Executive Summary
The Department of War (DoW) has elevated Applied Artificial Intelligence (AAI) as one of six Critical Technology Areas to deliver tangible results for the Warfighter. The Department’s message is clear: we need to leverage AI to transform decision making, make the force faster and more lethal, and improve efficiency. AAI is meant to live inside command-and-control systems and enterprise processes, producing intelligent workflows that can be fielded quickly and safely.
Legion is a clear, ready, already-fielded example of AAI. Legion’s secure agent-orchestration platform uses agentic AI workflows and human-augmented automation to transform how organizations execute intelligence analysis, current operations, training, mission planning, and administrative and business process work. Its deployment in demanding defense environments demonstrates what “applied” actually means: coordinated AI agents integrated into existing tools and data, operating in classified and air-gapped settings, with humans firmly in control of high-stakes decisions.
DoW’s Applied AI: From Concept to Operational Capability
In the DoW framework, Applied AI is not primarily about inventing new algorithms. It is about embedding AI into command-and-control, intelligence, logistics, and enterprise systems in a way that materially improves the speed and quality of decision-making. The emphasis on “intelligent workflows” signals that AI is expected to operate across whole processes—collecting, triaging, transforming, and presenting information—not as a standalone chatbot or analytic widget.
The focus is also explicitly near-term. By elevating AAI as a priority, DoW is asking for capabilities that can be deployed in real units and organizations on a relatively short timeline, rather than speculative research programs. This implies several expectations: that solutions integrate with existing mission and enterprise systems rather than replacing them wholesale; that they handle real-world constraints such as classification, disconnected or contested environments, and legacy data; and that they produce measurable improvements in tempo, accuracy, or capacity. This is the lens through which Legion builds and deploys AI.
Agentic AI and Human-Augmented Automation
DoW’s language around “intelligent workflows” aligns directly with the emerging pattern of agentic AI. Multiple specialized AI agents can plan, reason, and act across tools and data sources in order to complete multi-step tasks. One agent might be responsible for searching and retrieving documents, another for extracting entities and relationships, another for drafting assessments or orders, and another for triggering actions in downstream systems.
These agents are coordinated by an orchestration layer that routes tasks, manages context, enforces guardrails, and maintains logs. Humans occupy the top of this loop: they set goals and constraints, initiate workflows, inspect intermediate and final outputs, and approve or reject substantive actions. The result is human-augmented automation rather than fully autonomous decision-making.
For warfighters, this pattern is attractive because it reconciles three hard requirements. First, decisions must keep pace with rapidly evolving data and events; automation is essential for handling volume and complexity. Second, commanders and staff must remain in or on the loop for high-stakes decisions, especially where lethality, escalation, or strategic signaling are involved. Third, systems must support accountability, with clear audit trails for what information was considered, what actions were taken, and why.
In this model, AI acts as scalable digital labor. It executes the repetitive, mechanical, or high-volume portions of work—retrieving data, cross-referencing sources, drafting routine products—while humans focus on interpretation, prioritization, and judgment. This is the core conceptual move behind AAI: shifting from AI as a curiosity to AI as an embedded labor force that expands the capacity and effectiveness of the force. Legion’s product is built deliberately around this pattern.
Legion Intelligence as an Applied AI Platform
Platform overview
Legion Intelligence is a secure, model-agnostic, flexible, agent-orchestration platform designed for defense, government, and mission-critical enterprises. Rather than presenting a single application, Legion connects to the systems and data an organization already uses—mission systems, data platforms, ERPs, collaboration suites—and enables AI agents to operate across them. The platform emphasizes security and compliance from the outset, supporting flexible and resilient deployments, including on-premise, air-gapped, and classified networks. Critically, Legion is deployable to the edge, providing an extensible and resilient common operating system - from the cloud to the field. Data ownership, encryption, and access control are not optional features; they are baseline assumptions.
