NASHUA, N.H. – Artificial intelligence (AI) has attracted growing attention across the aviation industry, but integrating AI into commercial flight operations presents challenges that extend well beyond developing more capable software.

Unlike consumer AI applications, aviation systems operate in highly regulated environments where every recommendation must be understandable and subject to human oversight. For pilots and air traffic controllers, AI may become another decision-support tool rather than a replacement for the people responsible for operating aircraft and managing airspace.

Researchers at the German Aerospace Center (DLR) recently concluded a four-year effort to examine how that relationship could work in practice. Through its LOKI project, DLR worked with aviation companies, research organizations, and public authorities to develop and evaluate AI systems that support both pilots and air traffic controllers. The project also identified best practices for introducing AI into safety-critical aviation environments.

Related: DLR tests drones, AI, and satellite data for disaster response coordination

AI shifts from automation to collaboration

Commercial aviation has relied on automation for decades. Flight management systems, autopilots, and collision-avoidance technologies already assist crews with routine flight operations.

AI works differently from traditional automation. Rather than following only predefined logic, these systems can evaluate large volumes of information, recognize patterns, and adapt to changing conditions. That capability allows them to analyze complex operational situations more quickly than conventional software, particularly when multiple variables compete for a crew's attention.

For aviation, however, those advances also introduce new engineering challenges.

Operators must understand how an AI system reached a recommendation and when human judgment should take priority. Developers must also ensure the software behaves consistently across a range of operating conditions before regulators will consider approving it for operational use.

Those questions shaped DLR's LOKI project. Rather than asking whether AI can perform aviation tasks, researchers examined how people and systems can work together without compromising safety or transparency.

Digital controller supports growing air traffic

One part of the project focused on air traffic management.

Air traffic controllers continuously monitor aircraft positions, coordinate route changes, and respond to changing weather and traffic conditions. As flight volumes increase, those responsibilities become more demanding. This is especially true in busy sectors where controllers may manage dozens of aircraft simultaneously.

DLR addressed that challenge by developing the Digital Interactive Reliable Controller, or DIRC.

Unlike traditional decision-support software that simply displays information or generates alerts, DIRC is designed to function as an AI-supported team member. The system can independently perform selected tasks while coordinating with the human controller on how responsibilities are divided throughout an operation.

Researchers tested the concept through simulations that paired human controllers with the digital controller.

According to DLR, the combined teams managed traffic volumes up to 25% higher than the current maximum workload defined for an airspace sector. Rather than replacing controllers, the system reduced routine workload so operators could concentrate on more complicated decisions requiring human judgment.

That distinction reflects a broader trend across aviation automation. Developers are increasingly designing AI systems to complement human expertise instead of removing people from the decision-making process.

AI helps pilots evaluate time-critical decisions

The project also explored how AI could assist flight crews during high-workload situations.

Pilots routinely evaluate weather conditions, aircraft performance, airspace restrictions, and operational procedures throughout a flight. During abnormal operations, those decisions are often made within minutes while crews continue flying the aircraft and communicating with air traffic control.

To support those scenarios, DLR developed the Intelligent Pilot Assistance System, known as IPAS.

The demonstrator analyzes operational information and presents decision-support recommendations during time-sensitive events, including selecting alternate routes or identifying suitable diversion airports. The system rapidly processes information that would otherwise require crews to review multiple sources independently.

Researchers also examined another important question: what information pilots need before they are willing to rely on an AI recommendation. The answer extended beyond accuracy.

Pilots wanted enough context to understand why the system recommended a particular course of action and what its limitations were. That transparency allows crews to verify recommendations using their own experience before acting on them.

Engineering trust into aviation AI

Those findings highlight one of the most significant differences between aviation AI and many commercial AI applications. Performance alone is not enough.

An AI system that consistently generates accurate recommendations may still receive limited acceptance if operators cannot understand its reasoning or predict how it will behave in unfamiliar situations. For that reason, the LOKI project devoted significant attention to the human factors associated with AI adoption.

Researchers evaluated what level of automation aviation professionals consider appropriate and how new capabilities should be introduced into operational environments. The project concluded that trust develops gradually rather than automatically.

Clear explanations, effective training, and incremental deployment all contribute to operator confidence. Human oversight also remains essential, particularly in safety-critical applications where experienced pilots and controllers retain responsibility for operational decisions.

Those conclusions align with broader work underway across the aerospace industry. Manufacturers, regulators, and research organizations increasingly recognize that successful AI deployment depends as much on human-centered system design as on advances in machine learning itself.

Building the next generation of flight operations

The LOKI project reflects a change in how the aviation industry views AI. Earlier discussions often focused on whether the technology could automate pilot or controller responsibilities. Researchers are now asking a different question: how can AI reduce routine workload while allowing aviation professionals to focus on the decisions that require experience and accountability?

Projects such as DIRC and IPAS suggest collaborative AI may offer a practical path forward for commercial aviation. AI can identify patterns and evaluate multiple options in seconds, while pilots and controllers contribute situational judgment and the ability to manage unexpected events that fall outside predefined models.

The greatest advances may come not from replacing human operators, but from designing systems that strengthen the partnership between people and technology while maintaining the safety standards on which commercial aviation depends.

Related: Germany targets aviation growth with 15-year industry roadmap