Explore Artificial Intelligence (AI)

Agentic AI: the next step in decision support

Most AI systems in use today work in a passive manner.

They complete a bespoke task when a user asks for it and wait for the next instruction. Agentic AI works differently. Give it a goal, and it works out independently which agents it needs to query, what data to pull, and how to assemble an assessment — with multiple AI agents collaborating to accomplish complex tasks with minimal human interaction.

Think of a military headquarters. It runs on specialised expertise: a terrain analyst, an intelligence officer, a logistics planner. Traditionally, a chief of staff coordinates between them, decides who to consult, and waits for each input before forming a picture. Agentic AI can take on that coordinating role itself: one agent might use satellite images to produce a terrain analysis — routes, obstacles, rivers, bridges, built-up areas — while another analyses enemy equipment, activity, and doctrine. Combined, commanders get a far richer, faster picture than either agent could produce alone.

Agentic AI is already a focus for development across the SitaWare suite, with particular potential for SitaWare Insight, our data-driven decision support tool, and SitaWare Headquarters, our C4ISR system.

No data, no AI

Agentic AI can rapidly accelerate the time it takes to use data, but only if that data can be trusted. As with any AI capability, credibility and reliability of the underlying information is essential, otherwise the system risks producing a confident but incorrect result. This is why Systematic's approach starts with the individual algorithms: understanding what each agent does and how it reaches its output, before combining agents into an agentic capability.

Humans remain in the loop

Agentic AI is an evolutionary step, not a replacement for the commander. It can process vast amounts of data and present options at speed, but it cannot decide whether a mission should prioritise speed, surprise, or logistics sustainment. That judgement stays with the human. The commander assesses the recommendations, makes the final decision, and carries the responsibility.

Implementing AI in the maritime domain

The Recognised Maritime Picture (RMP) is a cornerstone of naval operations, and keeping it accurate and current is decisive for mission success. With manning under pressure, correlating the sheer number of tracks is slow and prone to error — SitaWare Maritime's correlation and fusion engines lift that burden.

By feeding in historical data — ship locations, vessel characteristics, typical movements — SitaWare builds an accurate pattern of life and flags abnormal behaviour automatically. The same approach extends to ship classification: algorithms trained on image or silhouette data can identify vessel classes, even individual ships, giving commanders persistent surveillance from sources including shore-based sensors.

Naval warfare against a peer adversary moves fast. AI gives commanders the means to get inside an adversary's observe, orient, decide, act (OODA) loop — and Systematic continues to explore new capabilities for SitaWare Maritime.

Towards agentic AI

Systematic is building agentic capability across the SitaWare suite, including new features that let operators partner with a large language model (LLM) chatbot as a mission-support "wingman," with agents trained to follow established processes such as NATO standard APP-28 land tactical planning procedures.

SitaWare's advantage is its infrastructure: it already connects the strategic to the tactical level, draws on inputs from more than 50 nations, and keeps working in disconnected, denied, intermittent, or low-bandwidth (DDIL) environments via SitaWare Tactical Communication. That foundation is what makes agentic AI possible.

AI assisted toolbox

SitaWare’s AI Assisted Toolbox is designed to ease the workload of operators, in the first instance with regards to the processes required to build command layers in the software’s map interface.

The AI Assisted Toolbox eases operators' workload, starting with building command layers in the map interface. The algorithm learns from users' symbology choices, prioritising what's most likely needed for a given scenario — speeding up the process and limiting errors.

This knowledge can also be shared between users: joining a mission already in progress means drawing on what another user has already built. The benefit is clear at higher command levels, but it helps at the tactical level too — simplifying a dismounted commander's interaction with their device means less time head-down, more time aware of their surroundings.

Want to dive even deeper into the technology area?

We have selected three recent technology-focused news and blog articles and our dedicated AI eBook for the defence domain. For more knowledge on this topic, visit our News & Knowledge section.