This award recognizes an organization that has developed and implemented a groundbreaking AI solution, demonstrating exceptional technical innovation. The winning solution must showcase advanced AI capabilities and illustrate how these technologies have been effectively applied to address complex challenges. This award honors not only technical excellence but also the scalability of the solution and its tangible impact on business outcomes or industry practices, setting a new standard for AI-driven innovation and establishing a benchmark for future technological advancements.

Modern air combat is defined by a rapid OODA loop this where Saab’s AI work begins. When a pilot is simultaneously managing sensor data from multiple on-board systems, receiving feeds from off-board assets, interpreting command network traffic, and flying the aircraft, the limiting factor is human cognitive capacity. Saab initiated the Gripen AI programme to address exactly that constraint: to build a fighter that actively participates in understanding the battlespace, not merely one that displays it. The goal from day one was not a technology demonstrator. It was fieldable AI, embedded in an operational aircraft, supported by a rigorous assurance and certification process, and designed to evolve over decades of service.

Four objectives have guided the programme throughout. First, to augment the pilot rather than replace them: AI should reduce cognitive load and sharpen decision quality, whilst keeping the human in full control of every consequential action. Second, to deliver operationally relevant AI: capabilities that perform in contested, degraded, and genuinely unpredictable conditions, not only in laboratory settings. Third, to build a reusable framework: an architecture and toolchain that can be applied across platforms and domains, and upgraded incrementally. Fourth, to set a credible benchmark for safe and responsible AI in defence, one grounded in transparency, traceability, and verification rather than in aspiration.

The technical challenges were severe and interlocked. AI functions must execute within tightly bounded time budgets on airborne-qualified hardware, with no tolerance for unpredictable latency. Real combat data is scarce, classified, and highly variable, forcing Saab to combine synthetic data generation, high-fidelity simulation, and carefully curated operational data to train and validate models. Conventional AI approaches conflict with certification frameworks that assume deterministic, traceable behaviour; reconciling learning-based components with safety-critical standards required new hybrid methods and new assurance practices developed from first principles. The human-machine interface had to support pilots across a wide range of experience levels, under extreme stress, without overwhelming them or eroding their trust in the system. And the entire capability had to be designed for upgrade over a service life measured in decades, without creating an unmanageable validation burden with every iteration.

Saab’s answer is a multi-layered AI architecture that keeps perception, reasoning, and human interaction as distinct, independently verifiable layers. The perception layer applies machine-learning-based detection and classification across multi-sensor inputs and datalinks, improving track quality, threat recognition, and clutter rejection. The reasoning layer uses probabilistic models, optimisation algorithms, and rule-based tactical logic to generate and rank options, accounting for mission goals, rules of engagement, fuel state, weapons load, and threat projections. The interface layer translates all of this into intuitive, contextually adaptive pilot aids that surface the right information at the right moment.

Critically, Saab does not rely on black-box models. The system is designed to explain itself: indicating the primary factors behind a recommended manoeuvre, flagging the threat drivers underpinning a weapon suggestion, and doing so in a form that a pilot under pressure can absorb and act on. Data-driven components are constrained by explicit rules and safety monitors. A disciplined MLOps-like process, adapted for defence certification requirements, governs model versioning, configuration management, and regression testing. The same AI components that operate in the live aircraft run in a digital twin and simulation ecosystem, enabling rapid scenario iterations to accelerate both learning and validation.

The results are measurable. Pilots report significantly clearer situational awareness in complex, multi-domain scenarios where unmediated sensor data would otherwise overwhelm. AI-supported recommendations allow more options to be evaluated in less time, improving both survivability and mission success rates in operational exercises. Intelligent automation handling sensor management and navigation functions shortens the path to proficiency for new pilots and reduces workload for experienced ones. The AI components and engineering practices developed for Gripen are now being adopted across other Saab platforms, establishing a common, scalable AI capability. And for partners and customers across the defence ecosystem, Saab’s work provides a concrete, credible reference for how responsible AI integration into mission-critical systems can be done.

This is not AI at the edge of a portfolio. It is AI at the heart of one of the world’s most advanced combat aircraft, certified, operational, and already shaping the industry around it.