Main content

Why Amex GBT embraces AI with executive oversight

George Lawton Profile picture for user George Lawton April 4, 2024
Amex GBT has launched an ambitious program for adopting AI across engineering, customer care, finance, and the workplace. However, it is cautiously embracing these new technologies with awareness of look-out areas requiring executive oversight and humans in the loop.


American Express Global Business Travel (Amex GBT), a leading software and services company for travel and expense, has announced a new technology and management initiative to embrace new AI tools across engineering, customer care, finance, and the workplace. 

For clarity, Amex GBT UK operates under the umbrella of Global Business Travel Group, in which American Express holds a minority stake, allowing it to function independently from the parent company. It is a leading software and services company for travel and expense with a presence in over 140 countries. 

It’s important to note that the company did not just approach this as a tech thing. It also appointed new senior management tasked with overseeing AI rollout and managing various risks across specific areas. This includes:

  • Marilyn Markham, the new VP of Engineering & AI Strategy, who will lead the team in fostering the secure, compliant, and scalable adoption of AI capabilities across the organization.
  • Erica Trevino, the new VP of Operations System Strategy & Optimization, focused on using AI to achieve greater efficiencies on the frontlines for global travel counselors.
  • Jake Hautly, the new VP of Finance Systems Strategy & Optimization, dedicated to modernizing finance processes and creating greater confidence in financial forecasting.
  • Neil Kirk, the new VP of Technology Services, who will oversee initiatives for staff productivity with AI components.

Markham says their goal was to intensify their focus on current AI projects to drive innovation and efficiency while also bolstering governance of AI adoption: 

Responsible innovation is really at the core of this offering and of our overall approach to AI. As a former bank holding company, we are held to extremely high standards of oversight, so we are ensuring our legal, privacy, compliance, procurement, and cybersecurity experts are actively engaged in learning how AI changes the landscape for their specialty and are empowered to advise the organization as we move forward in a world where AI is prominent. 

As we innovate and transform our offering with AI, we remain committed to the safety and security of our global customer base so we will be refraining from exposing them to untested or unreliable AI technologies until thoroughly vetted and proven. Recognizing the evolving legal, privacy, compliance, and cybersecurity landscape surrounding AI, we remain vigilant in staying informed and proactively mitigating associated risks.

Purposeful innovation

Markham explains that Amex GBT decided to organize the AI initiatives around specific business drivers such as traveler care, finance, engineering, and the modern workplace to ensure it leads with the desired business outcomes. As a technological capability, AI has captured people’s imagination through popular services like ChatGPT. It was important to stay focused on tangible business opportunities while also staying open to new tools that can help get there. 

It is starting with AI as an assistant to employees already performing the tasks that need to be done. But issues like hallucination need to come with checks and balances to avoid creating new problems. Markham says:

It is important for us to keep a knowledgeable human in the loop to help us ask AI the right questions, review the output from AI, and provide a feedback loop for improvement. Where we are invoking AI in a programmatic process, we insert a human checkpoint to review and correct before it reaches our customers.

With that in mind, the firm is also starting with internal uses first. Immediate benefits include quicker responses to inquiries, more time for consultation with travel experts, and smoother interactions. Meanwhile, knowledgeable employees can intervene when problems happen and correct them before they impact the customer. Once it has proven the technology, it plans to integrate some of these features into customer-facing products. 

Building on an existing foundation

It's important to note that Amex GBT is not approaching new generative AI opportunities fueled by DALL-E fantasies from a dead stop. It already had a running start with prior efforts to develop an AI development lifecycle to solve real business problems safely. Internally, it built the Amex GBT Trip Planner, and it also acquired Egencia from Expedia in 2021. 

But Markham sees the new GenAI capabilities as part of a breakpoint with classic AI approaches:

Classic AI (NLP, predictive analytics, and computer vision) revolved around the basic understanding of data in all its forms to produce classification, detect patterns, and produce identification. It extracted what was already there. New generation AI (large language models, image and sound generation) builds on top of the classic capabilities adding the ability to create new data and understand with nuance and context. Taking this notion further, new generation AI is not limited to the dataset it is presented with, it leverages the pre-trained data it possesses to extrapolate, inform, and transform.

In the past, it used Machine Learning to inform rules engines and other sophisticated programming to provide pertinent trip offers, such as Egencia’s Smart Mix and Amex GBT Trip Recommender. Also, their chat team has used NLP tools to build chatbots that help with service efficiency. Expense teams use computer vision to ingest receipts and auto-complete expense forms.

One specific area where gen AI shows promise lies in bridging the language data gap. A good example is responding to new trip inquiries by email. Amex GBT is one of the few travel management companies that owns sits online booking tools. This means it has the technology to produce a trip offer that is in policy, pertinent, and practical. However, this process requires a structured request. Large Language Models (LLMs) can help understand the email, extract the relevant data, and transform it into the API format needed to query the booking engine. 

Markham argues this can also free up staff to spend more time with customers:

In this scenario, the human aspect is also important. Travel counselors who spend time understanding, then researching, and building itineraries will soon be able to focus on curating the proposal. This will also free them to be more consultative in live scenarios.

Caution required

However, caution is also warranted because the new AI tools can break in novel ways. Markham thinks ‘look-out areas’ is a better term than ‘challenges’ for these new cautions. Look-out areas it has identified as requiring attention and mitigation include privacy, quality, and scale. She explains:

AI’s creativity is not conducive to predictability and consistency. This is problematic when integrating AI for programmatic use. We find the need to test extensively and continuously to ensure the same question yields the same answer. We also find it necessary to insert post-processing to eliminate undesirable variability and monitor the quality of the output.

Amex GBT has its roots as a bank holding company. This means it has already developed a culture for maintaining the highest standards of data privacy, compliance, and governance. This is paramount when considering putting data into AI applications. 

My take

The US Government recently announced that all departments would now be required to appoint a Chief AI Officer. This is certainly a step in the right direction because AI, and particularly the new gen AI tools, introduce a host of legal, ethical, privacy, and workplace issues that need to be addressed across specific domains. 

Amex GBT’s example of appointing AI leads across engineering, finance, customer, and employee sides is a smart move. It also has a greater chance of balancing the incredible amount of AI hype against practical opportunities to generate business value safely. 

A grey colored placeholder image