the head of airports and ground transportation.

Before joining AWS, I spent about 20 years working in the travel industry, airports, airlines and travel technology. I'm part of the AWS for travel and hospitality team. Our team has expertise in airports, airlines, ground transportation, restaurants, accommodation and lodging, casinos and cruise lines.

We provide guidance to our travel and hospitality customers and we help develop industry solutions. We also identify partners who can help our customers on their digital transformation journey - partners with experience and expertise in travel and hospitality that can help our customers reinvent the travel experience.

Travel has been built on innovation. In fact, flight itself has been seen as a symbol of humanity's inventiveness and creativity. When you think about inventions and innovations, very often you get this kind of cascading effect where multiple inventions come together to solve really big problems.

A great example of that is the invention of powered flight in 1903 where the Wright brothers on their journey to inventing the Wright Flyer, they had to invent lots of things like the wind tunnel and wing warping. Wing warping is the twisting of the edges of the wings in opposite direction to control flight.

Once the Wright brothers invented the Wright Flyer, we had to invent so much more like airports and airlines and travel agencies. That spirit of innovation continues today.

A real nice example is the invention of the QR code. Once we invented the QR code, we needed to invent standards for the QR code to be used for travel. With that, we could invent putting the QR code on the boarding pass and that allowed us to invent things like self-service boarding and other self-service operations in the airport.

Then we invented putting the boarding pass itself on the mobile phone to make travel even easier. Just a quick poll - how many people here when they came used a boarding pass on their mobile phone? Well, that's pretty good. That's pretty good. You know folks, this is a tech conference. So next year I want to see all hands up for using your mobile boarding pass.

When you think about your journey, there are a multitude of systems and technologies that take place in the airport in order to facilitate travel. Considering there's almost 100,000 flights every day, it all works pretty smoothly. But sometimes it doesn't and even when things don't work well, the most common complaint that I hear from people - and if you think about your own experience, the complaint you most often feel - is "I wish I'd known".

Like, I wish I'd known the line on the other side of the terminal was shorter or I wish I'd known how long it was going to take me to get through security. You also think, I wish the airport or the airline had known. You want those organizations also to have that situation awareness so that they can do things like forecast the length of the queue and balance the queue between different parts of the facilities.

As a modern traveler we want and we expect to have real time information on our journey, but airports are complex operations with multiple systems and organizations working across the globe. In order to have that real time personalized information we need to have and we need to use artificial intelligence.

Now, artificial intelligence, I think it's fair to say has captured the imagination of all of us in this room. I think a lot of people outside the room. And we've already seen the cascading effect of innovation in artificial intelligence. For example, we've seen the invention of Amazon Alexa and computer vision. And of course, generative AI.

We have seen from surveys that industry leaders in travel and hospitality really believe in the power of artificial intelligence to improve their operations and improve the travel experience. We have lots of organizations who are already getting started, as we've seen today during the announcements at re:Invent, getting started using artificial intelligence.

Today we're going to hear from two industry leaders on how they're using artificial intelligence to improve the passenger experience and improve operations.

Manchester Airport Group is the largest UK owned airport operator operating Manchester, Stansted and East Midlands Airport. In my career, I've had the pleasure of working with Manchester Airport many times. I've seen firsthand how they use technology to improve operations and improve the experience. I've also personally flown through the airport many times for business and pleasure and I've experienced how those technology investments then translate to a better experience for me as a passenger.

TUI Group is one of the largest travel organizations in the world. They have thousands of travel agencies, they have their own cruise ships, they have their own hotels and they have their own aircraft. I've also personally experienced the seamless TUI vacation going to Cancun Mexico on a family holiday and we all flew in a TUI Dreamliner and stayed in an all inclusive TUI hotel. My kids were amazed that the hotel had its own branded aircraft.

So today we're first going to hear from Mark Jennings, CIO of Analytics and Artificial Intelligence and Customer Experience at TUI.

Thank you very much, Bob. Thank you. So, yeah, good morning, everyone. Thank you for coming and attending and listening to what we have to say. I'm going to take you through some of the experiences we have in modernizing our business, how we're working with AWS and how they're helping to shape our future and how they're helping us to use AI to develop better travel experience products and services for our customers and for our colleagues.

A little bit about the company that I work for. So TUI, as Bob explains, we're one of the largest leisure travel organizations in the world. We operate over 16, more than 16 cruise ships. Now we have approximately 480 of our own hotels. We have an extensive in resort destination service around the world and we take over 20 million customers on holiday every year.

Our ambition is to deliver great experiences to our customers. We're continuing to grow and build our customer base. At the same time, we want to do that, we want to deliver a better experience and a better service. We want to do more and make sure that the products and the services and the portfolio is extensively grown.

