NATO’s major exercises are only as useful as the realism they create, the pressure they place on participants, and the lessons they help the Alliance carry forward.
That is the purpose behind AI in Audacious Training, one of Allied Command Transformation’s 2026 Beacon Projects, delivered as part of the wider Audacious Training initiative. The project is using artificial intelligence and digital tools to help modernize how NATO exercises are designed, controlled, and refined. Rather than treating AI as a distant concept, the effort focuses on practical support to the training teams responsible for building complex, realistic scenarios.
A practical workbench for exercise teams
The NATO-Maven Smart System, or N-MSS, is central to this work. Rather than a “black box” AI system, it functions as a digital backbone and day-to-day workbench that exercise teams can use to build, manage, and respond to training activity in near-real time.
One of the most immediate applications is the development of Main Events List/Main Incidents List, or MEL/MIL, content. These storylines, events, and “injects” drive the decisions and actions of the participants during an exercise. Traditionally, this process requires significant manual effort from experienced planners.
The future is undoubtedly more digital. We need tools that make us quicker, support the decision-making cycle of the headquarters, and reduce the staffing burden. The Maven Smart System is one example of how we are implementing a digital system across ACO and training headquarters to use and exploit it within the warfighting system.
– Colonel Kevin Rafferty
JWC Deputy Chief of Staff Operations
Through the Exercise Content Creation Operations Hub, the NATO-Maven Smart System can help generate an initial exercise storyline and inject set that planners can then save, edit, refine, and reuse. Alongside the Opponent Operations Hub, the Exercise Content Creation Operations Hub was coordinated and developed under the oversight of the Joint Warfare Centre, helping ensure both tools are aligned with exercise requirements and the broader direction of AI in Audacious Training. This supports faster iteration and more flexible scenario development without removing human ownership of the exercise design.
During execution, the Opponent Operations Hub helps speed the interaction cycle between allied actions and opposing force responses. When “injects”, or actions taken in the scenario, from allied forces occur, the NATO-Maven Smart System can propose credible opponent actions more quickly and align them to the overall scheme of manoeuvre. This helps opposing force cells keep pace with a dynamic exercise environment and supports more responsive opposition play.
The NATO-Maven Smart System is also being developed to reduce friction in Exercise Control (referred to as “EXCON”). Through the Exercise Control Briefing Environment, the system is increasingly becoming a place where briefs, updates, and battle rhythm events can be conducted, reducing reliance on multiple disconnected documents, slide decks, and ad hoc coordination methods.
Keeping development aligned across NATO training centres
Close coordination between the Joint Warfare Centre and the Joint Force Training Centre is central to the project’s next phase. AI in Audacious Training is not intended to create isolated tools for one headquarters or one exercise series. The Joint Warfare Centre, the Joint Force Training Centre and Headquarters Supreme Allied Commander Transformation are developing governance and interoperability efforts so that the NATO-Maven Smart System development supports exercises across both training centres and contributes to the wider NATO training enterprise.
That means aligning data standards, workflows, requirements, and capability priorities from the outset. Recent engagement between the commands has focused on digitizing the exercise environment, refining platform requirements, and identifying emerging AI use cases. It has also helped clarify which functions should be integrated into the system and which may still require external tools. NATO’s SANDI, or Data Science and AI Sandbox, is also helping shape how digital and AI-enabled capabilities are matured for practical use.
The Joint Force Training Centre is an active participant in this development, ensuring the application supports tactical echelon requirements to enable fully integrated all-in exercises. Following our staff engagement at the Joint Warfare Centre in April, we came away with a clear picture of where the NATO-Maven Smart System is headed and where the Joint Force Training Centre fits within that roadmap. Our teams are already working alongside the Joint Warfare Centre counterparts in the content development cycle for STEADFAST DUEL 2026, and we are planning to employ the system for exercise content generation in LOYAL LEDA 2027. Getting hands-on with these tools now – rather than waiting for them to mature in isolation – is exactly how we ensure that what gets built actually works for both training centres to support the entire NATO warfighting system.
– Brigadier General Zoltan Barany
JFTC Deputy Commander and Chief of Staff
The Joint Force Training Centre’s engagement has already included practical experimentation. During LOYAL LEDA 2026, a small team conducted an initial hands-on trial by configuring a dedicated AI agent with exercise reference documentation to support the generation of scenario events during the unscripted “dynamic play” phase. The trial helped test functionality, identify constraints, and assess how the toolset could apply within the Joint Force Training Centre’s exercise planning cycle.
That experience is now informing the Joint Force Training Centre’s contribution to follow-on content development work, including structured user feedback on how the tools support scenario development, opposing force response, and future exercise preparation. The goal is to automate where it adds value in exercise preparation, execution, and evaluation, while preserving operational requirements, user needs, and the professional judgement of exercise staff.
Measuring what “better” means
As the project matures, the Joint Warfare Centre and the Joint Force Training Centre are establishing a measurement plan to compare current workflows against AI-enabled ones. Potential measures include the time required to build Main Events List/Main Incidents List content, the number of scenario iterations planners can produce, response latency during execution, reductions in exercise control staff workload, and the rate at which exercise content can be reused. These indicators will help show whether the tools are increasing speed, coherence, and agility in ways that matter to exercise teams.
User feedback will also remain central. Exercise planners and subject-matter experts will continue to review the usability of the tools, the quality and relevance of generated scenario events, and the credibility of the actions of opposing forces in the storyline. The process will also refine how AI-generated outputs are reviewed, edited, and approved by exercise staff, ensuring the tools improve speed and consistency without reducing control, fidelity, or professional judgement.
Over the coming months, the immediate priority is operational testing in preparation for , where the Exercise Content Creation Operations Hub and the Opponent Operations Hub will be used and assessed against exercise requirements and subject-matter expert judgement. This practical use will provide the evidence needed to confirm what works, what needs adjustment, and where the tools best fit within the exercise design process.
AI in Audacious Training shows how Allied Command Transformation’s Beacon Projects are intended to work: identify a practical problem, develop usable capability quickly, test it with real users, and scale what works. As the Joint Warfare Centre continues development and the Joint Force Training Centre remains closely involved, the project has already moved from concept toward practical implementation across NATO’s training enterprise.