AI or CTO? Rethinking Airline Customer Experience in the Age of Agentic AI

I clearly remember my time as Managing Director when I launched Expedia UK in the late-1990s.

The internet had already started to change airline distribution, but the industry was still carrying many of the old habits behind the scenes.

Tickets were still issued on paper.

Customers called to book, change, confirm, chase, correct, and ask for reassurance.

Travel companies were handling large daily ticket volumes through manual workflows, couriered documents, and call-center teams built around human processing.

At that time, we were issuing around 4,000 tickets a day.

That volume could easily generate more than 1,500 service calls.

Every ticket created a trail behind it, a payment question, a delivery issue, a schedule change, a name correction, or a customer who simply needed to know that everything was under control.

That period taught me a lesson I still believe in today: Technology does not change travel simply because it arrives, it changes travel when the work behind it is redesigned.

The internet changed the front end; customers could search and book differently but behind that click, people were still doing the hard work.

Then e-ticketing changed the model; online booking matured, self-service became normal, mobile apps became part of the journey and chatbots arrived to handle basic customer questions.

Each wave removed some manual work, each wave created more scale and helped airlines and travel companies process more passengers with fewer repetitive steps and each wave also brought back the same question: What should technology handle, and what still needs people?

We are asking that question again now, only this time, the technology is not e-ticketing, online booking, or a basic chatbot. It is AI, and more specifically, agentic AI.

And for airlines, the question is not just whether AI can help the call center but the bigger question is whether having CTOs are financially viable and reasonable?

AI or CTO? Airline Customer Experience in the Age of Agentic AI

Before going further, it is worth clarifying what I mean by CTO.

In this context, CTO is referred to the old airline City Ticket Office, the local airline office that once handled sales, ticketing, changes, refunds, trade support, and customer problems close to the market.

For many years, the CTO was the airline’s local face, customers walked in, agents called in, corporates had names they could rely on, and complex issues were handled by people who understood the market as well as the airline’s rules.

As ticketing moved online and call volumes increased, much of that work moved from city offices into centralized call centers.

The customer did not always walk into an office anymore; they called, emailed or used web forms. The airline still had to serve the same needs, but the service point had shifted.

Either call it CTO or Call center, it is the place, physical or operational, where airline service, local sales, trade support, ticketing knowledge, market understanding, and customer accountability come together.

The Airline Customer Experience in the Age of Agentic AI question matters because airline customer experience is not a simple support function or just answering calls, also it is not just sending a refund status; telling a passenger whether a bag is included.

It is where operations, revenue, distribution, payments, loyalty, airports, travel agents, local markets, and human emotion all meet.

The traditional airline service environment

Before we talk about AI, it is worth looking at what airline customer service actually covers:

A normal airline call center or service office is often responsible for reservations, fare questions, booking changes, cancellations, disruption support, refunds, baggage, loyalty, special assistance, payments, trade queries, corporate accounts, group bookings, and complaints.

That is a wide scope.

A reservation call may involve fare rules, availability, ticket conditions, seating assignment, agency ownership, payment status, and schedule changes.

A refund request may involve the original payment method, ticket restrictions, voucher rules, disruption policy, consumer regulation, and the difference between voluntary and involuntary cancellation.

A baggage case may involve airport handlers, tracing systems, delayed delivery, damage claims, compensation rules, and sometimes urgent personal circumstances such as medication, business clothing, or a family event.

A disruption case may involve rebooking, rerouting, hotel accommodation, meal vouchers, missed connections, checked baggage, loyalty status, and a passenger who is already under stress.

This is why airline services are so different from many other industries.

A passenger may contact the airline with one question, but that question can quickly touch five or six departments.

The other issue is volatility.

Airline service demand is not evenly distributed.

Weather disruption, airport closures, ATC delays, strikes, schedule changes, baggage failures, fare sales, refund policy changes, and large group travel periods can all create sudden spikes.

A quiet morning can become a major servicing event by the afternoon.

That is the reality airlines have always had to manage.

They need reliable service, but they also need cost control.