So if our business is trying to modernize and expand and grow, how is technology playing its role in supporting the business on its journey? If we wind back seven or eight years, we identified that technology needed to develop, needed to modernize. We needed a strategy that we needed to execute.

Our strategy was a simple acronym that we have - ABCDS. So we needed API, we needed to use and exploit big data. We wanted to move to the cloud. We also needed our dev ops to have a faster lifecycle and the ability to implement much quicker. And lastly S was around security - we absolutely want to have a security first approach, protecting our customers' data and our colleagues data.

What did we do? So we reached out, we worked with AWS, we partnered with them and they helped us in our very first migration to the cloud and that was some of our web stack and that was very successful. That gave us the confidence that we had supported, we had engaged with the right partner on our journey.

It also gave us the confidence to be bold as a business and to overcome some of the skepticism that we had at the time about moving away from on-prem to the cloud. Without moving to the cloud, all the other things as part of our tech strategy, we wouldn't have been able to do. It was the one thing that has helped us to unlock and modernize our environment.

As we've moved then over the years, over the last 7 to 8 years, we've constantly reinvented. We've looked to accelerate our development and we're now in that period where we're a much more mature organization working in the cloud and with the product set, but there's still more that we can do.

So what is the next step? What is the latest thing that is going to help us to be better, going to help us to be faster, going to help us to be more effective and efficient?

We absolutely understand that generative AI or AI in its broader sense that does incorporate machine learning and data science, we believe that there are huge opportunities throughout the whole customer and colleague lifecycle - all the way from planning and inspiring someone when they're looking for a holiday, all the way to post booking where you're getting a CSQ or customer survey back.

How are we going to use it? How can we identify a number of initiatives and capabilities that we can significantly improve? So what does this mean? And what's the approach we've taken? Working with our board, there are some top initiatives - the top 10 initiatives, these are big complex change activities. They require lots of resource, they require significant decision making by senior stakeholders.

These things are being pushed top down and people brought on the journey. But the nature of AI and AI is such a broad, there's such democratization of the capabilities that are available and it's moving so fast. So as a business, we don't want to just have top down ideas, initiatives pushed through, we need to bubble stuff up from the bottom.

To that extent we're trying, we are rolling out an extensive set of processes where we can engage and get feedback from people right across the business around the ideas and innovations and the things that they believe that they can change. We do that, we will take their ideas and then we are running fast proof of concepts.

When we're looking at this, we have a top down bottom up approach and together that will make us very powerful, that will deliver a very powerful set of outcomes and an extremely... It will then leave no stone unturned in our ability to deliver change.

As a business, we want our cake and eat it. We do want to be bigger, we want more passengers, we want more revenue. At the same time, we want to drive down costs and be more effective. Can you have both? We believe AI is the first thing that will really enable us to develop that change.

Where have we come from? Like many organizations, we have grown up through a series of acquisitions and mergers. We have many different systems that were old, data was disjointed, nothing was joined up. The definition of a booking wasn't the same definition of a margin - all historically not aligned. This meant you cannot exploit that.

However, over time, we've managed to overcome some of these challenges and that's helped us to certainly reduce some of our costs. It helps us to go faster, but this isn't enough. We need to do more. We need to be better partnering AWS and using the new technology will help us do more.

If we've overcome some of the challenges today, what is our future focus? What are the key things that we want to do in terms of exploiting that data and exploiting our systems?

If there's one thing that we knew during COVID, we learned from COVID, it was that we are nothing without our customers. Whilst we thought we may have been exploiting our customer data and engaging with them very well before COVID, I think what we understood is we needed to double down on our efforts.

So we needed to better harmonize and source this customer data. We needed to do more with it. We needed to understand, we need to have a better set of goals and targets for what we were going to do with data. We also needed to encourage our users and our business people to exploit and self serve with that data to make better decisions.

Lastly, how do we personalize? We have all this data - if we're sourcing it, if we're encouraging people to use it, what does that mean? Well, we want to take the data and we want to use it to turn it into personalized offers and experiences that our customers will enjoy.

So what is the first thing? What is the next thing we have done recently? In the last six months, we have started to implement, we have built our first foundational platform of which we have entitled the Intuitive Platform, the AI Platform.

If we wind back again, what were the challenges or things that we needed to tackle as a leadership team and as a data science and machine learning and AI team?

So the first thing is that whilst we started to, to have um a set of new tools, a set of new ambitions, a set of new objectives. We didn't actually necessarily have all the skills to be able to exploit the capabilities that were out there.

So what did we need to do?

We needed to uh recruit, we needed to train up, we needed to get the right people. Uh so we could match the tools and the skill sets to be able to exploit the technology that was available.

We didn't have a platform we had, well, we were trying many different things at the time. This led to us not being focused, not have not being able to exploit any particular one platform or technology that we were trying to use.

So we have overcome that hence the two intuitive platform.

