Amazon Web Services on how generative AI can streamline operations, empower staff, and deliver more seamless customer experiences

Which did you hear about more in 2023 – Taylor Swift or generative AI? While the two were certainly the hot topics of the year, there are many in the airport space that are wondering if generative AI will be useful to them or is just another buzzy topic. FTE sat down with Bob Kwik, Global Head of Airports and Ground Transportation, at Amazon Web Services to hear his take. Bob has over 20 years’ experience in the aviation industry and works closely with airports like Aeroporti di Roma, Manchester Airports Group, and Singapore Changi Airport, to help digitally transform their operations and passenger experience.

Bob Kwik, Global Head of Airports and Ground Transportation, Amazon Web Services

Several publications, including Reuters, shared analysis of the number of times generative AI is mentioned in earnings calls. In Q2, a third of S&P 500 companies mentioned AI in their calls. Clearly executives across industries are excited about generative AI – why?

Bob Kwik (BK): It is amazing to see how quickly and broadly interest in generative AI has grown. I think one reason was consumers could actually play with ChatGPT and experience themselves how powerful the technology has become.

While a lot of attention has been given to how consumers are using generative AI, I think there is an even bigger opportunity in how businesses will use it to streamline their operations, empower staff, and deliver more seamless experiences for their customers.

I think there is a bit of a mis-perception that the main use case of generative AI is a search engine or a chatbot. What makes generative AI stand out from the Machine Learning/AI being used over the last 20 years, is that these foundation models can perform a broad range of tasks, such as language processing, visual comprehension, and content generation. Companies are excited because these models can be customised to perform functions that are differentiating to their businesses, for example creating report summaries or generating product descriptions, using only a small fraction of the data and compute required to train a model from scratch.

How can businesses, specifically airports, use generative AI to help their customers?

BK: As I noted earlier, when people hear generative AI, the first thing that comes to mind is chatbots. That is definitely a use case for airports.

We are already seeing airlines, such as Delta, are using generative AI to help passengers easily get travel-related information. In addition to passenger use cases, airports are planning to use generative AI to help their staff find answers and respond to complex custom questions about ground transport, airline, policies, travellers’ rights, and more.

Airports can also use generative AI to improve passenger engagement through social listening, sentiment analysis, automation of passenger engagement and real-time language translation.

Another way we’re seeing airlines experiment with generative AI is to minimise disruption management, which airports could also do. For example, United Airlines is using generative AI to give passengers alternative travel options, such as flying into a different city and driving, in order to get to their destination. It could also be used to resolve missing baggage and flight connection: Think of a passenger arriving at an airport and their bag doesn’t make it, or they miss a flight connection. With many airlines operating at the airport, usually served by a local handling agent, it can be confusing to know who to talk to or how to contact the airline. Airports could use generative AI to provide passenger information for all the airlines operating at their airport in a way that’s customised, fast and, relevant for each passenger.

Can generative AI also be used to improve operations?

BK: Absolutely. Airports can summarise day of operations reports, extract key performance indicator (KPI) data, and identify trends. This can also be applied to back-office documentation, like operational reports, manuals, and contracts.

For example, with Amazon Q in QuickSight, AWS’s business intelligence service, users can ask their dashboards questions like “Why did retail revenue increase last month?” and get a response in the form of a mini dashboard that illustrates the factors that influenced the increase in retail spend at the airport.

These are a few examples. I continue to see new and innovative ways our customers are using generative AI and the list of examples continues to grow as our customers explore new use-cases and become more familiar with how the technology can be used in their operations.

What is unique about the generative AI offering from AWS?

BK: Generative AI isn’t a single technology, it’s made available through different types of technology. You can think about is as three layers of technology.

The bottom layer is the infrastructure used to train foundation models, or FMs, and run these models in production, known as inferencing. Even though it’s invisible to end users, training and inferencing can be the most time consuming and expensive part of building and using generative AI. Needs vary based on the type of uses. At AWS we give customers a choice on the actual chips that are used, from NVIDIA to AWS designed and built chips – Tranium and Inferentia.

The middle layer provides access to all the large language models (LLMs) and other FMs you need and to the tools you need to build and scale generative AI applications with them. This is where Amazon Bedrock lives. Companies use Amazon Bedrock to build generative AI applications with choice of LLMs and other FMs. For example, Meta’s Llama 2 is great for assistant-like chat, whereas Anthropic’s Claud is very good for generating creative content. Some, such as Amazon’s Titan Lite, are optimised for cost and speed.

Amazon Bedrock also allows you to use your own business context quickly, easily, and privately. It’s called RAG, retrieval augmented generation. For example, information on your operational procedures that can be used for a natural language query. It’s important to understand that any information you provide to Amazon Bedrock is private and doesn’t get used to train the model and can’t be seen outside your organisation.

At the top layer are applications built using FMs that can take advantage of generative AI quickly and without end-users needing any specialised knowledge. Companies like, Accor, Delta Air Lines, TUI, and United Airlines are just a few of the travel and hospitality customers using generative AI on AWS. Amazon have developed our own applications such Amazon Q, our generative AI powered chat bot assistants that securely and privately works with your data and information, and Amazon CodeWhisperer, which helps software developers build applications faster and more securely by generating code suggestions in near real-time.

Is there a way for me to try out the technology?

BK: We’ve built Amazon PartyRock for anyone to build their own generative AI application. It’s incredibly easy to use. Just describe what you want to build and PartyRock will create it. For example, in the PartyRock app builder enter the description “What I need to pack based on my travel destination” and it will generate an app with a field to enter the destination and the resulting packing list. You can then play around with it, for example adding seasons or time of year as a criteria, or add a generated image of the destination.

How has security been built into AWS’ generative AI services?

BK: As with all technology, security and privacy are paramount, and generative AI introduces new considerations. Built with enterprise-grade security and privacy in mind, Amazon Bedrock makes it easy for customers to protect sensitive data. Your content is not used to improve the base models and is not shared with third-party model providers – protecting customers’ valuable intellectual property.

How does being on the cloud enable airports to get more out of generative AI?

BK: To get the most value from generative AI you need to build it into your existing applications and business processes. It’s much easier to do this if the applications already run on the cloud. To get high-quality outcomes from generative AI you need a strong data foundation. That data is used to fine-tune the models to work for you, so it needs to be up-to-date, complete, accurate, discoverable, and available when needed. The most effective way to store, analyse, and process your data is by using cloud data services. This is similar, for example, to how airports are using data on AWS to forecast demand on airport resources.

What is your advice for any airports exploring how they can use generative AI?

BK: The best way to start is to learn a bit more about the technology. We have a 10-minute Generative AI for Executives and a 90-mintute Generative AI Learning Plan for Decision Makers learning plan.

For airports who want to build AI solutions we have AWS Skill Builder classes for AI, ML, and generative AI that can help their tech teams get up to speed. If airports prefer to work with a partner there are many Travel and Hospitality competency partners that have airport experience and their own generative AI practices, including Accenture, DataArt, Slalom, and TCS.

Bob Kwik is the Worldwide Head of Airports and Ground Transportation for AWS. His role is to support customers on their cloud adoption journey. He brings over 20 years of experience in travel technology. Prior to joining AWS, Bob worked for leading travel technology companies, and held regional and global leadership roles in sales, business development, technical design, and product creation. He has lived and worked in Europe and the USA, and has travelled extensively for business and pleasure. He holds a Masters degree in Engineering from Trinity College Dublin.



Comments are closed.