AI is transforming travel, but its deployment doesn't depend solely on the performance of the tools. At Compass '26, the event organized by Sabre, two speakers emphasized three concrete conditions: user buy-in, the definition of understandable limits, and the maintenance of human oversight.
The exchanges were reported on July 22, 2026 by PhocusWire. They brought together Isabel Macdonald, described as a senior behavioral scientist at Irrational Labs, and X. Eyee, CEO of Malo Santo, a company specializing in the evaluation, governance and compliance of artificial intelligence systems.
These presentations do not constitute a quantitative study or a measurement of AI performance in travel. Instead, they provide professional insights into the conditions necessary for its adoption and the risks associated with insufficiently regulated systems.
Summary
The adoption of AI is not solely based on technology
According to PhocusWire's report, Isabel Macdonald explained that a powerful tool alone is not enough to guarantee its use. Teams and clients need to understand their role in the process and how the technology contributes to the final result.
His presentation highlighted the importance of involving users in the deployment process. This involvement can concern testing phases, feedback, or defining the tasks assigned to the tool.
The statement remains declarative: PhocusWire does not publish in this article any adoption rates, nor any comparison between several travel companies, nor any numerical results allowing us to measure the effect of this participation.
Presenting an AI capable of everything can create unrealistic expectations
Isabel Macdonald also emphasized the need to give users a clear understanding of a tool's capabilities. In behavioral science, this framework is sometimes referred to as a "mental model.".
Applied to AI and travel, the principle consists of specifying the tasks that the system can perform, how to transmit a request to it, and the situations in which human intervention remains necessary.
Conversely, presenting a solution as capable of meeting all demands can encourage unforeseen uses. This doesn't automatically demonstrate a decrease in trust, but it does increase the risk of a discrepancy between the marketing promise and the product's actual performance.
According to the specialized press, the challenge is therefore not simply to increase the number of features. It is also necessary to define a scope of use that is sufficiently understandable for employees and travelers.
AI agents raise a governance question
X. Eyee's presentation focused primarily on so-called agentic systems. These tools are designed to chain together several actions to achieve a goal, rather than simply producing a textual response.
In the travel industry, such systems could theoretically be used for searching for services, preparing a reservation, processing requests, or executing certain operations. However, the PhocusWire article does not specify which tools Sabre or the companies involved have actually deployed on a large scale.
X. Eyee indicated that companies developing their own AI agent infrastructure must integrate compliance and governance mechanisms. He specifically mentioned security, design flaws, hallucinations, and failures that can propagate between different stages of a process.
These risks were outlined by the speaker. The source article does not document any specific cases of airlines, hotels, agencies, or booking platforms experiencing such failures.
Humans remain responsible for decisions made with AI
The message presented at Compass '26 was not to dismiss automation. X. Eyee instead distinguished between tasks that can be handled by models and responsibilities that remain with people.
Technology can perform repetitive operations, process large volumes of information, and generate proposals. However, defining rules, managing incidents, validating sensitive decisions, and maintaining ultimate responsibility remain the responsibility of the organizations using these systems.
This distinction is particularly important when AI is involved in a transaction. An error in a booking, payment, file modification, or pricing condition can have operational and financial consequences that the company must anticipate.
Human supervision alone does not guarantee reliability
Keeping a person in the process is often presented as a form of protection. However, it does not constitute an automatic guarantee.
To be effective, this supervision requires that employees have sufficient information, defined procedures, and the real ability to correct or stop an action. A purely formal human validation does not necessarily detect an error produced by the system.
The PhocusWire article does not detail the recommended control methods, applicable legal responsibilities, or levels of autonomy that would be appropriate for each tourism operation. It primarily reports the principles outlined by the two speakers.
What this discussion changes for AI in travel
For agencies, tour operators, accommodation providers and technology companies, the discussions at Compass '26 remind us that deploying AI is not limited to choosing a supplier or a model.
It involves determining which data can be used, which operations can be automated, who controls the results, and how an error should be handled. These decisions relate to the organization of the company as much as to its technological strategy.
Team training is also an operational point. Employees must be aware of the available functions, the limitations announced by the supplier, and the cases in which further verification is required.
However, the source article does not allow us to conclude that companies applying these principles will automatically gain a competitive advantage. No financial, commercial, or operational results are provided to support this conclusion.
This debate should be distinguished from Sabre's communication
Compass '26 was an event hosted by Sabre, during which the group's leaders also presented the company's transformation. PhocusWire notes that Sabre covered the cost of its reporter's attendance at the event.
This disclaimer does not call into question the accuracy of the reported statements, but it must be retained in the editorial assessment of the source. Furthermore, the contributions of X. Eyee and Isabel Macdonald should be presented as professional analyses, not as universal conclusions validated by an independent study.
AI and travel: technology advances, responsibilities remain human
The discussions at Compass '26 highlight a factual point: the development of AI agents forces travel companies to simultaneously address issues of usage, safety, compliance and liability.
They don't demonstrate that humans will prevent all errors, nor that AI will replace entire teams. Above all, they show that deployment choices are made by companies and that these companies remain responsible for managing the tools they provide to their employees or customers.
Article published on July 22, 2026 based on information available at that date.
Sources
- https://www.phocuswire.com/news/technology/ai-human-behavior-governance-travel-sabre
- https://www.malosanto.com/
- https://irrationallabs.com/
👉 Access all our tourism reports, statistics and analyses on Tourisme Stats.
➡️ Check out our full report on B2B Tourism to discover the key trends, figures and strategies for 2026.
👉 Stay informed about business tourism trends: subscribe to InfosTourisme Inside .
📘 Also check out our practical pages for tourism professionals .
📩 Send your press releases .