And lastly, we needed to exploit real time data. So we ha we're very good with capturing batch offline data but but being able to source online data, then we could be able to do much more and be much more powerful in the capabilities and the outcomes for our customers.

So we've done that.

So what did we do?

We are we, we partnered aws and we, we have exploited their platform. The model you see here is a pretty standard model source, you have your data sources, your pipelines feature engineering store, all of this stuff. It is very uh it's very standard of the things i would call out that most that we focused on uh particularly that we were very keen to uh to make sure that we maximized.

Uh so how do we source data and then experiments so very key for us that we were able to demonstrate to ourselves and to our colleagues within the business that as we roll out new pieces of work, new initiatives, there's a benefit. So making sure we had a baseline set of results and then being able to measure a of every test that we did.

We wanted to do this as quickly as possible, monitoring stuff, getting feedback, learning that cycle continues uh endlessly.

So the products that we uh actually chose in the end, uh we've uh focused and doubled down on are amma saves maker and personalized um using these products. And this tool set has given us a framework and a set of products that now means that our data, scientists and our engineers, they can focus on building stuff, they can focus on testing and it has enabled us.

Not just my words, i can, i'm about to show you some data. These are the things that we have actually been able to, to demonstrate with facts and figures in terms of how we have got better uh as an engineering team, as, as a technology team, as a business team.

So we are now much faster in terms of the things that we can execute. So previously, so going from having uh an initial idea for, for a model um to it being uh built as it were, it's about 100 and 28 day, 100 and 20 days, we are now significantly faster. We can see that three times faster, approximately 40 days on average in terms of the output.

We have now been much more. We've in, in, in the same time period where we might have only have three models not just built actually deployed. We have now significantly grown that you can see the numbers in the same in a similar time period. We have 15 models live and we've actually built 31 models that are being tested, developed ready for deployment and in terms of upskilling and right skilling the pe the the the the people that we need previously uh because those people weren't in the right team, we had lots of hand offs.

This again became unproductive. So by getting the right people with the right skills sets in the right teams, they become much more self dependent or self sufficient. Therefore, we have less handoffs and again, we can go much faster, been a very good successful story.

But if that's the the tech side, how we deployed this, so we're using it in our web personalization and the models that we are built are helping us to deliver and improve our customers online experience and do this with, with real time uh interfaces and updates the products that we're focused on. They're not necessarily new.

So recommend us. Uh these things have been around for a while. Uh in fact, we've been using a third party uh hotel recommend a recommender for a number of years, but we've now been able to do build some of these products ourselves. We've got them live and we have a number of uh of, of new things uh in the pipeline.

So our hotel and room upgrades i will touch on. I'll give you some feedback, uh image personalization, the, the the trip advisors, these are all things that help us deliver the right product to the right customer at the right time.

So if i said we had a, we had a, a third party product that we were using before our first challenge or the first task uh uh that the uh data science and an a i team took on was to map uh or to match and better what the third party products could do and do that at a lower cost.

Uh again, very successful, they have driven through uh the the in terms of their up clicks and the click throughs uh in her. So it's a much better output, delivering much more benefit cos these things will turn into, into to bookings and the fact that they are now we're now in control of the code, the algorithm, we're now um much more confident.

And if you can see from, from, from the numbers on the slide that our original business case, we believe we will outperform it by 11 times. Um uh our original business case, which is, which is a fantastic result and something i've never heard of at, at tuy, it's a, it's an amazing result. Again, that's down to the team, the the platform, the experience, the learnings, the optimization being able to do that quickly and effectively and iterate its its excellent use the same approach with our room up room upgrade model.

So this again is delivering increased upgrades and bookings. We do this by showcasing the best products to the right person at the right time. Again, we're getting more upgrades that are booked now and we're getting again, a better click through uh and conversion rate.

This one not, not quite the 11 times improvement on the business case. Uh so maybe the team have some work to do. But again, it's a very positive story. And again, it frees up our team not to not to just continue to develop this. They continue to work, they can now work on new products rather than being stuck. Continue reiterating this to get to a point in time where it's as good as good as the business case.

So what was our experience before? Like many things? We were in terms of room upgrades, we had many options that we were offering to our customer. This was would obviously lead to, to in confusion and indecision.

So how, what do we need to do? Simple changes highlighted uh messaging needed to be improved, the ux but also reducing the number of options by understanding the customer, the segment, their party size, the party composition, their preferences and then being able to surface that up to the customer at the right time during the booking process.

So giving them a little nudge to the right product that has been very successful, more bookings, more profit, happier customers. It's a great success story.

So if we talk about delivering the, the right message, uh sorry if we're doing the right product to the right customer at the right time. The last thing is the right message.

So the, the next uh initiative that we have worked very closely with aws is about improving our content. And our seo uh and the, the the product that we have been able to exploit, working with the, the amazon prototyping team is to get experience of bedrock.

