AI and travel predictions 2026-2030: who says what
The travel industry produces AI predictions at a rate exceeding its bookings. Some announce agents will execute a third of sales by 2030; others, that all this is smoke with a good press department. The interesting part is that both camps publish numbers — and most will be checkable. This article lines them up in a table, explains how to read them, and takes on a commitment: to review every year who is getting it right.
The table of dated predictions
| Prediction | Who | Horizon |
|---|---|---|
| Task-specific agents will grow from less than 5% of enterprise applications (2025) to 40% | Gartner | end of 2026 |
| 50% of companies using generative AI will launch agentic AI pilots or proofs of concept | Deloitte | 2027 |
| 30% of travel bookings will be executed by AI agents | IDC projection, cited across industry forums | 2030 |
| Up to 50% of searches will run through AI engines | Emérito Martínez (Jalala/ESADE), March 2026 — projection with no source cited | 2026 |
| A 12-18 month competitive window before agent adoption generalises | converging specialist estimates (Kerwyn Arias, via Hosteltur; Adam Swart, Lighthouse — a vendor) | 2027 |
| Significant corporate workforce changes from AI within 18-24 months | Sébastien Bazin, Accor CEO (June 2026) | 2028 |
And across the aisle, the sceptics
The opposite camp also signs with data. Ewout Steenbergen, Booking's CFO, declared in March 2026 that the traffic arriving from AI assistants is "very small" and "more or less stable", with no real underlying trend — coming from the player with the most to lose, that may be realism or strategic denial. McKinsey supplies the organisational friction: close to 90% of AI use cases remain in pilot phase, and only 2% of travel executives have agentic AI broadly deployed (with Skift, 2025). Thomas Reiner, of Altimeter Capital, formulated in January 2026 the most quoted middle position: OTAs won't die suddenly but suffer a slow structural decline, like the Yellow Pages — permanently damaged, never deceased.
And the figure both camps must explain is the same: extensive use of AI for trip planning jumped from 13% to 30% in one year (Skift Research survey, published by McKinsey in November 2025), while willingness to delegate payment stays stuck at a single digit. Adoption soaring above, trust frozen below: each camp underlines the half that suits it.
How to read a prediction without being fooled
Four filters I apply to everything in that table — and to whatever the industry publishes next year:
- Projection or measurement? "30% by 2030" is a projection: a model with assumptions. "From 13% to 30% in a year" is a measurement: something that happened. The latter always outrank the former.
- Who speaks and what do they sell? The "12-18 months to adopt" window is spread, among others, by technology vendors — it can be true and simultaneously be their sales pitch. Booking's scepticism is also interested: minimising the threat reassures its investors. The incentive doesn't invalidate the data; it obliges you to seek the counterweight.
- Does it have a checkable date and magnitude? "AI will transform travel" can never be verified; "40% of enterprise apps by end of 2026" can. Only falsifiable predictions deserve memory.
- What does the author risk if wrong? Bazin put a deadline (18-24 months) on a statement about his own company: that is skin in the game. An unsigned trends report, none.
The precedent that counsels humility
Anyone with years in the industry will remember that booking without leaving the metasearch engine — "Book on Metasearch" — was announced as inevitable for a decade, until it died quietly in 2026 with the closure of Tripadvisor Instant Booking. Directional predictions can get the destination right and miss the calendar by ten years, which in business is missing everything. My rule, after fifteen years watching "imminent revolutions" reach a tour operator's operations: take the direction seriously, take the calendar with a calendar.
This blog's commitment
Every prediction in the table has a date. When they arrive — end of 2026 for Gartner and Martínez, 2027 for Deloitte and the "window", 2028 for Bazin, 2030 for IDC — this blog will return to them and publicly record who was right, who was wrong and what the errors taught. It is the exercise almost nobody does in an industry that publishes trend lists every January and never reviews the previous January's. It is in writing; see you at the first review.