Architecturally, Legion is indifferent to any particular model. It can orchestrate commercial foundation models, open-source models, and sovereign or government-furnished models side-by-side, along with traditional software services. This allows organizations to evolve their model strategy—swapping models as policy, performance, or licensing needs change—without re-architecting workflows.
Crucially, Legion integrates into operational environments rather than expecting users to leave their existing tools. Its technology underpins SOFchat, deployed across U.S. Special Operations Command and integrates into mission command and collaboration environments that operators already rely on. That means AI agents can work on the same operational and intelligence data that humans use, in the same context, with the same security posture.
Evidence of fielded impact
SOFchat is available to tens of thousands of personnel at USSOCOM and subordinate units where it supports mission planning, intelligence analysis, training management, and monitoring of real-world situations. Intelligence products and operational summaries that previously took hours now take minutes. In training environments, scenario building, skill validation, and eval writing have been significantly accelerated, freeing instructors and staff from repetitive checks and documentation tasks.
In each case, the impact is not abstract. Analysts spend less time hunting for documents and formatting reports and more time evaluating sources and refining assessments. Operations staff spend less time manually compiling situation reports and more time considering courses of action and risk. Trainers spend less time fixing scenario inconsistencies and more time coaching. These are precisely the kinds of operational improvements that distinguish applied AI from experimentation.
Applied AI Use Cases: Legion Across the Mission Spectrum
Within the DoW AAI framing, Legion’s impact today can be understood across three major domains: intelligence analysis; current operations and planning; and administrative and business processes.
Intelligence analysis
Decision superiority starts with better intelligence, delivered faster. Legion’s agentic workflows automate tedious portions of the intelligence cycle while keeping analysts clearly in charge. Agents can ingest reporting from structured databases, unstructured documents, chat logs, and mixed media. They index and organize this information so that it can be queried by entities, topics, and events. When an analyst asks a question, agents retrieve relevant material, cross-reference sources, and assemble a first-pass synthesis, complete with citations to underlying documents.
Analysts then review, correct, and extend these drafts. They can challenge assumptions, add context that models cannot see, and reshape products for their intended audience. The net effect is that analysts spend less time on mechanical tasks—copy-pasting, reformatting, searching across fragmented systems—and more time on the expert reasoning that cannot be automated. Intelligence summaries, command briefs, and updates move from being scarce, manually produced artifacts to being continuously refreshed products, updated as new reporting arrives.
Legion can also support continuous monitoring of ongoing situations. Agents track new reporting against previous baselines, flag changes that matter, and propose updates to existing running estimates. Instead of analysts manually polling for updates, the system surfaces deltas proactively and updates in the appropriate format. This shift from pull to push—while still allowing analysts to validate and curate the outputs—is a hallmark of applied AI in the intelligence domain.
Current operations and planning
Current operations and planning are where intelligence, logistics, and command converge. Here, Legion helps compress planning cycles, stabilize documentation, and reduce the cognitive load on staff. Because the platform is integrated with mission command environments, agents can read from and write to the same systems operators use to track tasks, statuses, and orders.
Consider the daily production of situation reports and fragmentary orders. Traditionally, staff officers collect updates from multiple systems and chats, reconcile inconsistencies, and then draft structured documents in standard formats. With Legion, agents can pull data from those systems automatically, populate templates, and generate coherent drafts that staff refine. The staff officer still owns the product and the decision; the difference is that they begin from a near-complete draft rather than a blank page.
Planning workflows see similar benefits. Agents can assemble the relevant intelligence, prior plans, and doctrinal templates; pre-populate annexes and graphics; and document proposed courses of action. Human planners concentrate on evaluating these COAs, understanding trade-offs, and engaging with commanders, rather than spending the bulk of their time on data gathering and formatting. In training and readiness contexts, agents assist with building and validating scenarios, checking for internal consistency, and generating supporting materials, again freeing humans to focus on evaluation and coaching.