And our key objective uh is as we are scaling out as we want to offer more hotels uh and more products to our customers. How can we scale up the production of content at tui?

Tui has uh many different brands made use of c customer segments and we have very strict editorial uh and content guidelines um that, that meant this, this uh to date or previously. Um there are manual processes that we used and a manual content narration to, to hit the extremely high standards within um uh editorial guidelines.

So our objective was to demonstrate that we could use technology and a i to both uh automate it and scale it and deliver to the same level of standards and quality uh that our customers and oh, sorry that our colleagues expected.

We also wanted insights into how much this would cost, how much we could scale it and how fast we could go uh by using bedrock, we would be, we are able to do this in an, in an environment that's safe, secure, where we can use our own data and also with, with a partner that can help train and upskill us.

So what are the building blocks within uh the the process around content and seo naturally as they always, you have to source the data at the beginning, you have to what's gonna input uh into your process? Very simple. But again, pretty standard approach to, to uh a i.

Uh so we started with the prompt engineering. So we clearly, we didn't jump straight into fine tuning. The first thing we always do in this space is can we solve the problem using prompt engineering? We did that get the output uh generate. So it would then generate hotel descriptions.

We would then also uh check that uh against some seo rankings and scores, we would evaluate it and we would repeat and repeat and repeat. So uh i the feedback uh and learnings on how the content was generated would be provided by our content team, the seo scores and rankings we would get by uh querying out to search metrics uh and getting a ranking score this iteration we would then store in a dynamo database where we could retrieve it. Uh and check the data.

It would also enable the customer team to add comments. So if things weren't quite right, we could then use that to repeat back into our learning cycle as it happens.

The prompt engineering on its own didn't quite meet the, the the tone of voice standards that the uh the the content team um was suggesting were needed to, to, to deliver the right message to the right customers at the right time for the right product.

So therefore, uh we um decided that we needed to, to use the fa fine tune model, a model that will be able to fine tune, to be able to better deliver the right tone of voice. That was the key um key learning that we understood here.

So what did we learn from that? So, actually the, the winning model. So we want, we tried and evaluated many models that bedrocks offer us different sizes, uh different different configurations of models. And we found that the, the winning set was to use two large language models.

So that was llama two to do the fine tuning of content and then clawed to uh to be able to deliver the right formatting of the, the output and the descriptions.

Llama two was hosted on sage maker and then claude two was, was accessed from amazon uh bedrock services via an api we were able to automate this. All the content generation was automated when we used uh uh step functions to manage the, the controls of the uh the calls.

We then also then do the call out to the api for uh the seo scores. This was great. So we proved that we could do it.

Uh the, the uh the content uh that we generated, we did a blind test with our content team, the quality of the content uh that we generated automatically through bedrock. And the, and the products was of the same quality that the content team had uh had written themselves.

So we did blind scoring. We didn't tell them this was content they'd written, this was content that the automated system was proven. It was of the same when we took then the output of the automated content through uh through bedrock and that the services and large language models, we proved that we had a 4% uplift in our sco ranking again, a positive output.

And then uh the other thing is um when today we have to create aaa hotel description page takes about eight hours to do. Once we have a trained model, we can do this now in less than 10 seconds. And the training of the model that we built is cost tens of tens of dollars.

So it the the output is not, is more cost effective. Uh it's to the same quality and standard and it's also much quicker. But again, it's, it's brilliant news and again, it's been a great set of learnings that we've had.

So if i could look forward if there were some, some four takeaways for you

So the first thing that I would recommend is training and skilling your people. Um and that's not just technology people, it's people across the business, educate them around AI, the opportunities that it offers and then also give them the tools, give them the skills to do the job and help them on their journey. They're not going to get it right first time, uh let them learn and encourage them, pick a partner and work intensively for them. There's lots of focus, lots of things happening. Uh we've chosen AWS uh and very successful story so far. Uh measure and monitor everything you do. So change it. This is a period of huge change. Some people may be resistant to it. Let's use facts and figures as the evidence to help justify the great work that you do when you do it. But also I in any um difficult conversations, you may have to overcome any resistance to change and then lastly celebrate success. So the the world is changing so rapidly. Uh there's a new release every week. Uh there will always be the next best thing to do. Um so don't wait uh crack on with the things you need to do. Um stop talking, get stuff done um and celebrate those wins that you have. Um it is gonna be a difficult uh and bumpy ride probably for some of us on the journey. Um but yeah, make the most of it um and have fun. Thank you very much.

Thanks, Mark. That was awesome. Um you know, when people ask me what artificial intelligence is, I'll often say it's like having the ideal human expert help you perform a task, an expert who understands your needs, understands all the information that's available and presents the answer in a format that you can understand. And we got two really great examples of that from Mark using Amazon SageMaker to present the options to you, to upgrade your hotel room that is personalized to your needs in a way that's easy to understand. And then using Amazon Lex to create the descriptions of the hotel property in a way that customers and the organization requires. So we've booked our last minute vacation. We're all gonna go to the airport now and we're going to hear from Nick Woods, CIO of Manchester Airport Group on how he's using artificial intelligence to improve operations and the passenger experience. And we're going to start with the video if it plays...

