American Express AI tool helps 5,000 travel agents handle complex requests

American Express has deployed generative AI and machine learning across its customer support operations, with its Travel Counselor Assist tool now serving all 5,000 travel counselors across 19 markets. The credit card company credits these AI implementations with improving customer experiences, reducing call times, and increasing travel bookings while boosting its Net Promoter Score.

What you should know: American Express has rolled out two major AI-driven customer service initiatives that leverage both traditional machine learning and generative AI.

• The first project, launched in 2022, uses machine learning to measure customer sentiment on telephone calls with about 12,900 support agents across the US, Canada, Australia, New Zealand, and Mexico.
• Travel Counselor Assist, which began as pilots in late 2023 and expanded to all counselors in October 2024, enables travel agents to use generative AI for trip preparation, itinerary creation, deal searching, and personalized recommendations.
• The generative AI tool also summarizes interactions between travel counselors and customers, streamlining the support process.

How it works: The Travel Counselor Assist tool transforms how agents handle complex customer requests in real-time.

• Travel counselors can tap into generative AI to prepare before speaking to customers, then create itineraries, search for deals, find hotels, and generate customer-specific recommendations for restaurants and vacation activities.
• The system handles incredibly complex inquiries, from securing Taylor Swift concert tickets to planning elaborate multi-destination trips with specific dietary requirements.
• Rather than placing customers on hold for research or requiring callbacks, agents can now solve itinerary requests in real-time, respecting customers’ time.

Why this matters: The AI implementations are delivering measurable improvements to both customer satisfaction and operational efficiency.

• American Express is seeing an increase in travel bookings as generative AI has decreased the time customers need to create their travel plans.
• The company has increased its Net Promoter Score since rolling out the machine learning sentiment technology.
• “We handle many millions of phone calls, and we have machine learning in the background, which gives every single call feedback to the customer care professionals,” says Anthony Devane, American Express EVP and head of global strategy and enablement.

The big picture: American Express designed its AI strategy around flexibility and employee buy-in rather than rushing to implement cutting-edge technology.

• The company built its generative AI solution to work with multiple large language models, avoiding vendor lock-in as AI capabilities evolve rapidly.
• Employee engagement was prioritized through focus groups and encouraging customer support agents to experiment with AI tools.
• About 100 automation ideas are currently being tested, many originating from employees who actively participate in implementation discussions.

What they’re saying: Leadership emphasizes the importance of bringing employees along on the AI journey rather than imposing changes.

• “I’ve always had the same idea — bring people on board the journey versus expect them to arrive at a destination,” Devane explains. “We have to earn the right to win hearts and minds of both our customers and our colleagues.”
• Gary Kensey, EVP and CIO for the Global Services Group, stresses the need for CIO involvement: “If you’re going to get AI right, it’s not about developing a strategy and handing it off to technology and saying, ‘Ok, now tell me where we could leverage AI.'”
• Arun Chandrasekaran, a Gartner analyst, notes that while chatbots and natural-language business intelligence aren’t new individually, “embedding them seamlessly into mission-critical travel workflows at scale is a meaningful step forward.”

Key safeguards: American Express maintains rigorous testing and human oversight across all AI implementations.

• The company uses sandbox environments with redacted customer data to test accuracy before deploying any technology.
• Human oversight is maintained for all automated actions or recommendations, regardless of transaction size.
• Some automation ideas are paused due to high hallucination rates or technical challenges, with the company returning to iterate as AI intelligence improves.