They need speed, but they also need accuracy.

They need automation, but they also need human judgment when the case becomes sensitive.

This is where bots first entered the picture.

New Episode: Airlines started using bots

Airlines did not start using bots because they wanted to remove customer service.

They started using them because the volume of repetitive work was too high.

Passengers ask the same questions every day:

What is my baggage allowance?

Can I check in now?

Is my flight delayed?

Where is my refund?

Can I change my seat?

What documents do I need?

How do I add a bag?

For these questions, automation makes sense.

A bot can answer quickly, it can work outside office hours.

It can collect a booking reference before passing the case to a person.

It can reduce avoidable pressure on the call center and the City Office.

In that sense, bots helped CTOs a lot.

They took some of the routine volume away from local offices and service teams, allowing human staff to focus on more complex cases.

That was the right idea but basic bots came with their own limits.

Most early bots were rule-based, they followed scripts, keywords, and fixed menus.

They worked when the passenger asked the expected question in the expected way and struggled when the situation was unclear, emotional, or unusual.

That is where airline services become difficult.

A passenger may ask, “My flight changed, what should I do?” On the surface, that looks like a simple question but behind it there may be a missed connection, checked baggage, an elderly passenger, a booking made through an agent, a non-refundable fare, a visa issue, or a need to arrive the same day.

A basic bot sees a category. A good service person sees a journey at risk.

This difference matters.

The Air Canada chatbot case in 2022 became an important warning for the industry:

A passenger relied on chatbot guidance about bereavement travel, the information did not match the airline’s actual policy, and the airline was later ordered to compensate the passenger.

The lesson was not that bots are bad but that when an airline channel gives an answer, the passenger holds the airline responsible, that point becomes much more important as we move from chatbots to agentic AI.

Because a chatbot answers, Agentic AI can act.

And now comes the Agentic AI:

What is changing with AI and agentic AI in Airlines?

The next phase is not just better chat, that is the first mistake many companies make as they see AI as a smoother chatbot, a tool that can write better sentences, sound more human, and reduce friction in digital service.

That is definitely not the case, the real shift is from conversation to action.

I had one of those moments recently where this whole AI discussion stopped being theoretical.

I was travelling from Nice to Australia through the Middle East, and it was not one neat ticket from A to B.

It was separate tickets, different airlines, different rules, and the sort of itinerary where you already know that if something goes wrong, nobody is going to make it simple.

Then the US-Iran situation escalated while I was on the move.

And I remember thinking, “Well, this is exactly what you don’t want to happen when you’re halfway through a journey like this.”

The local office was closed, the time zone was working against me, I was tired and my bags were moving through the system, I was trying to work out whether the rest of the trip was still going to hold together.

So I called the airline expecting the usual routine: wait on hold, explain the situation three times, and probably be told that they could only deal with one ticket at a time.

But that is not what happened.

To my surprise the AI-led service picked up the situation far better than I expected.

It did not just look at one flight in isolation but seemed to understand that I was mid-journey, that the tickets were separate, that the route had become sensitive, and that I needed a practical way forward, not just a polite answer.

It helped work through the rerouting options and treated the trip as one connected journey, even though commercially it was not as clean as that.

And that, for me, is where agentic AI starts to become interesting.

Not when it says, “Your flight is delayed.”

But when it understands, “This passenger is already moving, the route is exposed, the office is closed, the tickets are messy, and a decision is needed now.”

That feels much closer to real service.

At the same time, it also shows the risk, if AI is going to help reroute passengers, interpret disruption, and join up separate pieces of a journey, then the airline needs very clear rules behind it. Otherwise, a helpful tool can quickly become a liability.

So yes, that experience made me more optimistic about AI in airline service, but only when the airline has done the hard work behind the scenes.

A large language model can explain, summarize, translate, classify, and draft. It can help a service agent write a better email, summarize a long complaint, translate a customer message, or find a policy faster.

An AI agent can interpret intent, decide which systems are needed, call those systems, recommend a path, prepare an action, and in some cases execute it.