TLm 2325 is an approach, prepare for arrival, receive automated approach arrivals, routine activated tom 2325 approach vector optimized using time based separation technology. Welcome to Manchester. Please follow the green lights to stand 107.

Um we essentially have about 7000 people that work for the business, but operating and working on our campuses, it's about 40,000 people. Um, so it's a big old operation that you have the opportunity to optimize and to think about. And when you really think about aviation and what it's doing, it's about how do you get aircraft into little boxes in the sky is efficiently possible to get them to their destination. It's just a big data problem. It's just thinking about these things in ways that actually how do we understand what's going on and optimize them as we go through? And that's what gets me really excited about thinking how we do this.

We've just relaunched our, our corporate strategy and we're really centering around three core things that we're looking at in our mission to essentially be the number one choice for our customers and for our airlines that, that want to fly from our facilities. Number one is all about delivering a world class customer service and doing that through a way that we're operationally excellent in everything that we do. Secondly, it's about being able to sustainably grow our airports and through that transformation program over the next five years, we aim to deliver about £120 million worth of value into the business from, from essentially reimagining what we do through digitizing what we do through transforming it. AI will play a big role in, in helping us to make that make that happen and thirdly, we were a founding member of the Jet Zero Council in the UK. Essentially looking at how do we decarbonize aviation? We've set an ambitious target for ourselves to be net zero by 2038. Um and were the first airports in the UK to be carbon neutral. So there's a lot of thinking going on around, how do we do that? Um real decarbonization drive, how do we drive energy efficiency? And it was great seeing yesterday in the sustainability section um over at the Venetian, some of the ideas that were coming around smart cities and things like that, that they're similar to the things that we're thinking on and working on today.

So um they're the kind of things that that as a business are driving us forwards and then how do we translate that into technology? So I will be next week relaunching my tech strategy to, to our team. So I took it to ExCo to our executive committee a couple of weeks ago. And really, we're building our mission around being the world's most intelligent airports about digitizing our operations around getting real good handle on what is going on, having a single version of our operating plan that is continually optimized and updated through artificial intelligence and made available to anyone who needs that at the right time to really go and drive the, the efficiency and the and the value back into the operation. It's about personalizing journeys for our passengers and for our airlines as well. Um thinking about how do we better understand their needs, provide differentiated products and services to them and deliver a much better experience as people come through. And in that, we're thinking about things like virtual queue scenarios. We're thinking about personalized products and services that, that customers want to engage with and consume from us. And it's about smart infrastructure. It's about again, better understanding our buildings, our assets, our energy management and being able to control them in a very dynamic way.

The video that I showed at the start was really starting to think about and, and kind of inspired by really um the amount of Alexa devices that I've got at home that my wife absolutely hates but have got the good morning routine that turns on the lights, turns on the coffee machine opens the blinds. How do we actually bring some of that simplicity around driving automation and routines into the airport's environment from what's historically been very siloed processes with different people who don't really communicate to each other. It's very hard on multiple systems to go and automate and think that through, through what we're building, we're building a way to bring that together and make it easy for us to go and um really drive that level of efficiency and automation into the environment.

We've been on a journey of product management for the last two years now. So we essentially think about two core flows. We think about the passenger journey, passenger will turn up in their car, they'll go through a series of products, check in security, the departure lounge, boarding and back the other way through immigration and, and out of the building and the aircraft does exactly the same. It comes in out the sky lands, taxis through our aprons parks does a turnaround process and then flies off again. And if we think about those two journeys, we think about the products that we discreetly within them and how we optimize them, link them together to drive an end to end overall efficiency and flow. That is how we're starting to deliver a lot more capacity for without having to spend money on, you know, significant new terminals and runways and infrastructure. We're doing that as well because we see the opportunity in doing it. But actually by driving this thing, we get far more value out of what we're doing.

So our approach, this is our intelligent airport architecture. Um and really, it's about building proper event driven, all knowing all sensing systems, right, the way across our airport operations that have that operational plan that we've built, we start 360 days out but continually evolve and update right the way through to the day of operation on the aircraft, actually leaving and taking off from, from our, from our facilities, having sensor technology in understanding the radars, understanding what's happening from our car parks using computer vision to think about how we turn the aircraft around and all the aircraft on time and performing the way they need to be feeding all of that sensor technology into our um into our planning systems which are all driven by artificial intelligence to continually optimize. If you think about the sheer number of moving parts that are happening across an airport on a daily basis, you're taking tens of thousands of events and messages every hour and thinking and feeding those into, um, into your AI engines to just continually think. Am I parked at the right stand? Can I move this to somewhere else? Is there a um, an aircraft pushing back slightly late?

