Passenger Terminal World • Passenger Terminal World January 2026 • Change agents

Airports are moving from traditional to generative AI assistants as part of a broader shift toward agentic AI

The release of OpenAI’s ChatGPT in late 2022 was a watershed moment in the development of artificial intelligence, according to Verena Dollberg, AI lead at Fraport, which operates several airports worldwide, including Frankfurt in Germany.

“We had been working with AI long before ChatGPT came along, but its arrival suddenly put AI on everyone’s agenda,” says Dollberg. “You finally had management awareness, and everyone wanting to do something with AI.”

One consequence of the emergence of so-called generative AI solutions – such as ChatGPT – has been an improvement in public-facing chatbots that convey information to passengers. Frankfurt Airport’s virtual assistants comprise off-the-shelf solutions from a range of software providers.

“The market is still rather new,” continues Dollberg. “So far we haven’t decided on a single solution but are trying out several solutions for different areas. In the end, it’s better to have one provider, but we’re still in the experimental phase, where you have to wait and see which product emerges as the best.”

Earlier this year, Istanbul’s Sabiha Gökçen Airport launched Savvy, a generative AI digital assistant for passengers. Based on OpenAI and running Microsoft Azure’s large language models (LLMs), Savvy handled around 5,000 queries from passengers in its first six months of operation, according to Ismihan Baysal Anderson, the airport’s IT and automation director.

“If you think about classic chatbots, you are guided to different boxes,” she says. “If you want information about flights, you select the ‘flight’ box. If it’s about parking, you select the ‘car park’ box and you are guided to further boxes where you can ask questions. But Savvy is not like that. You say, ‘Hello,’ and then you can have a conversation with it.”

Savvy was launched as a proof of concept (PoC) in November last year and rolled out to the public in March 2025. To illustrate the chatbot’s usefulness in airport operations, Anderson points out that although the 24/7 customer service center handled twice the number of queries as the AI during the first six months, it required 12 human agents to do so.

Anderson notes that the generative AI assistant does not replace the work of customer service staff. Rather, it is helping Sabiha Gökçen reach a new customer base that might not ordinarily make contact with the airport.

“Customers who prefer to communicate in writing, usually those who don’t speak English or Turkish, are translating their questions and posting them to Savvy and getting responses,” she says. “So Savvy is increasing our interaction with passengers by opening up this new channel of communication.”

The types of questions being posed by passengers are also helping the airport refine the AI’s functionality, says Anderson. A common query, for example, is the cost of airport parking. “We are integrating Savvy into our car park system so it can run a calculation,” she says. The chatbot can also access data from the airport’s management system to provide passengers with live flight information.

The airport is working on a WhatsApp integration, but there are currently four other ways for customers to interface with Savvy: via the airport website and mobile app, and via QR codes and flight information screens throughout the airport.

The implementation of public-facing AI tools such as Savvy brings with it a range of challenges, including public trust. Anderson hopes the integration of the tool into the airport website and on information screens throughout the terminal will help improve consumer confidence in the software. “People tend to take things more seriously when they see them used in official places,” she says.

Another problem with public-facing AI is its capacity for misuse. To guard against this, Savvy has been trained only to answer questions relevant to the airport experience. “There are restrictions on how it can respond,” reports Anderson. “It’s not going to talk to you about politics or give you a soup recipe.”

In the interests of cybersecurity, the chatbot does not have direct access to airport management systems. Instead, it accesses data from these systems via other airport programs.

Fraport has also had to set security controls around how it interfaces with generative AI. In 2024 it set up FraportGPT, a company-secure LLM available to Frankfurt Airport staff. The tool is meant to prevent the company’s sensitive data finding its way into the cloud and out of the airport’s control.

The move from traditional to generative AI assistants is part of a broader shift toward so-called agentic AI – autonomous systems that can undertake complex tasks with minimal human oversight.

Besides customer service, another area of airport operations where agentic AI has the greatest near-term potential is in logistics, according to Enrique Gonzalez Serrano, head of applied innovation at Aena, which operates major airport hubs including Madrid-Barajas and Barcelona-El Prat.

“In logistics, automated guided vehicles and drones already operate with a certain degree of autonomy outside airports, and Aena is working intensively to deploy them in the management and operation of its airports,” says Serrano.

The airport operator has set out a roadmap for the rollout of AI in five phases, with the final phase involving “the implementation of autonomous agents capable of observing the environment, making decisions and acting without human intervention.”

In the case of airport delays, Serrano envisages future AI agents that can “automatically redistribute resources, inform passengers and adjust flight schedules.” With regard to passenger flow, the agent could “optimize routes, waiting times and staff allocation, thereby improving both experience and efficiency.”

He gives the example of a future AI agent responding to road traffic issues, which, although outside the direct control of the airport, can affect airport operations: “Imagine an AI agent capable of monitoring road traffic and public transportation in the city and surrounding areas, identifying a problem with airport access and then interacting with airlines and handling agents to increase resources at check-in when a surge in passenger arrivals is expected.”

A more broad-brush vision of where the technology could lead is offered by Dalila Haidar, Microsoft’s senior industry advisor for travel, transportation and logistics for the Middle East region: “Looking ahead, Microsoft envisions airports evolving into intelligent, self-orchestrating ecosystems powered by agentic AI. Over the next decade, autonomous agents will anticipate passenger needs, optimize resource flows and coordinate across airlines, ground services and air traffic control to deliver frictionless travel. This transformation will enable hyper-personalized journeys – from automated check-in to dynamic boarding – and significantly reduce delays and emissions through predictive scheduling.”

