Travel executives are confronting a paradox: agentic AI promises automation and efficiency gains, yet it is already pushing up technology costs across distribution, operations, and infrastructure.
The Upfront Costs Are Real
At the Skift Global Forum in September 2026, leaders from Amadeus and Spotnana noted that agentic AI is adding concrete expenses to travel distribution before the industry has determined who will foot the bill or how value will be captured. Elena Avila of Amadeus observed that costs are here now, while revenue offsets are not guaranteed. Steve Singh of Spotnana suggested suppliers or even travelers might eventually pay if the benefits materialize.
Infrastructure and Talent Demands
Legacy airline and hotel systems were built in silos over decades, making seamless agent-to-agent interactions difficult. Companies must overhaul these systems, publish machine-readable offers, and implement standards for interoperability. Talent acquisition adds further pressure: attracting top AI engineers from big tech firms carries a premium, and running sophisticated AI systems at scale is not inexpensive.
Survey Signals on Adoption and Impact
McKinsey and Skift research shows 90 percent of travel organizations use generative AI in some capacity, yet only 2 percent report widespread agentic AI deployment. While some executives report productivity gains and modest cost savings from broader AI use, the shift to fully agentic capabilities requires new governance, cloud-scale infrastructure, and process reinvention.
Opinion: Short-Term Pain, Long-Term Questions
The industry’s enthusiasm for agentic AI is understandable—autonomous agents could handle complex itineraries, rebookings, and personalized offers with minimal human intervention. Yet rushing to build the plumbing risks inflating tech budgets without clear ROI timelines. Fragmented systems already drive operational costs; adding agentic layers without addressing underlying interoperability may compound the issue rather than resolve it.
Travel companies should prioritize internal use cases where they control the data and outcomes before betting heavily on consumer-facing agents. Standards and commercial models for revenue sharing and liability must evolve in parallel. Until then, agentic AI looks less like an immediate cost saver and more like a significant new line item on the technology ledger.