In all of these cases, AI is not a separate system operating on the side. It is embedded in the planning and operations workflows that the DoW cares most about, shortening OODA loops while preserving human control.
Administrative and business process workflows
DoW’s AAI vision extends beyond warfighting. To become an “AI-first” organization, AI must also transform the administrative and business processes that consume vast amounts of staff time. Legion addresses this through secure process automation that applies the same agentic model to back-office work.
In a defense context, this includes workflows like travel and voucher processing, contract documentation, personnel actions, training records, and compliance reporting. Legion’s agents can scan inboxes, identify relevant messages, extract key fields, populate forms, and generate draft responses or records. They can move data between email, case management systems, and document repositories without manual re-entry.
What ties these use cases together is the concept of digital labor. The AI is not replacing human oversight, but it is performing the same sort of repetitive, rules-driven tasks that junior staff or clerks might otherwise handle. Supervisors and specialists remain responsible for approvals and exceptions, but routine work is handled at machine speed. This is how AAI translates into real capacity gains: it returns time to people who are currently trapped in low-value work.
Legion is Applied AI
Viewed against the DoW concept of Applied AI, Legion stands out in several dimensions. First, it is decision-focused rather than demo-focused. The value is expressed in concrete improvements: intelligence products produced more quickly and consistently, planning cycles shortened, training and administrative burdens reduced. Each of these directly affects the speed and quality of decisions or the amount of human attention available for them.
Second, Legion is workflow-centric rather than model-centric. The platform does not ask users to care which specific model is performing a given task. What matters is that workflows spanning multiple systems—data platforms, mission tools, ERPs, email, documents—can be defined, executed, monitored, and refined. The orchestrator coordinates agents, enforces guardrails, and captures logs. That is precisely what the DoW describes when it talks about intelligent workflows embedded into existing missions and processes.
Third, the system is human-augmented and auditable by design. Humans define objectives, approve high-impact actions, and retain ultimate authority. Every agent action is logged, and outputs can be traced back to inputs. This aligns with the cultural and regulatory expectations of national security organizations, which must be able to explain how decisions were made and which information was used.
Finally, Legion spans both mission and enterprise domains. It is as capable of supporting intelligence analysis and mission planning as it is of supporting finance, procurement, or engineering workflows. That breadth is essential to any serious notion of an “AI-first” department, because it recognizes that warfighting effectiveness depends not only on what happens in operations centers and on the battlefield, but also on the institutional machinery behind them.
Implications
The DoW’s Applied AI priority sets a high bar: AI must be operational, embedded, secure, and accountable, with impact measured in better, faster decisions and reclaimed human capacity. Legion Intelligence shows that this bar is not abstract. Agentic AI workflows, properly orchestrated and supervised, can already deliver meaningful gains in intelligence, operations, and administrative domains. The DoW must also provide this at a cost that doesn’t break the Department’s budget, and in a way to ensure the government remains firmly in control of their data and avoids lock-in from model, data, compute, or hardware vendors.
For organizations translating the AAI framework into concrete projects and acquisitions, several implications follow. Platforms that can orchestrate agents across existing systems will deliver more value than point solutions that live in isolation. Human-augmented automation should be treated as the default pattern for high-stakes and high-friction work, ensuring both speed and control. And AAI initiatives should be anchored in specific workflows—INTSUMs, SITREPs, mission plans, training scenarios, financial close, and similar processes—where time savings and error reductions can be measured and iterated on.
By these standards, Legion is a strong example of Applied AI in action: not a promise of future capability, but a fielded reference implementation for how DoW’s AAI vision can be realized in real units, on real networks, supporting real missions.
Points of Contact
Ben Van Roo, CEO and Co-Founder, ben@legionintel.com
Chris Hume, CBO, chris@legionintel.com
Nick Weir, VP of Mission Engineering, nick@legionintel.com
Brian Lampert, Director of Public Sector, brian.lampert@legionintel.com