These kinds of use cases do not seem far-fetched, given the extraordinary pace at which AI is developing. However, there are many challenges to overcome before they become commonplace. A key hurdle is bringing together all the different stakeholders required to introduce automation on an airport-wide scale.

In Frankfurt, for example, the airport is working on an AI-based camera solution that uses real-time data to make the turnaround process more transparent and efficient. A collaboration between Fraport and Lufthansa, the solution is called Seer and involves filming every step of the aircraft handling process, with the AI system automatically timestamping each process step.

The timestamped data is collated into a central database that all the stakeholders can access. For the project to be a success, it will require data sharing among many partner agencies – something the airport industry has traditionally not been very good at, according to Dollberg.

“With AI it’s even more important because it’s all about data,” she says. “And the more data you get through cooperation, the better the product you can create for everyone.”

Fraport has also been gathering data to analyze passenger shopping behavior within Frankfurt’s terminals. “We aggregate data from multiple systems and evaluate it using modern analytics platforms,” says Dollberg. “This allows us to identify patterns that help us tailor services and offerings more effectively and avoid operational bottlenecks.”

The collection of passenger flow data relies primarily on stereo cameras, she says. “These datapoints are combined with other information to create a comprehensive view of passenger movements and preferences. Based on this, we can make operational decisions, reduce waiting times and improve the overall quality of the terminal experience.”

Dollberg refers to the process more broadly as “data intelligence,” which she defines as “the ability to efficiently, securely and purposefully use data from various sources to generate actionable insights for operational and commercial decisions.”

She adds, “In short, data intelligence is the key to turning large volumes of data into actionable insights. It helps us streamline processes, allocate resources efficiently and deliver a personalized, pleasant experience for our guests.”

In spite of the challenges, the airport industry is gradually coming around to embracing the inevitable shift to AI. “Airports are not broadly reluctant anymore – they’re cautious, but accelerating adoption,” says Haidar, who offers as proof Microsoft’s recent integration of its Azure-based analytics platform at Sydney Airport, and a project at JFK in New York that has used AI as part of a unified data strategy that streamlined decision-making across operations, finance and safety.

“The hesitation is less about skepticism and more about integration complexity, regulatory compliance and resource constraints,” reports Haidar.

The migration to an AI future is provoking concerns among airport staff, who are naturally reticent in the face of such groundbreaking technology. Dollberg calls this aspect of her job “change management,” and admits that she was surprised to find pushback against AI even at the higher tiers of management.

“We didn’t expect it at the managerial level but we had senior staff telling us, ‘No, it’s too complex, AI can’t do that,’” she says. “So it’s really at every level. You have regular employees who are afraid of losing their jobs, and you have management who are afraid of giving up decisionmaking to a system that is not in their hands anymore.”

Serrano says that Aena’s objective is not to replace staff with AI but for the technology to “act as a complement for staff, allowing each worker to contribute greater added value to airport operations while AI handles repetitive and less complex tasks. This approach ensures a gradual and steady transition to higher-value roles without compromising the quality of human work.”

At Sabiha Gökçen, Anderson says that the threat to jobs is a bigger concern for higher-skilled workers than for their blue-collar colleagues. “Turkey is a developing country, an emerging market, so replacing manual workers with robotics is still very expensive for us to implement,” says Anderson.

On the other hand, highly skilled workers such as IT staff are more at risk from the changes heralded by AI, particularly if they refuse to adapt to them. “It’s important that you learn about it and how it can change the way you work,” believes Anderson. “There are those who are resisting it, but in the future I think there won’t be any place for these people.”

In May 2025, the US Federal Aviation Administration (FAA) published its first guidance on automated guided vehicles (AGVs) – self-driving vehicles designed to be used as part of the ground transportation fleet at airports.

In many ways, airport environments are well suited to autonomous vehicle use, as they are controlled spaces less prone to the uncertainties of public roads. According to Aena’s Enrique Gonzalez Serrano, the Spanish airport operator is looking seriously at the use of AGVs, and has participated “in various international working groups” exploring the technology. He adds that the organization is considering how the vehicles could be used “for logistical tasks on the apron.”

“AGVs combine environmental perception with autonomous decisionmaking for their safe movement,” says Serrano. This last factor is important because the ongoing growth in airport traffic is leading to greater risk of ground collisions, which IATA estimates could cost the industry globally US$10bn a year by 2035.

For its part, the FAA’s latest guidance mandates that for the time being, self-driving vehicles can only be used in “non-movement areas” away from where aircraft are loaded and unloaded.

According to a report published in July 2025 by Singapore’s Civil Aviation Authority, 30% of jobs in aviation could be “redesigned” due to the impact of AI over the next five years. What goes for Singapore surely goes for much of the rest of the world, where the extraordinary potential offered by AI heralds opportunities and threats for today’s airport workforce.

Given these circumstances, Fraport’s Verena Dollberg believes it is important to address honestly the anxieties of workers with respect to this paradigm-shifting technology. “I try to be very honest because as someone dealing with AI every day, I’m also a bit afraid,” she admits. “So I understand. But at the same time, we’re not going to stop it coming.”

A big part of helping airport staff overcome their AI-phobia, notes Dollberg, is helping them identify “the single use case where it can help them.” She recalls an encounter with a senior member of staff who was deeply hostile to AI but was won over after she used it to write a blessing she needed to offer at a baptism. “I love it when people have this moment where they see how they can use the technology to their advantage,” Dollberg says.

In addition to questions over how AI will tranform roles, the advent of AI poses dangers to airport infrastructure, says Dollberg. The ability of hackers to attack critical infrastructure has been heightened by the extra capabilities offered by AI, so, she reports, “We tackle these dangers in our training.”