Frankfurt Airport shows how AI delivers real operational gains
Frankfurt Airport is tracking aircraft turnarounds with computer vision and AI, generating precise timestamps across more than 30 operational events. Leaders from Fraport, Lufthansa Group's zeroG unit, and FraAlliance shared practical lessons from the deployment at the International Airport Summit Berlin.
The shift from guesswork to data-driven decisions is measurable. Before AI systems arrived, operators made turnaround decisions with incomplete information. Now they see exactly when unloading starts, when boarding begins, and when an aircraft will be ready to push back.
From blind spots to visibility
Ground operations teams worked without seeing the full picture. Christian Ritter, Head of Product & Principal Data Scientist at zeroG Lufthansa Group, said operators were "steering highly complex processes like the turnaround without holding the full information into what is happening. This leads to sub-optimal decision-making, which in turn leads to delays and cancellations."
Computer vision changed that. Lena Luftschitz, Project Lead Digitalisation Airside Operations at Fraport AG, said the difference is stark: "Before, we didn't really know a lot about what was going on during the turnaround process or how far along we were. Now we suddenly have a huge amount of data available to us."
The data itself is not the goal. The goal is knowing when an aircraft will be ready to leave the stand. That single prediction lets airports make smarter decisions about gates, stands, and ground crew allocation.
Avoiding data overload
Operational teams already work under pressure. Systems that flood users with every available data point create noise, not clarity.
Frederik Jean, Leader Consulting & Strategy at FraAlliance, said the focus must be narrow: "What we want to achieve is picking out what's not going according to plan." Ground handlers benefit directly. When AI timestamps show exactly when ULD unloading begins, supervisors stop sending crews on best-guess timing and send them when they are actually needed.
This is a workforce productivity gain. It reduces idle time and waiting.
Design around how people actually work
The technology matters less than how it fits into daily operations. Ritter said the team visited ops control centres and asked what information operators actually wanted. "It's not about the tech, it's about how you integrate it into their daily work."
That approach paid off in adoption. Luftschitz shared an example: "One of the turnaround managers told us, 'It was really great seeing what I suggested actually appear in the dashboard I now use.' That really helps with acceptance."
Education matters. Ritter said operators need to understand how systems work and what they cannot do. "People are still the decision makers. We just want to give them a tool, and they need to understand the tool to make the most out of it."
Building trust in a monitored environment
Introducing cameras and AI into airside areas raises legitimate concerns. Workers worry about surveillance. The team did not dismiss these fears.
Ritter said the solution was clear boundaries: "We made sure we don't detect individual people. We track the process, not performance." Engagement with workers' councils and strict data protection measures were essential from the start.
Jean added that transparency about the system's purpose builds trust. "If you show transparency from the beginning and make it clear this is a supporting system, not a replacement for humans, you can really gain trust." That trust-building does not end at launch. Ritter said the team still collects feedback and updates the system to keep it relevant.
Collaboration between airport, airlines, and handlers
Frankfurt's progress accelerated because three parties aligned around shared goals. Jean said the shift from individual optimization to shared data and metrics was critical. "In the beginning, everyone optimises their own systems. What you need is a common vision, shared KPIs and shared goals."
Data sharing is the hardest step. Once trust is established, momentum builds. Luftschitz said the benefit is simple: "Airlines, ground handlers and service providers all benefit if everyone has the same information."
Start small, expand carefully
Safety-critical operations cannot tolerate risk. The team used proof-of-concept pilots and incremental rollouts. Ritter said the approach is necessary: "We cannot just stop operations for a day. So, we keep the risk low at the beginning, start with a pilot, then gradually expand."
Fallback processes remain in place so operators can revert to existing methods if needed. This staged approach also helps with regulatory requirements. Jean noted that GDPR compliance is "not a showstopper, but a hurdle" when addressed in parallel with technical development.
What the data shows
Partial visibility is still a win. Ritter said tracking 80% of turnaround events is "much better than tracking nothing." That incomplete picture still eliminates blind spots and improves decision-making.
For operations teams evaluating AI for Operations, Frankfurt's experience shows the path is not about futuristic automation. It is about visibility, collaboration, and trust. Understanding how to turn raw operational data into AI Data Analysis that supports human decision-making is where the value sits.
The practical lesson: be brave enough to start, small enough to learn, and collaborative enough to scale.
Your membership also unlocks:
