Much of the travel discussion in 2026 appears to arrive at the conclusion that artificial intelligence (AI) agents will play a growing role in how people research and book trips.
That conclusion is supported by the evidence, but it captures only part of the change underway. Phocuswright’s 2026 research found that 44% of travelers would book directly within an AI platform, while 40% would allow an AI assistant to handle flight and hotel bookings on their behalf.
The immediate response has been a race to build assistants, trip planners and conversational booking products. By 2030, however, these features are likely to be widely available, limiting the differentiation that access to an AI model can provide.
A more consequential question is whether the systems beneath those interfaces can support what the agent has been asked to do.
An AI assistant may understand that a traveler wants a refundable room near a conference venue. Completing that request still requires reliable inventory and infrastructure capable of booking and servicing the reservation.
Much of the travel industry still lacks that connectivity. For this reason, the next stage of AI adoption will depend on standardized data and systems capable of operating across company boundaries.
Moving from conversation to transaction
Generative AI has changed how travelers express intent. A traveler can describe a trip in ordinary language instead of entering a destination, dates and filters into separate fields. While that improves discovery, an answer is different from a completed booking.
The distinction becomes clearer when AI moves from assisting a traveler to acting for one. Recommending a hotel requires the system to find and summarize information. Booking it requires the system to verify live availability, interpret the rate conditions, identify the correct room and execute payment.
The partnership announced by Sabre, PayPal and Mindtrip in February offers an early view of how the industry may address this. Mindtrip provides the conversational experience, PayPal contributes payment capabilities and Sabre connects the interaction to real-time travel shopping and booking infrastructure.
Hotel distribution remains deeply fragmented
Fragmentation has shaped travel distribution for decades. Each participant has developed its own systems. Connecting a new supplier or distributor can require considerable engineering work before the companies can exchange usable information.
Air distribution has moved toward shared standards. The International Air Transport Association launched New Distribution Capability, an XML-based standard that enables airlines, travel sellers and technology providers to communicate using a common framework.
Hotel distribution has not coalesced around an equivalent framework with comparable industry-wide acceptance. A room can be described differently by the hotel and an online travel agency. Amenities and cancellation conditions may also appear in different fields or formats. Rate parity arrangements can add further complexity when hotels package or name comparable rooms differently across channels.
A human travel agent can sometimes recognize these inconsistencies and ask for clarification. An autonomous system needs structured information that allows it to determine whether two offers are truly comparable.
This becomes particularly important in the “ask and book” model described in research from the NYU School of Professional Studies and Boston Consulting Group. As hotel discovery moves toward conversational requests, properties will increasingly depend on how accurately AI systems can interpret their inventory.
AI adoption is moving faster than operational readiness
Research by HBX Group found that 65% of travel distributors already use AI. Yet many remain at an early stage of implementation, and scaling the technology continues to be difficult.
A company can introduce an AI customer service tool without rebuilding its booking infrastructure. Giving the system authority to change a reservation is more demanding because it must also understand the supplier's rules and update the financial record.
Similarly, the Global Business Travel Association found that data fragmentation continues to limit visibility and performance optimization, while gaps in hotel distribution and technology impede the travel experience.
Improving the model will not resolve those operational weaknesses. An AI system cannot reliably transact with inventory it cannot interpret. It also cannot reconcile a booking when payment and reservation data sit in disconnected systems. Fragmentation has effectively become a recurring tax on innovation.
Autonomous booking requires autonomous payments
Today’s online travel payment process is designed around a person completing a checkout and authorizing the purchase. An autonomous agent changes that sequence because software may initiate the transaction under conditions established in advance by the traveler. Payment networks are already preparing for this model. Visa Intelligent Commerce, for example, is designed to embed credentials and authentication into AI-initiated transactions.
Autonomous commerce will give stablecoins a much larger role in B2B settlement and agent-to-agent transactions, since their ability to operate globally around the clock and support programmable transfers makes them particularly well suited for this.
I expect that, over time, stablecoins will become the default payment rail for transactions initiated and completed by software. In travel, this will allow agents to pay suppliers across jurisdictions while keeping each transfer connected to the underlying booking.
A connected foundation for travel
Shared infrastructure requires cooperation among companies that also compete with one another. That has historically been difficult in travel because each participant has an incentive to protect its own interests.
The payments industry, though, shows how competitors can operate through common infrastructure. Travel needs the same approach. Another isolated operating system would add another connection for the industry to maintain. A connected operating layer would allow everyone to exchange standardized information while retaining control over their own products and relationships.
By 2030, many travelers will begin planning their trips by telling an AI agent what they want. The AI agent's usefulness will depend on whether it can complete the booking safely. As this happens, the companies building AI interfaces will shape how travelers interact with the industry, but the companies modernizing the underlying infrastructure will determine how much those interfaces can actually accomplish.
About the author...
Mathias Lundoe Nielsen is the founder and CEO of
Entravel Group.