Across the defense and national security community, the question is no longer whether artificial intelligence (AI) will be adopted, but whether it can be trusted to perform when it matters most. According to the Stanford 2025 AI Index, 78% of organizations are using AI in at least one business function and 99% are actively investing in the technology. Yet according to McKinsey’s 2025 Workplace AI research, almost all companies are investing in AI—92% plan to increase their AI investment—while only 1% of those organizations believe they have reached AI maturity. That gap between access and readiness is beyond a strategic concern; in environments where mission outcomes and human lives depend on the quality of information, it is an operational one.
Access is Not the Same as Readiness
The instinct to equate access with capability is understandable, but it is a dangerous assumption in high-stakes environments. Think of an airman fresh out of boot camp, handed a bag of keys to every fighter jet on the flight line. The access is real. The readiness is not. The same logic applies to AI: deploying a model, standing up a co-pilot or completing a proof of concept does not mean an organization can depend on that capability in the field.
Many organizations are already asking smart questions: Which model fits best? How much Graphics Processing Unit (GPU) capacity is required? Should we build our own Large Language Model (LLM)? These are important considerations, but the first question is: what standard does this capability have to meet once it enters real workflow?
The Prompt is Not the System
There is a persistent misconception that and AI system can be judged by quality of its interface. A clean prompt window, a fast response and a well-formatted output do not reflect the operational substance of the capability underneath. The real system is what the user never sees: what context gets retrieved and from which sources, and which identity and access boundaries are enforced. Enterprise AI is not a smarter interface. It is an operational layer, and organizations that evaluate it only by its outputs are missing everything that determines whether those outputs can be trusted in consequential decisions.
Ungrounded AI Creates “Trust Drag”
When a model responds without access to the right context, drawing instead on general training data that may be months or years out of date, the answer may sound confident, fluent and plausible. But confidence is not the same as context. Consider a doctor asked to provide a diagnosis before the nurse has gathered the patient’s vitals or reviewed their health history. The doctor may still sound authoritative, but the judgment is unanchored. Grounded retrieval is dynamic: before the model responds, the system reaches into approved, current, organizationally controlled data sources and passes that context to the model at runtime. The result is an answer that is not just plausible, it is anchored, explainable and traceable.
Without that grounding, organizations experience what can be called “trust drag.” Users begin to re-check outputs, manually validate claims, compare responses against other sources and eventually hesitate before acting on AI-generated information. Over time, they route around the system entirely. The AI capability that was supposed to accelerate operations starts working in reverse, and the burden of assembling context, verifying provenance and bridging knowledge gaps shifts back to the analyst, the engineer and the operator. The complexity was never eliminated. It was simply relocated.
Operational Trust is the New Standard
IBM has reported that among organizations that experienced breaches involving AI, 63% had no AI governance policy in place, or were still developing one, and 97% lacked proper AI access controls. The Cloud Security Alliance (CSA) found in March 2026 that 68% of organizations could not clearly distinguish AI agent actions from human actions. These are not abstract governance concerns. They are operational control problems, and they arrive the moment AI begins participating in consequential workflow. The relevant questions are concrete:
- What data can the system touch?
- Which sources can it rely on?
- What tools can it invoke?
- What actions can it trigger?
- What can be monitored, attributed and audited after the fact?
Meeting that standard requires three things working together. First, infrastructure built with security as a foundational design principle, not added after deployment, covering compute, storage, networking and GPUs. Second, continuous observability and control over AI models, agents and datasets throughout their full lifecycle, not just at deployment. Third, deployability across every environment where the mission requires it: enterprise data centers, cloud and the tactical edge. In operational communities, the mission does not happen under ideal conditions. It occurs in austere, disconnected, high-pressure environments—and the AI capability must function there too.
Nutanix addresses this through Nutanix Enterprise AI, which delivers end-to-end observability into models, agents and datasets, combined with the Nutanix Cloud Infrastructure for secure compute and networking and Nutanix Central and Cloud Manager for global identity and access management, governance and automation. The goal is a single, consistent operational layer that can be deployed anywhere and trusted everywhere.
Building the Capability That the Mission Requires
The measure of AI maturity is not whether an organization has access to models or impressive demos to show stakeholders. It is whether the capability can be trusted under pressure, with the right controls, visibility, boundaries and operational confidence to rely on its outputs. In the Department of War (DoW) and U.S. Special Operations Command (USSOCOM) environments, speed without trust is not an advantage; it is friction. Achieving operational trust requires treating AI not just as a feature or a tool, but as part of the infrastructure—starting with the fundamental questions behind the prompt.
To explore these concepts in depth and see how Nutanix approaches operational AI for mission-critical environments, check out the full presentation.
Carahsoft Technology Corp. is The Trusted Government IT Solutions Provider, supporting Public Sector organizations across Federal, State and Local Government agencies and Education and Healthcare markets. As the Master Government Aggregator for our vendor partners, including Nutanix, we deliver solutions for Geospatial, Cybersecurity, MultiCloud, DevSecOps, Artificial Intelligence, Customer Experience and Engagement, Open Source and more. Working with resellers, systems integrators and consultants, our sales and marketing teams provide industry leading IT products, services and training through hundreds of contract vehicles. Explore the Carahsoft Blog to learn more about the latest trends in Government technology markets and solutions, as well as Carahsoft’s ecosystem of partner thought-leaders.
