The US Air Force’s third Decision Advantage Sprint for Human-Machine Teaming, known as DASH 3, marked another step in bringing AI into real battle planning workflows.
The event brought together US operators and coalition partners from Canada and the UK to see how AI tools could speed up and expand multi-domain decision-making.
Led by the Advanced Battle Management System Cross-Functional Team, DASH 3 gave industry participants a chance to work side by side with operators to test AI systems designed to provide a “decision advantage.”
Generating Courses of Action
The sprint focused on generating courses of action (COAs), which are structured sets of operational choices aligned with a commander’s intent under tight time and resource constraints.
These plans are complex, spanning long-range kill chains, electromagnetic battle management, space and cyber operations, or agile combat tasks like rapidly relocating aircraft.
Traditionally, building COAs is a manual process that takes minutes and often yields only a handful of viable options.
In DASH 3, AI tools generated multi-domain COAs in under a minute, factoring in operational risk, geospatial routing, timing, force packaging, and refueling requirements.
Comparative assessments showed AI solutions were up to 90 percent faster than human planners.
The most refined outputs had a 97-percent tactical validity rate, while human-developed solutions took about 19 minutes and less than half were only deemed similarly viable.
The air force emphasized that AI did not replace human judgment but amplified it. Instead of offering a single “best” answer, the tools reportedly expanded the range of workable options, letting commanders weigh risk, trade-offs, and strategy under pressure.
Participants noted that AI absorbed much of the analytical workload, freeing humans to concentrate on critical decision-making and approvals.
DASH 3 also showed that machines could propose multiple COAs in parallel, enabling decision branching at a speed and scale humans can’t match in compressed timelines.
Working Through Challenges
DASH 3 worked through practical limitations that continue to shape AI integration in operational planning.
One recurring challenge has been weather. While real-world missions are affected by changing conditions, the experiment did not yet incorporate dynamic weather data into the AI models.
Instead, operators simulated disruptions like airfield closures, delays, and degraded operations through “white carding.”
Future iterations aim to integrate live weather data.
Another focus was AI reliability, including the risk of “hallucinations,” or incorrect outputs that can occur in large language models.
Developers implemented safeguards to minimize these failure modes and monitored performance throughout the sprint.
Lessons from the 2025 DASH are expected to inform future experiments in 2026, focusing on improving AI trust, reliability, and multinational coordination.