That is a very different level of responsibility.

A language model can explain a refund policy. An AI agent may check the booking, assess eligibility, apply the rule, trigger the refund, update the customer record, and create an audit trail.

One improves communication while the other changes operations.

This is exactly where agentic commerce becomes important:

agentic AI is not simply about answering questions, it is about intent, authority, orchestration, payments, fraud controls, and system connectivity.

For airlines, intent is the big opportunity.

Customers do not think in booking classes, fare families, refund rules, NDC schemas, or interline logic, they think in outcomes.

They say, “I need to get to London tomorrow morning, avoid a red-eye, sit near the front, and keep the fare flexible.”

That sentence contains intent, to let the airline know about timing, comfort, flexibility, and preference.

A traditional booking flow forces the customer to break that request into fields.

Agentic AI can start from the full intent and work backwards into the systems.

But that only works if the airline’s systems are ready.

AI needs access to live availability, fare logic, payment rules, baggage rules, loyalty context, refund policy, disruption options, and escalation paths. Without that, it may sound confident without being operationally useful.

That is why agentic AI is not only a customer service tool, it is an operating model question.

The real challenge is orchestration

The hardest part of airline AI is not writing a good answer but connecting the answer to the right action.

A passenger may ask one question, but the solution may require multiple systems.

Reservations, loyalty, baggage, payment, refunds, disruption handling, CRM, fraud screening, and agency servicing may all be involved.

Agentic AI needs an orchestration layer. It needs to understand the customer’s request, decide what information is needed, call the right systems, apply the right rules, and return a useful next step.

This is where basic automation fails.

A standard API may answer a narrow question but customer intent is often broader than that. Someone may not ask, “What is my fare rule?” They may ask, “Can I still get there tonight without losing my money?”

That question requires judgment, policy, inventory, payment logic, and service sensitivity.

So airlines need to define what AI is allowed to do.

Can it only read information? Can it recommend an action? Can it prepare the action for a human? Can it execute with approval? Can it execute without approval?

Those are different levels of risk.

If leadership does not define these boundaries, the technology may define them by accident.

That is not a strategy.

Payments and refunds change the stakes

As long as AI is only answering questions, the risk is manageable.

Once AI can take payment, issue refunds, redeem miles, rebook flights, offer compensation, or sell ancillaries, the airline enters a very different environment.

At that point, the airline has to think about authorization, passenger consent, payment security, fraud screening, fare-rule accuracy, refund justification, audit trails, and liability. These are not technical details. They are leadership questions because airline service touches money all the time.

Refunds, vouchers, chargebacks, fare differences, ancillaries, local payment methods, loyalty redemptions, and agency bookings can all sit inside what looks like a simple customer service conversation.

This is why governance matters.

A confident answer is not enough.

The airline needs a controlled answer. And when the case is sensitive, it may still need a human answer.

Fraud and policy abuse will also change.

Traditional fraud systems rely heavily on behavior. They look at how a user browses, what device they use, how long they take, whether they compare options, and whether the journey looks normal.

An AI agent may not behave like a normal customer, it may go directly to the product, fare, refund path, or payment point. That can be efficient, but it can also create blind spots.

Bad actors will use AI too. They can test cards faster, search for loopholes faster, manipulate promotions faster, and abuse refund or voucher policies faster.

So agentic AI is not only a service topic. It is also a risk topic.

The answer is not to block AI. That is unrealistic. The answer is to decide which agents are trusted, what they are allowed to do, and when human review is required.

Again, the issue is not technology alone. It is judgment.

Where CTOs still matter?

This brings us back to the City Ticket Office.

If a CTO only exists to do repetitive manual work, AI will absorb much of that work. And frankly, that is not a bad thing.

Experienced local teams should not spend most of their day repeating baggage rules, basic flight status updates, or standard refund timelines. Automation should remove that burden.

But that does not make the CTO irrelevant. It makes its real value clearer.

