AI won’t modernize the federal government if it never leaves the data center.
Josh Wilson is the CEO of LMI
Over the past several years, the federal technology conversation has centered on artificial intelligence: pilots, models, copilots, automation layers. Those investments matter. AI is already reshaping analysis, logistics, cybersecurity, health and defense operations. But modernization is not the same thing as operational adoption. Fielding technology is only the beginning. What matters most is whether systems can operate, adapt and improve under real mission pressure.
Operational adoption is proven in these conditions. It is tested when networks degrade, when environments are contested, and when latency carries consequences. Modernization strategies aside, systems either hold up in those scenarios, or they don’t. This is where the next wave of federal modernization will be decided.
For more than a decade, federal modernization focused on enterprise IT: cloud migration, platform consolidation and digitized workflows. That work was necessary, but based in dashboards. Missions don’t turn on dashboards. They turn on the flight line, in shipyards, across supply chains and in space. Ground is earned at the edge.
If intelligence remains centralized — if data must travel back to distant environments before it becomes actionable — we introduce fragility into systems designed to create advantage. When connectivity drops, performance drops. When latency increases, decision quality suffers. When adversaries disrupt networks, centralized architectures become liabilities.
Modernization that depends on ideal conditions is not modernization. It is optimization for the demo environment.
The next phase of federal modernization must extend intelligence to where missions actually operate. That means building operational systems, not just software overlays. It means designing hardware, software, data flows and security together from the start, with the assumption that environments will be unpredictable and connectivity will be intermittent.
At the edge, systems must be capable of collecting data, processing it securely and enabling action locally. They must operate autonomously when disconnected and synchronize seamlessly when networks are restored. Security cannot be bolted on after deployment; it must be intrinsic to how information is generated and transmitted. And modularity must be built in, so systems evolve without requiring wholesale replacement every time threats shift.
Consider logistics. Item-level visibility is powerful in a controlled warehouse environment. It is transformational in contested conditions, where degraded networks and fragmented systems obscure readiness. Edge-enabled architectures allow data to persist, synchronize and inform decisions even when connectivity is unreliable.
Or consider homeland security operations in remote terrain. When drones and sensors collect data but rely entirely on centralized processing thousands of miles away, latency becomes operational risk. Systems that process and act locally, while maintaining secure integration with enterprise infrastructure, shift that risk profile entirely.
These are not abstract design choices. They are architectural decisions that determine whether technology enhances mission performance or simply adds complexity.
There is also a governance dimension to this shift. Missions increasingly span agencies, domains and classification levels. Data-sharing frameworks must enable collaboration without introducing paralysis. The goal is disciplined interoperability: systems engineered to move securely across boundaries while maintaining integrity and trust.
Modernization, then, is not simply about introducing AI into existing processes. It is about reengineering systems so that intelligence, resilience and security are embedded from the point of data creation through enterprise synchronization. It requires technologists who understand both software and physical constraints. It requires acquisition strategies that reward integrated design rather than stovepiped procurement. And it requires leaders willing to measure success not by the number of pilots launched, but by the durability of systems deployed.
AI will continue to evolve rapidly. Models will improve. Capabilities will expand.
But sustained advantage will not come from the most advanced algorithm in isolation. It will come from engineering resilient systems where intelligence operates securely at the edge and integrates seamlessly with enterprise infrastructure.
Modernization is more than AI. It is the disciplined engineering of operational capability — built to perform not under ideal conditions, but under real ones.