Airlines Built Brands for Humans. AI Agents Don't Care.

Tashkent, Uzbekistan (UzDaily.com) — The aviation industry has built a tractor, and it's using it to haul shovels. The question debated at IATA's 2026 Annual General Meeting was not whether AI will transform airlines — but whether the sector can govern it before it outruns them.

That metaphor came from Nikhil Ravi Shankar, CEO of Air New Zealand, and it landed as the session's defining provocation. "We need to start using it for what it's designed for," he said, describing an industry equipped with transformative tools it is deploying on peripheral tasks.

The panel — drawing together airline executives and AI specialists — ranged across dynamic pricing scandals, agentic workflows, synthetic workforces, and, in its most unexpected turn, the existential question of whether an AI agent can be made to fall in love with an airline brand.

Dynamic Pricing: From Statistics to Mind-Reading

The session opened on the topic generating the most passenger anxiety: dynamic pricing. Megan O'Keeffe, Vice President for Airline Solutions at Amadeus, drew a careful distinction between what airlines have always done — revenue management systems adjusting fares based on historical data and demand forecasts — and what is now technically possible: deep learning models determining in real time which price to show each individual passenger based on their profile.

Andrew Sellers, Vice President of Technology Strategy at Confluent, offered an instructive parallel from hospitality. AI pricing models in the hotel sector, he explained, have learned to probe corporate travel policies: if a company's policy allows accommodation up to $650 in a given market, the system begins pricing rooms at $649.90. "That's not surprising, but it's exactly why it matters" to monitor whether AI systems are genuinely oriented toward broader benefit rather than pure margin maximization, he said.

The core technical problem is interpretability. Previous generations of AI favored models that could explain their decisions. Today's deep learning systems and large language models are vastly more powerful but near-opaque. "We're extremely poor at understanding why they make certain decisions. These systems are inherently stochastic — probabilistic. We simply don't always know what they'll do," Sellers said.

The Non-Discrimination Red Line

Ravi Shankar identified what he called a "red line": non-discriminatory pricing practices must hold regardless of the sophistication of the tools applied, anchored either through regulation or industry self-governance.

The concern is not abstract. Abha Dogra, Chief Product Officer at IBS Software, described a scenario where hyper-personalization becomes ethically untenable: a system with access to loyalty program data and travel history could theoretically identify that a passenger is traveling to a funeral or for a medical emergency — and price accordingly. "If the offer is dynamically formed, hyper-personalization can work against the customer" at their most vulnerable moments, she warned. "How much to restrain that aspect of personalization is not yet calibrated, integrated, or regulated."

Disruption Management: Where AI Already Delivers

If dynamic pricing is contested ground, disruption management is where panelists unanimously agreed AI is already producing tangible value. When a winter storm closes a hub and an airline must simultaneously rebook thousands of passengers, reposition crews, and reassign gates across multiple airports, the problem is, in Sellers' framing, axiomatic: constraints are defined, resources are finite, rules are known. AI handles it decisively.

O'Keeffe identified this as the industry's near-term trajectory — a shift from AI as decision-support to what she called "supervised decision-making with a move to action": systems capable of detecting unexpected demand, updating network planning, and rebooking flights autonomously, but with mandatory human checkpoints embedded in the workflow.

From Chatbots to Agents

Sellers characterized the current moment as aviation "dabbling at the periphery" and identified the leap to agentic AI — systems capable of conducting dialogue while simultaneously coordinating complex, multi-step workflows — as the sector's pivotal threshold.

He offered a case from his own career: early in his time with the US Air Force, generating daily aviation tasking orders required hundreds of man-hours and was considered unautomatable. Today, airport clients perform the equivalent task in fully automated fashion — not because the models became exponentially smarter, but because data governance finally caught up. "As data comes into proper shape, it will unlock more potential and more use cases than we've ever seen."

50% Synthetic Workforce — and Who Governs It

Ravi Shankar looked furthest into the future. He said he now spends significant time on questions that would have seemed science fiction a few years ago: how to integrate an artificial workforce into a human one; what an airline's operational profile should look like if half its staff are synthetic agents; and where the boundary sits between carrier and passenger in a world where AI gives customers tools to independently "manage what an airline is."

On regulation, his position cut against standard industry lobbying instincts. Attempts to regulate the technology itself are probably futile given the pace of change, he argued. The more durable approach is principle-based: non-discrimination, safety as an absolute priority, trust — and allow specific implementations to evolve. He also proposed what may be the session's most unconventional governance model: using AI to regulate AI. "Accompanying everything we try to do with AI with an AI solution that monitors it" — an algorithmic system of checks and balances — has become a practical working heuristic at Air New Zealand.

The Brand Loyalty Question Nobody Can Answer

The session closed on a problem Ravi Shankar's team is actively working on without having resolved: how do you make an AI agent love your brand?

The question is less whimsical than it sounds. As AI agents increasingly mediate between consumers and services — booking travel, managing itineraries, arbitrating between competing offers — airlines that spent decades building emotional loyalty in humans face a scenario where their customers have delegated those decisions to a system indifferent to any carrier. "We've all been building brands designed for people to fall in love with them. We are trying to figure this out, and we don't have the answer yet," Ravi Shankar admitted.

Dogra provided the session's closing note — and its most grounding one: "Caution is good, but excessive paranoia is harmful. This technology is pervasive. It is here to stay."