Spanish agencies: 46% now use AI, only 4% profit from it

In September 2023, ObservaTUR put the share of Spanish agencies using artificial intelligence at just 7%, almost always a customer-service chatbot. Two years later, in September 2025, a Drimer study put that figure at 46% — and added a second figure that completely changes how the first one should be read: only 4% get a real, measurable benefit from that use. The study doesn't publish the exact sample size, so the figure deserves that caveat, but the pattern —fast adoption, lagging profitability— is the same one documented by other sources with sounder methodology, and that does make it reliable as a diagnosis.

Using it isn't the same as profiting from it, and the gap runs across the whole sector

The gap in Spanish agencies isn't an isolated case. It's the local version of a pattern that repeats, with different numbers, across all of AI-powered travel: Phocuswright (February 2026) finds that more than 60% of travel companies are experimenting with or scaling agentic AI, but only 6% truly scale it and 22% are beginning to; McKinsey documents that 90% of sector executives use generative AI but barely 2% report widespread agentic use. The pattern repeats so often that it's no longer a Spanish anomaly: trying AI is cheap and fast; turning it into a process that moves the bottom line demands strategy, clean data and time, and that's what almost nobody has done yet.

Javier Caballero, a partner at McKinsey, sums it up with a line that applies to any agency, however small: the potential of agentic AI isn't only in automating tasks, but in transforming how the sector's problems get solved. Automating one isolated task —drafting a quote, summarising an itinerary— is what the 46% do. Redesigning the whole process around AI is what, still, almost nobody does.

A nearby example of how it is done is TUI France, with 215 stores. As its head of information systems, Sébastien Botte, explained in an interview published in May 2026, they started with the least risky part: a search tool that understands requests in plain language across their own catalogue (late 2024) and, later, an assistant that compares products and drafts the email to the customer with the options they liked (summer 2025). They tested it first in 15 agencies and then rolled it out to another 200. Booking and payment were left for last: their ChatGPT app only searches, and payment happens on TUI's website. The team is minimal, three technical people and a product owner, and what has taken the most work is not the technology but complying with GDPR. Botte sums it up in one idea: in the end the agent decides, and AI is only a support.

The travel agent isn't disappearing — but their trust advantage is eroding

The narrative that "AI replaces the agent" doesn't hold up well against what travellers themselves say. According to Ávoris Travel Insights (April 2026, 700 travellers), 46.4% believe AI will complement the work of agencies, and another 33.1% that it will replace part of their functions without making them unnecessary. Most prefer AI as support within the agency —42%— and only 14.1% would want it integrated decisively.

The figure that should really worry any agency isn't the replacement one. It's this: the trust travellers place in other travellers' opinions online (33.8% rank them as their first source) and the trust they place in their agency (33.2%) are, in practical terms, tied. And in that same report, only 4.9% put an AI tool first. The picture is more nuanced and, for the sector, considerably more uncomfortable: it's not that AI has won the trust, it's that the agent has stopped having a clear advantage over almost any other source of opinion.

Where Spanish agencies' money is going — and where it's stopped going

The Beroni/ObservaTUR Barometer (May 2026) confirms the symptom from the budget side. Spanish agencies' investment in 2026 is rising in back-office systems (from 16% to 24% of declared technology investment), in CRM (from 15% to 19%) and in web (from 35% to 37%): automation and customer data, the technical foundation of any serious AI project.

And here is the problem that explains much of that 4%: investment in training falls from 59% to 46%, and in hiring, from 30% to 24%. The sector buys the tool and cuts the person who would need to learn how to get value from it. Josep Bellés, of Beroni, sums it up this way: technology is no longer a differentiator, it's a prerequisite. The problem is that a prerequisite without the training to use it well doesn't generate any return — it just adds one more subscription to the expenses column.

A Spanish ecosystem that already exists, even if almost nobody names it as a whole

Behind this 46% there's a Spanish AI travel-tech ecosystem that's rarely told as a whole: Pipeline (API integration), Beroni (forecasting growth of around 45% in 2025, by its own account), Travel Compositor (part of the French group Travelsoft), Grupo Iris (automation), Drimer and Ávoris's CAT-AI. Héctor García, of MOGU, summed it up at UNAV's II Tecno Travel Agency forum in Málaga (May 2026): "now we can all be an OTA" — the democratisation of tools that used to belong only to the big distribution chains.

That democratisation had a very concrete example at that same forum: a live demonstration of a ChatGPT copilot built in half an hour, with no technical knowledge, for an agency's daily tasks — with another workshop rightly warning about the risks that come with that ease: phishing, ransomware and payment fraud, now with AI-cloned voices to make the call believable.

What I would do if I ran a Spanish agency today

In fifteen years inside a tour operator I've seen the difference between "we have a tool" and "we have a process that shows up in the bottom line": a CRM that had to be rebuilt from scratch because the first implementation had failed, and the automation of many very different processes. With AI the pattern is the same, and so is the order: first, clean data (rates, policies, customer history in a single place, not three), then a single task automated from start to finish —not ten isolated tasks half-done—, and only then the training of whoever uses it every day, not the other way round. The Beroni barometer documents exactly the opposite mistake: buy first, train later (or never).

If your agency is in the 46% that already uses AI, the question that gets you closer to the 4% that profits from it isn't "which tool do I buy now?" It's: which complete process, start to finish, am I going to redesign around it this quarter — and who on my team is going to get paid time to learn how to do it well?