We want to help you keep up with how brands across the travel industry are exploring - and using - generative artificial intelligence solutions from companies including OpenAI, Google and more. So we're surveying technology professionals at leading travel brands, and we'll be publishing their answers periodically here.
David Thompson, chief technology officer of American Express GBT, is the latest to offer his insights on generative AI and its impact on the travel industry.
We began working with generative AI in … We have been using AI across our portfolio of services for years. Generative AI can be used to create many types of content, images, text, video. If you consider applications used specifically for text and providing written answers to questions, we are actively researching applications that use open source large language models (LLMs) among others to increase productivity for our teams.
Our current work with generative AI is focused on … Applications using LLMs have the potential to enhance, not replace, existing solutions at Amex GBT. There are several potential use cases spanning across internal functions.
Our main effort for the moment is in traveler care operations. The LLM can make sense of any amount of information presented in any unstructured or structured way. In a few seconds it can answer questions across any area involved in a booking and generate trip or approval emails at the same time. This will allow travel counselors to complete many small tasks in a very short time, improving handling times.
Travel policy is often unstructured and nuanced, but LLMs can be used to read and interpret this information and help travel counselors quickly answer users’ questions. For example, fare rules often come in large chunks of unstructured text. LLMs can help travel counselors quickly make sense of it and see how it applies to a user’s travel policy.
The biggest challenge for us related to generative AI is … For us, one of the main challenges of generative AI is the prevalence of “model hallucination” or errors in the answers it provides. These errors undermine the trust of the user, and constantly retraining models to avoid them is costly. This is why we are looking at open-source LLMs that are not trained on a set of data, but work with vector databases, thus mostly avoiding the need to retrain the model.
Moreover, companies, travelers and suppliers don't always have the same priorities, and our industry is constantly changing. Managing the expectations of all these groups in real time requires emotional intelligence that for the moment can only be provided by humans. Innovative companies will experiment with long context LLM with vector databases to seek out the value it could add to managed travel, while recognizing the need to add a layer of human intelligence as well.
For the travel industry overall, we see the most potential for generative AI to … Generally speaking, AI has great potential to personalize the travel experience and bring productivity to travel teams. LLMs go a step further as they can learn from large quantities of unstructured data and make sense of it, it can help applications take over the repetitive, simpler tasks that travel counselors must fulfill and free them up to focus on creative problem-solving and better serving traveler priorities.
One year from now we expect to be using generative AI for … We are prioritizing developments that can increase productivity for our traveler care teams. LLM can help us build veritable co-pilots for travel counselors. They can help across any task or situation, whether it’s managing a routine booking or staying on track while helping a stressed traveler. These developments will help our travel counselors manage large amounts of unstructured information to perform tasks quickly, freeing up time to better serve our clients and their travelers. And as always, we will continue to ensure that our solutions strike the right balance between technology and human interaction and intelligence.