Large Language Models and the Modernization of Department of War Workflows

By: Travis Hartman, CANA Senior Operations Research Analyst

“Welcome to the party, pal.”- John McClane (Bruce Willis) in Die Hard (1988)

For decades, the Department of War has relied on development and acquisition pathways designed for a different era of technology. Those processes were built around long timelines, rigid software cycles, and highly specialized systems that often required extensive user training and institutional knowledge simply to operate. While those approaches once represented the gold standard for reliability and control, the pace of modern operational demands is exposing a growing gap between how quickly missions evolve and how quickly software systems can adapt.

LLMs are starting to close that gap.

Across the federal space, organizations are realizing that artificial intelligence is no longer just a future capability reserved for advanced analytics or autonomous systems. LLMs are becoming operational productivity tools that can fundamentally improve how teams develop software, manage information, and interact with complex systems. The most immediate value is not replacing personnel; it is reducing repetitive tasks, simplifying workflows, and allowing personnel to focus more time on mission execution instead of administrative overhead.

For software development teams, LLM-enabled tools are accelerating many of the most time-consuming aspects of the development lifecycle. Tasks that previously required significant manual effort, such as code review support, documentation generation, test case creation, and debugging, are now increasingly augmented by AI-enabled workflows. Development teams can move faster, onboard new engineers more efficiently, and spend more time focused on operational outcomes. A simple example is documentation generation: README documents are often treated as lower priority tasks during development cycles, despite their importance, and frequently become stale or out-of-date. LLM-enabled workflows can generate and maintain README documentation automatically alongside code changes while also performing continuous code validation and formatting reviews in the background.

LLMs are also changing how operational users interact with software systems.

Much of the daily workload across government and defense organizations is repetitive, time consuming, and operationally necessary, but low value from a strategic perspective. Status reports, planning updates, message drafting, meeting summaries, data lookups, and workflow management consume thousands of labor hours across programs every year. LLM-enabled systems can now reduce much of this burden by automating routine administrative tasks and simplifying how users access information and complete workflows.

The problems become even more apparent in many legacy government software environments still in use today.

Under traditional development and deployment models, even relatively small software modifications can require extensive coordination, manual testing, and lengthy deployment cycles. At the same time, operational users often depend heavily on institutional knowledge to use systems effectively. User interface workflows frequently require planners and operators to memorize where to click, navigate through layered menus, and complete specialized training before becoming proficient in the software. The operational burden is not just technical debt within the codebase; it is cognitive debt imposed on both developers and end users.

This changes the interaction model entirely.

Instead of forcing users to adapt to software, LLM-driven systems can increasingly adapt to users. Natural language interfaces, guided workflows, and context-aware assistance reduce dependency on memorized processes and highly specialized training pipelines. Rather than navigating dozens of menus to complete a task, planners and operators can increasingly describe intent directly to the system using familiar conversational patterns.

That matters operationally.

Many legacy systems were designed around the assumption that users would require significant training before becoming operationally effective. In practice, this often meant that systems became dependent on a relatively small number of highly experienced personnel who understood the platform’s workflows, terminology, and interface logic. LLM-enabled interfaces invert that model. Users no longer need to remember exact workflows, button locations, or system-specific terminology to complete routine tasks. The interaction becomes more intuitive, easier to navigate, and significantly faster to learn.

For operational planners, this can reduce onboarding timelines and lower training requirements while improving usability during high-tempo operations. For development teams, it can reduce support overhead, simplify user adoption challenges, and accelerate feedback cycles between developers and end users.

None of this eliminates the need for human oversight, domain expertise, or rigorous validation. In defense environments especially, trust remains paramount. Subject matter experts, operators, developers, and program leadership remain responsible for validating outputs, assessing operational risk, and ensuring mission alignment. LLMs are not replacing decision-makers; they are accelerating their ability to operate effectively.

The organizations that successfully integrate these technologies will not necessarily be the ones with the largest budgets or the most advanced infrastructure. They will be the organizations willing to rethink legacy workflows, reduce unnecessary friction, and modernize how work itself gets accomplished.

The traditional development model is becoming increasingly difficult to justify. Systems designed around static workflows, long training cycles, and slow iteration timelines are increasingly misaligned with the operational tempo required today. LLMs are not simply another software feature layered onto existing processes. They represent a shift toward adaptive systems, faster iteration cycles, and more effective human-machine collaboration across the Department of War enterprise.

As government organizations continue expanding access to Generative AI and LLM capabilities, the question is no longer whether these tools will influence defense operations. The question is how quickly organizations can integrate them responsibly enough to keep pace with the mission demands ahead.

____________________________________________________________________________

ABOUT CANA

CANA is a woman-owned small business (WOSB) and Service-Disabled Veteran-Owned Small Business (SDVOSB) that empowers federal and commercial organizations to thrive in a global digital world through precise and adaptable technology solutions. We fuse our rigorous analytics and top-tier talent with deep expertise in complex logistics to design customer-centric, powerful solutions. We strive to create an environment that allows our Team and Clients more time to focus on the things that matter most.