We talk about turning an aircraft or talk about an, an airport day a little bit like a Formula One race. I know um the F1 was in Las Vegas a couple of weeks ago. It was great, um, great, great watching it, but actually it's the same, same kind of principle at an airport. You have a plan for that Formula One race and as soon as the cars hit the first corner, all those plans are out the window because the cars are in a different position and you're constantly revolving and replanning it. It's exactly the same in that airport environment. As soon as that first aircraft is slightly late or slightly pushed back, pushes back off stands late. You're constantly having to replan and optimize it. And today that's done in a series of silos or historically, it's been done in a series of silos by humans in the tower or humans in, in the operational control centers or people in the security halls trying to react and understand what's going on from the little limited visibility that they've got of their world. But as we bring all of these systems together, as we start bringing in um proper overview of everything that's happening and have one version of the plan that's fed from all those individual places. That's what's allowing us to really go and dynamically um optimize and automate that process so that we can, we can be much more efficient, delivering much better services and experiences for our passengers, planes leave on time more frequently. Um but also um much better efficiency for our airlines as well. So they want to engage with us and, and, and more, more five more terminals.

I'll give you some examples of some of the things that we've been then using that for. So I listened to John Hurley speak yesterday, um who is the CTO for Ryanair. So Ryanair fly extensively from, from Manchester, um Stansted and East Midlands. Um and we partnered with the Ryanair team really well and actually, it was um set up and provision through, through the AWS team because we're both big consumers of, of AWS AWS products and services. And we've used that event driven system to better start taking data from what is happening on the ground and feed them dynamically into Ryanair systems and directly into their customers app.

We have about 500 flights a day um flying from, from our terminals, tens of thousands of passengers all, all on Ryanair flights. And what we've been able to do by that is remove, someone's job was used to be to take the latest information and re key it into, into the Ryanair systems by, by literally screen scraping what was going on. We've fully removed all of that and automated it. We've been able to much better provide accurate information for passengers across their digital channels that they're getting from the airline across the screens that we are presenting um inside of the terminal facilities and through what we're presenting through our own websites and through people talking on the on the ground, in terms of asking questions of our staff, passengers are now getting a much more consistent message and what we're working on now is being able to extend that with much more information.

Um both from ourselves to drive Ryanair's operational efficiency, but that they're starting to provide back to work as soon as an aircraft is going to be a little bit late or something else, we know about it before um before it pops up and we can actually do something about it to be much more proactive about it. So that partnership is working really, really well um the one that we're working on is about version 360 where we're using um using AWS Panorama devices integrated initially into our CCTV platform to be able to take events and build algorithms around them to identify things that are happening across the airport. Whether that is that the airport aircraft is docked and you saw some of this at the start, the aircraft is docked on stand with the cleaning facilities have happened with the bags have come off yet. Feeding all while standing gate allocation to understand is the aircraft going to leave the stand at the time that we think it is and if it is going to leave the stand at the time, I think, or it isn't going to leave, what do we do about it? What's the next best stand that we want to go and allocate that to um we've been working with the AWS professional services team around this and our own internal engineering team to start going, developing those algorithms. And we're having a real success now actually being able to detect things very, very quickly. And for relatively low cost, we could go and buy some of this stuff off the shelf today, but they're built on all the technologies and we think that we can probably deliver this for 1/5 of the cost by doing it ourselves by building and using the latest technologies um that AWS are providing to us to be able to go and make it happen proper properly off the back of that.

We're then also starting to think about what other sensor technology we go and feed into that. So we're now starting to feed into things like the ground power units that are charging the planes. We're now starting to feed into things like um the runway lighting system so that we can actually dynamically route the aircraft to the gate right way. And again, it's by having this sensor technology is by having the AI to go and optimize it to think about the best routes in there to allow us to actually go and really drive value into the into the airport over the next 3 to 5 years, we believe that just from delivering better data, better insight and better planning, we can deliver about 15 additional based aircraft from Manchester and Stansted with a value to us in the region of £30 million. Um just from, just from actually optimizing what we're doing today. We don't have to build any new stands. We don't have to build any new terminals to go, make that happen. It's about using that big data, using artificial intelligence to really go and drive the optimization and to really increase the capacity of our airfields without having to do um you know, any significant building works to go and make that happen.

We're then also looking at how do we give this information back to our, um, back to our businesses and make sure that they can get the information of the staff working at our airports and we get the information that they need to be able to go and make this happen.

In Stansted, we have, um, a train line that brings in about 50% of the passengers to the airline into the airport is essentially a single track one way in one way out. If they have any problems on that line, which they do frequently, that can cause a massive cascade impact into how passengers are turning up at the terminals and therefore how they're flowing through that end to end journey into our facilities and into um into our security halls and and and onwards.

