Expert Q&A By Chris Brown, Public Sector CTO, Virtualitics
This Expert Q&A is part of the AI in Practice series, where Virtualitics experts share perspectives on the evolving role of artificial intelligence in national security.
At AFA Warfare, leaders from the Department of the Air Force and industry discussed how AI is moving from experimentation to operational execution. Chris Brown, Public Sector CTO at Virtualitics, joined the panel to share how AI is already supporting readiness and sustainment across the defense enterprise.
Below are key questions and insights from the discussion.
How is AI already delivering measurable readiness and sustainment impact?
AI is helping improve decision-making processes that support readiness and sustainment.
Replacing Spreadsheets. Across the Department, many readiness processes still rely on manual workflows and disconnected spreadsheets. AI enables predictive maintenance, supply forecasting, and regulated asset storage planning—allowing leaders to anticipate constraints before they impact operational availability.
Predictive Success. For example, AI-driven planning can support efforts to reduce manual re-storage processes for regulated assets and optimize storage against safety constraints such as net explosive weight limits. In sustainment environments, AI models can forecast parts demand months in advance—allowing teams to get ahead of long supply chain lead times rather than reacting when inventory hits zero.
Operational Wins. The result: fewer readiness disruptions, reduced manual workload, and better alignment between supply, maintenance, and operational tempo.
What are the biggest barriers slowing AI adoption in defense today?
Technology is not the primary bottleneck—data governance and accreditation processes are.
Two major friction points surfaced during the discussion:
1. Data Governance and Security Controls
AI systems must ensure that classified data inputs—and the resulting insights—align with user permissions and security requirements. Without clear governance, attribute-based access control (ABAC), and data lineage, scaling AI becomes risky.
2. System-by-System ATO Processes
When each new AI agent or workflow requires a full Authority to Operate (ATO) cycle, deployment speed slows dramatically. Treating AI development platforms as infrastructure—rather than bespoke systems—could enable more streamlined risk reviews while maintaining security standards.
If defense organizations want AI operating at the pace of real-world operations, accreditation and governance models must evolve alongside the technology.
Why is metadata critical to successful AI implementation?
Cataloging data systems is only the first step. Metadata is what makes AI operationally useful.
Large language models and machine learning systems rely on contextual understanding. Without metadata—clear definitions of fields, lineage, classification, and usage constraints—AI systems cannot reliably interpret structured or unstructured data.
Metadata enables:
- Accurate text-to-SQL querying
- Secure row- and column-level access control
- Data lineage tracking
- Cross-agency discoverability
- Reduced duplication across data lakes
In short, metadata is the connective tissue between data governance and decision advantage.
How should defense organizations approach decentralized data and AI?
The future is not another centralized data lake.
A more scalable model is data-on-demand:
- Leave data in authoritative systems of record
- Govern it at the source
- Expose secure APIs for controlled access
- Deploy AI and analytics alongside the data
This reduces unnecessary data movement, preserves lineage, and lowers risk while still enabling enterprise-wide insights.
AI should not require duplicating data. It should operate securely on top of existing systems.
What lessons should government leaders consider when deploying agentic AI?
Agentic AI introduces both powerful capabilities and new risk considerations.
Agents that can carry out predefined automated actions—rather than simply generate insights—must be deployed with:
- Clearly scoped permissions
- Restricted API access
- Network isolation
- Egress controls
- Transparent reasoning and interpretability
At an industry-wide level, early experiments with broad agent permissions have highlighted the importance of scoped access and safeguards.
The lesson: agentic AI should amplify human decision-makers—not operate as opaque, unconstrained systems.
Interoperability standards such as agent-to-agent communication frameworks and Model Context Protocols will also play a key role in enabling coordinated, secure deployments.
What does “AI at operational speed” really mean?
It means moving beyond pilots and proofs of concept.
Operational AI requires:
- Secure, governed data foundations
- Scalable infrastructure
- Streamlined accreditation
- Interpretable agent orchestration
- “Human-centered” decision support
When these elements align, AI shifts from experimentation to measurable mission outcomes—supporting readiness, sustainment, and decision advantage across the force.
How can the Department recruit and retain top AI talent?
Mission matters—but so does the environment.
AI engineers want:
- Meaningful, high-impact work
- Access to modern tools and compute technologies
- The ability to experiment without months-long access delays
- Clear career progression and learning pathways
- Leadership commitment to AI adoption
When friction—such as lengthy GPU access requests or outdated tooling—slows innovation, talent retention becomes difficult.
Reducing bureaucratic barriers and enabling rapid experimentation will be essential to attracting AI professionals capable of delivering operational impact.
