The Defense Department has always struggled to push new capabilities to the field quickly enough for them to matter. Across decades of conflict, the limiting factor has rarely been ideas or intent, but rather the machinery required to translate innovation into operational capability.
Software mobility — the ability to rapidly move software across environments, platforms, classifications, and operational contexts — is foundational to achieving CJADC2.
One of that initiative’s core principles is that data must move freely across the joint force. To achieve that, the software enabling that data to be transported, translated, and contextualized must move even faster.
That argument remains true. But a new and underappreciated challenge has emerged. Agentic AI is dramatically accelerating software development — not just for us, but for our adversaries.
And our deployment infrastructure was never designed to handle what’s coming.
The Coming Software Tsunami
We are entering the era of agentic software development. AI-assisted and agent-driven coding is already transforming how software is designed, implemented, tested, and iterated. What once required large teams and long schedules can now be accomplished by small teams in weeks.
This shift is already happening. Adoption has been rapid across the commercial sector and increasingly across the defense industrial base.
The result is straightforward: we are about to generate vastly more software at a far faster rate than our current systems were designed to handle.
Historically, if a system took two years to build and six months to deploy, engineering was the dominant bottleneck. Deployment friction mattered, but it wasn’t on the critical path.
That balance is flipping.
If mission software can be built in three months but still takes six months to accredit, integrate, and field, deployment represents roughly 70 percent of the total timeline. The bottleneck has moved downstream, and the systems downstream are already strained.
There are only so many field service representatives, cyber assessors, and accreditation authorities. The system is barely holding together under today’s software volume. A tenfold increase in output doesn’t bend that system; it breaks it.
A Strategic Asymmetry We Can No Longer Rely On
This becomes far more serious when viewed through a competitive lens. Our adversaries will have access to the same agentic development technologies.
When confronted with a novel battlefield problem, both sides will be able to design a new application in weeks. The difference won’t be who builds it faster. It will be who can put it into the fight first.
Here’s what makes this genuinely new: historically, the United States compensated for slow deployment pipelines with an asymmetric engineering advantage. Even when others moved faster operationally, they couldn’t match the depth of US talent and domain expertise needed to build sophisticated mission systems at scale. We could afford the friction because we were the only ones producing software worth deploying.
Agentic development erodes that advantage. The ability to create capable software is being democratized. As the barrier to entry falls, logistics and fielding speed become decisive.
Consider what this looks like at the extreme. Imagine aerospace companies worldwide could design next-generation autonomous strike aircraft in six months instead of five years. Assume both the United States and its adversaries have access to this capability.
Our adversaries roll their aircraft directly onto the runway. We finish ours and place them in a warehouse for eighteen months — waiting for certification, integration, and process compliance — before they ever take flight.
While unrealistic in the physical world, this is precisely the dynamic emerging in the digital domain. And unlike aircraft, software can be duplicated, iterated, and deployed at zero marginal cost. The asymmetry compounds quickly.
So, What Do We Do About It?
The honest answer is that there is no single policy change or technical solution that fixes this. But there are clear principles we must embrace.
- Treat deployment pipelines as mission systems. Accreditation, integration, and fielding are not administrative afterthoughts. They are operational capabilities that must be designed, funded, measured, and modernized with the same rigor as the software they support.
- Make software mobility a requirement, not a preference. Portable architectures, standardized interfaces, automated compliance controls, and environment-agnostic deployment models are no longer “nice to have.” They are prerequisites for operating at agentic speed. Programs that can’t demonstrate them should not be funded.
- Move to continuous, risk-informed accreditation. Static, point-in-time authorization cannot scale to agent-driven development. We need automated evidence collection, reusable security artifacts, and continuous assessment that match the cadence of modern software delivery.
- Push authority and capability closer to the edge. The further decisions are removed from operators, the slower adaptation becomes. Empowering operational units with pre-approved, composable software components, and the ability to deploy them safely, will be critical.
- Structure the public–private partnership around outcomes, not compliance. Industry alone cannot fix this, and government cannot mandate its way out of it. We need shared incentives, shared infrastructure, and shared accountability for outcomes.
Agentic software development is not coming. It is here. It will dramatically increase the volume and velocity of mission software across the defense ecosystem. If we fail to modernize how that software moves from idea to impact, we risk winning the coding race while losing the fight.
Software mobility was already essential for CJADC2. In the era of agentic development, it becomes decisive.
The question is whether we treat that as an operational problem worthy of genuine urgency, or another item on a modernization roadmap that never quite reaches the top.
Michael MacFadden is CTO at Sigma Defense.
The views and opinions expressed here are those of the author and do not necessarily reflect the editorial position of The Defense Post.
The Defense Post aims to publish a wide range of high-quality opinion and analysis from a diverse array of people – do you want to send us yours? Click here to submit an op-ed.
