Federal agencies are not failing to experiment with artificial intelligence. Instead, they’re struggling to scale it across their complex systems and missions. This challenge is becoming a defining issue for federal systems integrators (FSIs), according to Scott Stapp, chief technology officer and chief revenue officer at DEFCON AI. With over three decades of experience in government data management and decision-making, Stapp offers a clear insight into why scaling AI remains elusive — and what needs to change.

The Integration Gap is the Real Bottleneck

Stapp argues that the real problem isn’t the models themselves but the lack of integration between systems, services, and environments. “Agencies do not have the data fabric and ontology that allows things to cross,” he said. In other words, there’s no shared framework for data or information that enables AI tools to communicate and work together across different departments or agencies.

This is a critical issue because one of the core strengths of artificial intelligence is its ability to improve with more data. The more people who can contribute to an AI system, the better it becomes at making decisions and predictions. However, without a unified structure that allows these systems to share information seamlessly, progress remains limited to isolated pilot projects.

Stapp explained that in the commercial world, scaling AI is straightforward: data flows freely between platforms, models are built on shared inputs, and systems communicate with one another. But government operations are different. “It is not well connected,” he said. “It doesn’t talk. The services don’t necessarily connect and talk to each other.”

As a result, AI solutions deliver value in isolated pockets but struggle to expand beyond them. Stapp emphasized that the real challenge lies not in creating better algorithms or more advanced models — which are already being developed — but in building the underlying architecture that allows these tools to operate across systems, services, and environments.

Cloud Strategy is Shaping Everything

The government’s approach to cloud computing and AI is deliberate, driven by a desire for long-term flexibility rather than short-term convenience. Agencies are cautious about becoming locked into any single cloud environment, which means their strategies must prioritize portability and interoperability.

Stapp noted that this caution is reshaping how AI systems are designed and procured. “Once you start to look at those ontologies, where data and information can be passed seamlessly across multiple cloud environments, you’re going to see the use of AI tools grow drastically,” he said.

For FSIs, this means their solutions must also be able to operate across different cloud platforms without losing functionality or performance. The ability to move between environments is no longer optional — it’s a necessity for building scalable and sustainable AI systems in government.

The Return of the Prime Integrator

One of the clearest signals of change is the re-emergence of integration as a central function within federal IT projects. Stapp pointed out that historically, the government relied on prime contractors to integrate complex physical systems — such as aircraft or naval vessels. Now, this model is returning in the AI era.

“Right now, there are limited prime contractors able to do this work,” he said, but agencies are increasingly looking for partners who can help them connect disparate tools, data sources, and environments into a cohesive system.

Stapp highlighted DEFCON AI’s recent contract with the U.S. Marine Corps as an example of this shift. The five-year, $115 million prototype agreement focuses on logistics modernization and AI integration. It involves deploying new AI tools while ensuring they operate within a common data framework and open architecture.

“They want to bring all these new AI tools in, and they have us looking at all those tools to ensure that they fit within a data framework and open architected capability,” Stapp said. The Marine Corps aims to become an AI-first force — one where AI is not just an experiment but a foundational element of operations.

For FSIs, this signals a market shift toward orchestration. The goal isn’t just to deliver point solutions but to align vendors, tools, data, and environments into something that works seamlessly across the entire ecosystem.

Edge Constraints Force Smarter Design

Scaling AI also brings physical limitations — particularly at the edge of networks where computing power is limited. Latency can disrupt mission effectiveness, which means systems must be designed with real-time performance in mind.

Stapp explained that traditional cloud-native thinking assumes more data and compute leads to better outcomes. However, in environments far from an agency’s headquarters, this approach isn’t efficient or effective. “You want minimal amounts of data needed in that edge for the specific decisions,” he said.

This means designing systems that are distributed, efficient, and purpose-built for specific missions. The days of simply extending centralized architectures are over — agencies now need solutions tailored to their unique operational needs.

Government Data is Fundamentally Different

Another key challenge is the nature of government data itself. AI in the public sector is more complex than in commercial environments not just because of scale, but also due to other factors such as security constraints and adversarial conditions.

Stapp noted that “you have an adversary whose goal in life is to show you information that is not accurate and is deceptive.” This introduces challenges many commercial systems don’t face — including the presence of manipulated data, incomplete datasets, and classified data silos.

In defense environments, for example, Stapp pointed out that “the only people who have all the data are the people at the highest level, and those aren’t the people who are on the battlefield.” This creates a gap in real-time decision-making capabilities, forcing AI models to operate with less information while still producing reliable outcomes.

Modularity Isn’t Optional

Given these challenges, Stapp emphasized that modularity is not just good architecture — it’s essential. Each agency or service operates differently, and even when problems appear similar, the implementation rarely is.

“You must make it modular because really it doesn’t work with anything else,” he said. This means the same application often needs to be adapted across environments. “You have to do tweaks and changes to make the same application work with a different service.”

As a result, modularity isn’t just an option — it’s the only viable approach for building scalable AI systems in government.

Start Small to Scale at All

Despite the push for enterprise-wide AI adoption, Stapp argued that large-scale transformation isn’t the starting point. Instead, agencies should begin with smaller problem sets and gradually expand their capabilities.

This incremental approach reflects the realities of data access, classification, and infrastructure maturity. It also aligns with where AI is gaining traction now — in focused, mission-critical applications rather than broad, untested initiatives.

Stapp reiterated that the path to AI at scale in government runs through integration by aligning data, architectures, clouds, and mission needs in a fragmented environment. “AI won’t scale until integration does,” he said. The onus is on industry partners to help agencies address this challenge.

Conclusion

Scaling artificial intelligence in government isn’t about creating better models — it’s about building the right infrastructure to connect them. From data fabrics and interoperability to cloud portability and edge computing, the challenges are complex but not insurmountable.

For FSIs, the opportunity lies in becoming prime integrators who can bring together tools, vendors, and environments into a cohesive system. Modularity, flexibility, and a focus on real-world constraints will be key as agencies move toward more scalable AI solutions.

As Stapp concluded, “The number one thing with any customer is you’re there to solve their problem.” That mindset — paired with the right technical approach — will determine whether government AI can finally scale beyond pilot projects and into meaningful transformation. Readers should watch for how integration strategies evolve in the coming years as agencies continue to seek solutions that work across silos, systems, and services.

---

Original Source

This article is based on publicly available reporting. For the complete original story, visit the publisher’s article.