Why is travel's agentic AI investment stuck in pilot mode?

A report from McKinsey and Skift, based on surveys of 1,002 travelers and 86 travel executives, argues that travel and hospitality is adopting AI quickly in aggregate but has not converted that investment into transformative, function-specific agentic use cases. The report notes that only about 4 percent of companies in the Skift Travel 200 mentioned AI in their 2022 annual reports, rising to 35 percent by 2024, and that travel-industry venture capital going to AI-enabled start-ups rose from about 10 percent in 2023 to 45 percent by the first half of 2025. Despite this surge, the report states that "the vast majority of these more focused experiments have remained stuck in pilot mode," even as horizontal efforts — "enterprise-wide copilots and chatbots" — have scaled quickly but delivered only "diffuse, hard-to-measure gains."

McKinsey and Skift attribute the stall to the structure of the industry rather than to the technology itself. The report describes travel, and especially hospitality, as "highly fragmented, cobbled together in part from countless small to medium-size businesses," with no centralized data ownership across the ecosystem. That fragmentation, the report argues, "limits the network effects and feedback loops that typically accelerate AI performance," making it harder for travel companies to train effective models or deliver personalized, real-time experiences at scale. The report adds a cultural dimension: travel companies "tend to favor investment in human interconnection instead of tech innovation," treating technology as an enabler rather than a core business component, which leaves tech talent and investment lagging.

This sits alongside Why do most enterprise AI pilots fail to deliver returns?, extending the pilots stall pattern to a specific sector and giving it a sector-specific cause: not buy-versus-build workflow fit in general, but data fragmentation and a service-over-technology culture particular to travel's fragmented ownership structure. It also qualifies Does agentic AI succeed through process redesign or workflow layering?: Deloitte's prescription — redesign processes rather than layer agents on — presupposes a firm or ecosystem with enough centralized process control to redesign; McKinsey and Skift's travel data suggests that precondition may not hold across a fragmented, multi-owner industry.

The report is a vendor-commissioned survey (McKinsey with Skift, a travel-industry publisher) rather than independent research, and its percentages describe self-reported executive sentiment and annual-report mentions rather than measured agent performance or outcomes. It does not establish that resolving data fragmentation would be sufficient to move vertical use cases out of pilot mode, only that respondents and the report's authors identify it as a limiting factor. The implication the report draws, at the strength its data supports, is that travel's agentic AI investment is currently better explained by industry structure than by model capability.