So we're now starting to actually track that by integrating into the rail service and the rail companies and we're feeding that into our into again into our planning engines to be able to say we'd expected in this 15 minute period. X number of passengers to arrive. We know because we've seen a problem on the on the train lines that that's not going to happen. So we're constantly replanning and thinking about right. Ok. We think it's going to be a 20 minute delay before we're seeing that surge. How do we deploy and send people on their brakes now, um to make sure that they actually get their down time to be able to prepare for the passengers that are going to be coming through in the next 1520 minutes or so.

And what you start to see is that these things never happened in the past because they didn't have the insight and the visibility. They would just continue to go against the plan that they already had for their schedules, for their rosters and for everything else. Um and actually, when the passengers turned up, everyone was on breaks and things like that. So we're really, really starting to drive that efficiency into our operation from being able to take all of this data in.

If you also think about what happens in that environment in a security hall, you've got x ray machines provided by one company. You've got body scanners provided by somebody else. You've got passenger flow technology looking at how big your queue is and what your queue times are for it. You've got a different set of systems that is looking at your roster and your timing for it through the event driven platform that we're building, we're able to start bringing all of that data together. And instead of having, you know, the dashboard that's provided by the x ray provider on one screen and another dashboard provided by the body scanner person. We're aggregating all of that data together in real time to be able to give us the real throughput information that we need in one place and a really usable, easy digestible thing for um for for our operations to be able to do it. And we've seen significant uplifts in terms of efficiency of those operations. We've seen significant uplift list in the mp s scores that we're seeing for passengers actually queuing in those environments and going through them. Um and overall, it's creating a much better experience for everyone operating at the facilities.

We know that happy passengers that are less stressed are more likely to get into our departure lounges and spend more money. Um because they're comfortable, they're in good space and, and we know that they're not stressed out as they go through it. So while this stuff is brilliant for us from an operational perspective, it has a real material impact on the bottom line as well in terms of actually being able to um prove the correlation between a good experience in those security halls which let's face it, they're there to make sure that bombs don't get on planes and knives don't get on planes. That's, that's their core function. But actually by using the right information to deliver great experiences for passengers that are going through them, we know that actually they have a much better experience in the departure halls and and are more likely to to drive better revenues for us as well.

So there's a huge amount of work going on in this space. One of the challenges that we have had historically is we were using different systems to provide real time data and to provide batch data for for after the event. And the reality of having business logic in two different places around that meant that you'd always see inconsistencies around the actual numbers that were coming out when you actually looked at them in retrospect, which led to mistrust um by the operation and the data that we were producing because the advances now that we're seeing in the speed and pace that things like s3 can handle in the aws environments. We're moving all of our business logic into a single data pipeline, whether it's for real time, for services or whether it's for batch services, we're bringing all of that stuff together. So what, no matter when you look at the data, it always looks consistent. Um and again, that's had a real trust uplift for us in terms of the results that we're seeing from it. So the data teams are doing brilliantly in terms of driving that stuff forwards.

So what we do next, you know, we're not, we're not wrestling our laurels around this. When I took on the transformation role, i never saw it as a time band program. The whole point of moving to a product management approach around this is that the difference between a project and a product is products never die. It's about partnering with the business and cross functional teams. It's about building product management capability in our world that brings together people that think about not just technology services, think about the people and process, that's driving those products in the right way, that partner with the business, whether that's our operational colleagues, whether it's our commercial colleagues, whether it's our people in engineering and assets to create cross functional teams that stick together and stay together, that are constantly building and developing the road maps that are, that are driving the business forwards in terms of taking advantage of best technology, continually improving our processes and continually delivering best value for everybody.

So the things that we're now focusing, focusing on is continues to continuing to expand that integrated planning capability. So today, we have a single plan that brings in the flight schedule, as i said before, from 360 days out right the way to the to the on the day of operation. We've now started feeding into that all of our rosters for all of our staff. So we're constantly operate, optimizing our rostering and our staffing levels based on what's happening with that plan in a single place. The next step is we're just putting live and we've started putting it live in manchester. Um dynamically planning the airfield in real time. As i said before, you used to be a guy in a tower with a red book that says that easyjet don't like parking on stand 42 because they get the starbucks queue gets in the way for them. Well, now all of that's taken away, we're dynamically optimizing it in that same system that we're integrating to. And what comes next for me isn't that is starting to build in our planned and reactive maintenance schedules as well. So when we need to service the x ray machine, we know we can't use that one today. So what impact is that going to have on the queues or if the x ray machine fails? This is the next best thing that we need to do in that security of the hall to minimize the impact on our passenger experience.

