In its March 2026 Airline Tech Innovation Radar, aviation data firm OAG observed that agentic AI in travel had crossed a threshold from concept deck to live infrastructure, pointing to the partnership between Sabre, PayPal and MindTrip, announced in February, that aims to deliver the industry’s first end-to-end conversational booking pipeline by Q2. The same month, the Airline Distribution 2026 conference in Barcelona convened around a question that has come to define the year: will artificial intelligence restructure airline distribution, or simply coat existing systems with a layer of optimisation? The vendor side is betting on the first answer. Sabre launched its Continuous Revenue Optimizer, described as the industry’s first AI-native, classless revenue engine, in October 2025; Amadeus has been deploying AI filters across its distribution stack; Accelya, PROS and Flyr continue to push competing platforms.
Behind the product launches sits a less-discussed gap. According to OAG’s 2025 analysis, roughly 80 percent of IATA member airlines now apply some form of dynamic pricing — but the level the same report calls “truly dynamic”, where AI integrates real-time shopping behaviour, is reached by under five percent of carriers. The distance between adoption and capability is not closed by better algorithms. It is closed by a layer most coverage does not see: the tariff architecture beneath the algorithm — the rules, classes, route logic, and IATA-compliant structure that determine whether an AI engine has anything coherent to optimise.
To explore where AI-driven pricing ends, and human expertise begins, we spoke with Dina Zhumassultanova, an airline pricing and tariff architect specializing in ATPCO fare systems, interline e-ticketing, and global distribution. With over a decade of experience across multiple carriers in Central Asia, the Caucasus, and beyond, Dina has led integrations enabling ticket issuance across more than 100 countries, designed fare architectures compliant with international IATA standards, and led interline integration projects with partners including Hahn Air Lines GmbH and Sirena-Travel JSC. Currently serving as Airline Pricing & Tariff Architect at Qanot Sharq Airlines while simultaneously consulting for Georgian Airways, Dina manages fare architecture remotely — a working model that reflects the global, systems-based nature of her field.
Dina, thank you for taking the time. The question framed at Airline Distribution 2026 in Barcelona — whether AI is genuinely restructuring airline distribution or simply layering optimisation on top of what already exists — has dominated industry conversation this year. From your position inside the system, how does that question actually look?
It’s a useful question to be asking, and I think the honest answer is that both things are happening at once. There is real restructuring in the commercial layer — revenue management has moved from periodic analyst review to real-time decisioning in a way that genuinely wasn’t possible five years ago. But there is also a lot of sophisticated optimisation being layered on top of infrastructure that hasn’t fundamentally changed. The architecture beneath — how fares are structured, published, and distributed through ATPCO and the global distribution systems — still operates on the same foundational logic it has for years. The speed and intelligence of what sits on top has accelerated; the foundation has not. That distinction matters a great deal when you’re trying to evaluate what AI is actually delivering.
For someone who doesn’t work in aviation, it can sound like pricing is already largely automated. How accurate is this picture, and what does it leave out?
It leaves out the most important part. When people see the price changing overnight, it is the output of a revenue management system responding to demand signals. That part does involve algorithms. But the price itself is only the tip of the iceberg. Before any of it can happen, someone has to build the fare itself: define its structure, set the rules governing when it applies, how it can be changed or refunded, which markets it’s available in, and which booking classes it belongs to. That tariff architecture is built by a person, not an algorithm. The algorithm optimises within a framework that a human designed.
You’ve described a fare not as a price but as a structured set of rules that take into account booking classes, route combinations, seasonality, refund conditions, and other factors. In the current aviation market, these rules have to behave correctly across dozens of global systems simultaneously. What structure needs to be established before a revenue management algorithm ever touches it?
The foundation is the tariff architecture itself — built in compliance with IATA standards (the International Air Transport Association’s global framework governing how airline tickets are priced and sold) and published through ATPCO, the Airline Tariff Publishing Company, which distributes those fares to booking systems worldwide. That means defining how the fare behaves at every level — from booking classes and route logic down to seasonal restrictions and the conditions under which it can be sold. Every element has to be specified precisely, because this isn’t a system that tolerates approximation. Once that’s published, it flows into global distribution systems, into agency channels, and into booking platforms across dozens of markets. The revenue management layer then works with what’s already there. People often assume the price comes first and the rules are added to back it up. It’s actually the opposite: the architecture has to exist before the algorithm has anything to work with.
Building that kind of architecture clearly requires a very specific set of skills. You’ve trained through IATA courses in the Netherlands and Singapore, and completed ATPCO’s certification program in the United States. What was that experience like, and how did it shape the way you work?
