AI in Travel and Hospitality: Where 2026 Really Stands

Travellers already plan trips with AI and startups demonstrate dazzling capabilities. Yet inside most established travel companies, operations remain almost entirely untouched. The gap between the two is the whole story.

By James Lever, Principal Director, CTO Advisory for Travel (UK), DataArt

After a season of conferences – Phocuswright, AWS re:Invent, PUSH UX and Skift’s “Permission to Exist” – here is my blunt read on where the industry actually is. The headline isn’t a new model. It’s this: AI success in 2026 will be driven by data infrastructure, not algorithms. The companies pulling ahead aren’t the ones with the cleverest prompts; they’re the ones who invested in modern data platforms and organisational readiness three to five years ago, and are now moving from pilots to production.

Three paths are opening up. Leaders are scaling from pilot to production. Fast followers are correcting course after limited early success. And laggards are watching the competitive gap widen, often without realising how quickly it’s compounding.

What the winners do differently

The patterns are remarkably consistent. High-impact AI programmes are around three times more likely to have genuine senior leadership sponsorship and accountability for outcomes — not isolated experiments tucked away in a lab. Winners are roughly 2.8x more likely to redesign workflows around AI, changing how work actually gets done rather than bolting a model onto an old process. And top performers are about 1.6x more likely to point AI at growth and new capabilities, not just cost-cutting.

Where AI delivers value today

Plenty of it is real and shipping now. The implementations that work share the same traits: specific use cases, clean data, clear business outcomes and proper governance. The high-impact applications I’d watch are intelligent automation of manual back-office and guest workflows (ROI within months), personalisation that understands trip context rather than skin-deep customisation (early adopters report 7–12% revenue uplifts), dynamic pricing off real-time demand signals, conversational commerce layers for search and booking, and developer-productivity tools delivering 20–30% gains.

There’s a particularly large prize sitting in plain sight for ancillary providers. As one Phocuswright speaker put it, “roughly 25 to 30 million people a day arrive at airports and even less than 10% of them actually have transportation booked.” Ground transfers, lounges and loyalty optimisation are all underserved segments where AI-powered integration with journey data turns a gap into revenue.

The mistakes holding companies back

The failures rhyme too. Treating AI as a technology initiative instead of starting with a business problem. Building on poor data quality and weak governance. Disconnected use cases with no pathway to production. Treating probabilistic AI as if it were deterministic logic. Keeping tech teams as service providers rather than strategic partners. And, most consistently, approaching AI as a technical project rather than a business governance challenge. AI governance is not a technical problem — it’s a business governance issue that needs executive sponsorship, clear policy and defined accountability.

The existential threat: disintermediation

This is the part that should keep distribution leaders awake. AI agents don’t need OTA infrastructure. They don’t aggregate inventory; they aggregate the aggregators, without the operational burden of supplier relationships, payments or call centres. Skift’s scenario, adapted to travel, is uncomfortably plausible:

An AI-only travel retailer launches with no call centre. It promises 10-second refunds, automatic disruption management, zero fees, and visibly lower costs… Within hours it undercuts paid search with agent-driven creative.— Skift, “Permission to Exist”

This is no longer hypothetical. Expedia and Booking.com’s October 2025 ChatGPT integrations are a tacit acknowledgement — a pivot from consumer platform to agent infrastructure. At AWS re:Invent, Visa and AWS announced collaboration to let AI agents “transact securely and autonomously,” with Expedia Group, lastminute.com and others reviewing blueprints. As one AWS VP framed it: “Chatbots answer. Agents act.” Meanwhile, 37% of travellers now say they trust ChatGPT’s answers, and a growing pattern sees people plan in AI environments, then leave to book. The aggregation advantage OTAs built over decades erodes the moment agents reach supplier inventory directly.

“It’s going to disrupt everybody… there’s going to be disintermediation of every channel.”

Timelines vary — five to fifteen years for major disruption, though many believe sooner. Either way the imperative is the same: play offense aggressively. Strengthen the direct relationships that survive channel shifts, prepare inventory and content for agentic distribution, rethink commission economics, and position yourself as an essential service layer that AI platforms genuinely need.

Priorities for the next 18 months

If I had to compress the to-do list, it’s five things. First, data infrastructure — modern platforms, governance, real-time availability. Second, legacy modernisation; twenty-year-old technology cannot support the personalisation and flexibility now expected, which makes offer-order transformation and API-first architecture foundational. Third, organisational AI literacy — train everyone, not just the tech team, because the competency gap between management and developers is quietly slowing everything down. Fourth, structured experimentation: buy out-of-the-box for common problems, build custom only for what truly differentiates you. Fifth, production-ready governance with monitoring, clear data-access policy and human oversight on critical decisions.

The sector picture sharpens this. Hotels most urgently need transformation. Airlines must make rapid experimentation mandatory. OTAs and agencies face the most acute disruption risk and need to rethink commercial models now. Tour operators and cruise lines sit at an inflection point where rich, multi-day guest engagement is a personalisation advantage — if the data foundations exist to use it. And for ancillary providers, 2026 could be a genuine breakthrough year.

None of this is about chasing the next model release. It’s about whether your foundations can carry the weight of what’s coming. The window to prepare is narrowing — and the foundations you build today determine which of those three paths you’ll be on by the end of 2026.