Um and that, that, that integrating planning capability. I'm, i'm really, really excited about we're digitizing our airfields front to back. So these things have traditionally been the domain of the engineering teams. We'd have a ground. Um the runway, ground lighting will be provided by one set of people, the stand turnaround equipment by somebody else, the radar, by somebody else, the work that we're doing by products. This is actually bringing all of that thinking together into that end to end process, to digitize it. And that's what's going to get us our, our additional 15 based aircraft during our absolute busiest hour.

I think today in manchester, we do about 42 movements in the peak hour off the runways and we're going to deliver another 10 by actually digitizing this just by creating more efficiency in terms of how do you schedule those aircraft coming in and how do you line them up in an efficient way to be able to cross the airfield without um without directly queuing up. As soon as you see a queue of airline waiting to queue of aircraft waiting to get on the runway, they're all burning jet fuel, they're all taking time. It's all costing them money, but it means that we're not optimizing the use of that air of, of those runways as well.

So by digitizing it and the, the image that you can see of the digital twin that we're in the process of rolling out of that runway environment will allow us to go and deliver significant value and more efficiency from the assets that we've already got. And then the last one in there is all about dynamic flow management. So this is thinking about in the same way that we're digitizing the airfield. This is about thinking about how do we optimize um the that passenger flow that passenger journey through our terminals.

I've started thinking about the stream of passengers moving through almost a little bit like a sound wave. You know, it's a constant stream of passengers coming through from the front of our facilities onto the gates and onto planes and in parts they bunch up and it makes it very difficult for them to move through in terms of the experiences because queues form or um we get busy in certain parts of the departure halls. We know that when you get to a certain occupancy level in a departure hall, no one likes standing up anymore because they worry that they're not going to get a seat. So everyone stays sticking down or if they're not getting the right information, everyone stands around the flight information screens trying to work out when the gate is going to open and where they need to go to.

So we're starting to look at things like dynamic call to gate. So actually when we know that the occupancy of the hall is going to get to a certain threshold, that means that the, the um nps scores are going to start going down, that people are going to start having a nice less nice time and that our retail yields are going to be impacted. We can start calling those passengers to gate that are the best ones to, to pull out of there to keep it into a pleasant environment. Um so that we can balance the occupancy levels in there. But in doing that, we know it might take 20 minutes after us saying it for passengers to actually notice, get their bags together, get their family together and start moving to their gates. So because we've actually built this into a single plan that we can control our rosters, can we actually allow the queue to build up in security from what might be running at five minutes to 10 minutes to just slow people down a little bit, be getting into that departure lounge um to allow it to empty out, to make sure that everyone is having the right experience. And where i was talking about that wave where it bunches up in places, it's about spreading it out. So you have a nice even flow through every part of the terminal.

We'd have never been able to do anything like that before. With the silent processes, with different people with different systems without having that single visibility and without the power of artificial intelligence to be able to really go and say, actually, we can see all this happening and this is the best action for you to go and take and automating that entire process.

So if we can get this stuff right, you know, it, it's absolutely game changing to us and i can see a path for us to making it happen. So there are three of the big programs that we're that we're really um motoring on over the course of the rest of this year, i think, you know, for me, an airport is absolutely, it's a, it's a city in its own right as a technologist and i'm a geek at heart, you know, i get to play with, you know, a baggage system is a, is a manufacturing plant. The departure lounge is a, is a shopping mall. Um and that's before you get to the planes and the security services and the border and all the other things in there. But it's all just a big data problem. There's many moving parts, there's many events being created. If we can harness that, if we can pull it all together and really look at how we optimize and drive it, then actually be delivers, delivers benefits that for every part of the ecosystem, for ourselves, in terms of revenue for our customers, in terms of the experience that they get traveling through our airports and through our airlines, in terms of the efficiency of their operations and the value that they get out um of what we're doing.

So um yeah, it's an exciting time for us. Thank you. Thanks nick. You know, whenever i talk to nick, i always find it just fascinating, the multitude of systems and services that are required to run an airport. And you can really see why you need artificial intelligence to create that situation awareness for the operators and to create that personalized journey. And it's not just to create the information nick is using artificial intelligence to automate how the systems and services can respond to things like delays and interruptions in the airport. But i love nick's ambition to be the world's most intelligent airport and you can see from nick session that they're well on their way to achieving that vision. So mark nick, thank you so much for sharing how artificial intelligence is improving the travel experience before you go.

I know everyone's really keen to rush off and book their two week vacation flying out of manchester airport in their upgraded hotel room. But before you do that, we've got two chalk talks this afternoon, both in the, in the mgm grand in here at one o'clock. It's how to optimize the travel and hospitality operations with digital twins and at four o'clock, how to build the digital concierge with gene of a i. Uh please do complete the survey. It's a great way for us to understand and get feedback so we can improve. That's also an awesome way for you to say thank you to mark and nick by giving them five stars in the uh in the response. So thank you all very much.