According to new research presented at the Business Travel Show America, only one in 10 corporate travel buyers currently trust AI-generated recommendations for bookings, itineraries or travel policy support.
Similarly, just 10% use AI-powered dynamic pricing insights to guide purchasing decisions, although 30% say they plan to adopt them within the next year.
What travel managers want from AI
When asked how they would like AI to improve their work over the next 12 months, travel managers prioritised:
• Automation of routine admin to free time for strategy (25%)
• Real-time updates on disruptions such as cancellations and delays (23%)
• Stronger data-driven insights to optimise budgets and policies (21%)
• Personalised itineraries tailored to individual traveller needs (14%)
• Streamlined expense management (7%)
• More efficient reporting (7%)
Steve Clagg, Corporate Travel Technology Consultant and judge of the Innovation Faceoff award at the event, said the sector is in a period of “unprecedented evolution” where AI adoption will shape competitive advantage.
He warned, however, that inefficiencies such as manual workflows, fragmented systems and poor compliance oversight remain barriers to adoption. While agentic AI can autonomously handle tasks like rebookings and expense exceptions, and generative AI can deliver real-time personalised updates, Clagg stressed that human input and AI literacy are crucial.
“The success of AI is determined just as much by human input as the data it uses,” Clagg noted. “Corporate travel managers need to invest in their own training and understanding to build trust – only then can they improve traveller experience, compliance and budget management.”
With corporate travel spend recovering and expectations of employees rising, the sector faces a dilemma: balancing the promise of efficiency through AI with lingering doubts over trust, compliance and data quality. The research suggests adoption will rise, but winning over buyers will depend on transparency, training and clear evidence of value.