It’s worth mentioning that there’s no established path into this profession. Unlike law or finance, there’s no standard curriculum and no obvious career ladder. Most practitioners arrive through experience — working at an airline, encountering the system, gradually going deeper. The formal training helped me build a rigorous foundation for the work I was already doing in practice. That’s probably the most common route into this field.
The IATA training and ATPCO certification gave me the foundations that most practitioners in this field don’t have formal access to — particularly in the CIS region, where this kind of training simply isn’t available. The IATA courses gave me the international standards framework: passenger fares and ticketing, interline accounting, and proration. The ATPCO program in the United States went deeper into the practical mechanics — how fares and rules are structured, how the system behaves under real conditions. You come away from these courses with a clear understanding of how important precision is when every element you enter has global consequences downstream.
At Qanot Sharq Airlines, you’ve been responsible for building and maintaining the global fare structure across international markets — coordinating across revenue management, IT, and commercial teams to keep pricing consistent across all distribution channels. When you approach that kind of build from the ground up, where do you actually start?
Not with the price — and that surprises people. The first question is always: what is the commercial logic of this product? Once you have that, everything else flows from it — booking classes, exchange and refund conditions, the markets where tickets will be available, and the rules that apply in each. All of that has to be defined before anything is published. You also have to think about who else will sell these tickets — not just the airline’s own channels, but agencies, global booking systems, and partner airlines operating in different regulatory environments and currencies. The architecture has to work for all of them at once. Only once that structure is fully in place does it make sense to think about the price itself.
Unlike a pricing error on a retail website that might affect one storefront, when an airline publishes a fare through ATPCO, it becomes available to travel agencies and booking systems around the world almost instantly. You’ve led integrations enabling distribution across more than 100 countries. How does that scale influence the way you approach the work?
It means there is no such thing as a contained mistake. In aviation, errors in the tariff system are well-documented — business class seats sold at economy prices for hours before anyone catches it, with real financial consequences. That happens because the moment something is published, it’s everywhere. Working across more than 100 countries amplifies that. So the discipline you develop is primarily about building a way of working where you’re thinking through the downstream consequences before you publish anything. You check the logic, you verify the behaviour, you think about how this will look in every system it touches. In this field, that’s not overcaution — it’s just how the work has to be done.
You’ve noted that airline pricing follows the same international standards — IATA rules, ATPCO publication — whether you are talking about a Gulf mega-hub or a US legacy airline. Given that universal foundation, where does AI actually add value in the pricing process?
The standards are universal, but the commercial decisions made within them aren’t. That’s where AI adds genuine value — in the revenue management layer, where the new generation of classless and continuous pricing engines can read demand signals in real time and respond to booking pace, competitor moves, and market signals faster than any team of human analysts could. AI is very good at that kind of pattern recognition and optimisation at scale. What it cannot do is design the tariff architecture that those decisions operate within. The rules, the structure, the compliance logic — that still has to be built by someone who understands both the international standards and the commercial reality of the specific airline.
AI-powered pricing tools are attracting significant investment and ambitious projections right now. From your position working on the layer beneath all of that, where do you see the claims outpacing the reality?
The figures for what AI can do within a well-built pricing environment are real. Where I’d be more cautious is the assumption that the environment is already well-built. In my experience, it often isn’t. The tariff architecture may be incomplete, inconsistently maintained, or not fully synchronised across all distribution channels. AI layered on broken architecture doesn’t fix the problem — it amplifies it. So the gap I see is between what these tools can theoretically deliver and what the underlying infrastructure can actually support. The technology is moving faster than the foundations it depends on. That’s where the realistic conversation needs to happen.
In many technical fields, there is currently a debate about which roles AI will eventually absorb and which it won’t. For fare architecture specifically, where do you see this work standing in ten years?
I think the work becomes more important, not less. What AI is very capable of is processing volume and optimising within defined parameters. What it cannot do is define those parameters in the first place — decide how a fare should be structured, what rules should govern it, how it should behave across markets with different commercial conditions and regulatory requirements. That requires judgment, not just calculation. The tools around this work will change. AI will take on more of the routine monitoring and adjustment work. But the architecture itself — the decisions about how a fare is built and why — that stays with the people who understand both the standards and what the airline is actually trying to achieve commercially.
The industry has spent this year asking whether AI is restructuring airline distribution or simply sitting on top of existing systems. From the layer where I work, those are not really competing answers — AI sits on top of what humans build. That is not a limitation. It is just how the work is divided, and the people who design the layer underneath are not going anywhere.
