Airlines are turning to artificial intelligence to address passenger disruption, a challenge that has grown more visible as travelers expect instant updates and solutions. At the Grounded 2026 conference in Reykjavík, senior executives from more than 35 carriers—representing roughly 40% of global passenger traffic—examined how AI could reshape the recovery process.
Airlines’ appetite for AI decision-making
When asked how much authority they would cede to AI if the technology were ready tomorrow, 74% of leaders said they would let the system take some form of action rather than merely offer recommendations. Of those, 39% favored AI operating within predefined rules, while 28% were comfortable with autonomous escalation of exceptions. A smaller group, 26%, preferred AI to stay in an advisory role, leaving final decisions to humans. Only 7% entertained the idea of full autonomy for most passengers.
These figures indicate that the industry is not rejecting AI outright; instead, it seeks a measured handover of decision-making power, bounded by clear parameters.
Barriers to implementation
Despite the willingness to adopt AI, respondents identified significant obstacles. The top concern, cited by 41%, was the incompatibility of existing systems and data quality. Organizational readiness accounted for 27% of the resistance, and issues of trust and governance were mentioned by 18%. Only 10% pointed to the business case, and a mere 4% flagged regulatory hurdles. In total, 86% identified some form of technological, operational or organizational readiness as the primary constraint.
This change is evident in every request for proposal (RFP) written in the last 18 months, which now includes questions about model selection, decision scope, and autonomy levels, indicating that procurement strategies are evolving alongside operational expectations.
Irregular operations present a demanding AI use case because a single disruption can involve operational constraints, inventory limits, accessibility needs, and immigration rules, all of which must be reconciled quickly for thousands of travelers. AI promises to shift the economics of this complexity by moving beyond data interpretation to direct action, such as re-accommodating passengers without human intervention.
When participants were asked to pinpoint the biggest gap between passenger expectations and current airline capabilities, 41% highlighted the speed of information, 27% cited personal relevance, and 25% noted the quality of options offered. Only 8% emphasized the need for human contact at critical moments. The data show that passengers prioritize accurate and timely information.
Speed is becoming the defining metric. Passengers now learn of delays from flight trackers, social media, or messaging apps often before the airline can issue an official statement. As AI shortens the time needed to understand a problem, travelers will increasingly question why resolution still lags behind notification. Improving response times is therefore a key challenge for airlines.
During the 2025 Thanksgiving period, 31 million passengers flew in the United States with a 99% completion rate, a modern logistical miracle. Yet the remaining 1%, roughly 310,000 travelers, experienced disruptions that tested airlines’ recovery systems and highlighted the potential impact of AI-driven solutions.
Projected adoption timeline
Seventy-eight percent thought it would take years before AI could handle most disruption tasks, fifteen percent considered a near-term rollout realistic, and only seven percent believed immediate implementation was likely.
Risk and trust considerations
One central challenge concerns the type of errors an autonomous system might produce. A mis-chosen hotel represents a minor inconvenience, whereas assigning a traveler to a route that violates immigration rules poses a serious problem. When an algorithm acts for thousands of passengers, a single flaw can propagate widely before detection, creating distinct levels of risk depending on whether the system merely advises or is authorized to act.