The modern CTO should focus on the work where local judgment matters: commercial escalation, trade support, corporate handling, group travel, disruption recovery, local-language nuance, market-specific payment issues, sensitive complaints, revenue protection, and relationship management.

This is where AI can support the decision, but should not always own the decision.

A City Office knows when a travel agent question may represent a larger account. It knows when one group issue may affect future demand. It knows when a passenger needs empathy before policy. It knows when a technically correct answer is commercially wrong.

AI can process information. A City Office understands meaning.

And this matters even more because the line between B2B and B2C is becoming less clear. Passengers expect direct digital service, travel agents still need support, corporates expect speed and accountability, Groups need flexibility and local partners need someone who understands the market.

A passenger may buy direct and still need local help.

A travel agent may manage a booking that touches airline digital channels.

A corporate client may expect consumer-grade speed with business-grade responsibility.

This is where Borderless service becomes important.

Not purely trade, not purely consumer, not purely digital, and not purely human. Airlines need service models that can cross channels, markets, and customer types.

And that is why I do not think the future is AI replacing the CTO.

I think the CTO is evolving again.

It started as the City Ticket Office, the place people went when they needed someone local, someone who knew the market, the trade, the airline, and the customer.

Then a lot of that work moved into call centers, because the volume changed and the customer moved from walking in to calling in.

Then bots came in, because too many people were asking the same basic questions, and no serious service model can have experienced people spending their whole day repeating

baggage rules or refund timelines.

Now agentic AI is the next step.

And if it is used properly, it can take another layer of pressure away from human teams. It can understand the intent, connect the systems, prepare the options, and sometimes even solve the problem before a person gets involved.

But that does not make the human role disappear.

It just changes where humans are most valuable.

The best people should not be stuck answering the same simple question for the hundredth time. They should be handling the moments where judgment matters, where a travel agent relationship is at risk, where a corporate account needs care, where disruption becomes emotional, or where the technically correct answer is not the commercially right one.

That is how I see the next model.

AI handles more of the scale.

The modern CTO handles more of the judgment.

And the airline needs both working together, not competing with each other.

Next step is not AI instead of people, it is not a return to old manual models either.

The path forward is role clarity.

Bots should handle repetitive, low-risk questions. Agentic AI should coordinate structured actions where systems, policies, and permissions are clear. Human teams, including modern CTOs, should handle judgment, escalation, relationships, and trust.

Airlines should start by mapping the work. Which tasks are repetitive? Which are sensitive? Which are emotional? Which are commercial? Which create risk? Which protect trust? Which require local knowledge?

Then they should design the service model around those answers.

They should clean their knowledge bases, connect the systems that matter, define AI authority levels, build escalation paths, and measure more than call reduction. Lower call volume is useful, but it is not the same as better customer experience. Airlines also need to measure first-contact resolution, repeat contacts, disruption recovery, refund accuracy, complaint escalation, revenue protection, loyalty impact, and customer trust.

For Anjuna, this is a practical discussion.

The future of airline representation will not only be about selling more.

It will also be about servicing better.

Markets are becoming more complex.

Passenger expectations, trade needs, corporate demands, digital channels, and local relationships now overlap.

A generic call center is not enough.

The future requires human-centered service supported by smarter systems.

That is where models like ASC+ become relevant. (to learn more about ASC+ click here the goal is not simply to reduce cost, that is too small, the goal is to turn service into a revenue-protecting and revenue-supporting function.

Good service protects future sales. Good recovery protects loyalty. Good local support protects trade relationships. Good escalation protects the brand.

The airline industry has been here before. The internet changed distribution, e-ticketing changed fulfillment, self-service changed passenger behavior, mobile changed access, and chatbots changed basic support.

Agentic AI will change the next layer: intent, action, orchestration, and authority.

But the winners will not be the airlines with the most impressive AI demo. They will be the airlines that know where AI belongs, where human judgment still matters, and where accountability must never be outsourced.

A bot can answer, an AI agent can act and CTO can judge.

Airlines need all three.

The future is not AI instead of the City Office but AI connected to a smarter, more local, more human, and more Borderless service model.