The growing complexity of multi‑domain operational environments—land, air, sea, cyber and electromagnetic spectrum—has dramatically increased the volume and velocity of data arriving at command‑and‑control (C2) centers, straining the OODA (Observe‑Orient‑Decide‑Act) loop. Current AI systems in defence are largely reactive and isolated tools that still rely on human operators to fuse information, assess scenarios and craft courses of action.
This paper proposes a agentic AI architecture that enables AI systems to autonomously plan tasks, access heterogeneous data sources, invoke tools, and act audibly, thereby supporting decision‑making across the Brazilian Navy, Army and Air Force. The architecture’s core capabilities are:
- Autonomous planning – generate action sequences from mission objectives;
- Data source integration – unified interfaces to sensors, intelligence feeds and historical databases;
- Tool execution – invoke simulation, optimization or analytics modules;
- Auditable actions – record decisions as audio or logs for traceability.
Four application fronts are examined in detail:
- Decision support: provide commanders with multi‑option comparisons and risk assessments;
- Situational analysis: fuse cross‑domain perception data into a unified operational picture;
- Feasibility studies: use simulation models to evaluate resource needs and success probabilities of plans;
- Countermeasure suggestion: automatically generate defensive or offensive options when threats are detected.
Realizing these functions requires specific data and sensor access provisions: high‑bandwidth, low‑latency networking; standardized metadata schemas; and cross‑domain identity authentication. For security and permission safeguards, a layered authorization model is introduced, covering administrative, strategic, operational and tactical levels, complemented by audit logs, behavior monitoring and anomaly detection to prevent misuse or accidental actions.
With this design, agentic AI can accelerate information processing, improve plan quality, and enhance collaborative effectiveness in multi‑domain operations while preserving the ultimate decision authority of human commanders.
Review