AI pilots and use cases are proliferating across the federal government. Yet many agencies find that despite growing investment and experimentation, AI’s impact remains narrow. Systems support individual tasks, but mission-level transformation is harder to achieve. The core barrier is fragmentation: across systems, across workflows, and most critically, across data.
This fragmentation restrains impact because AI’s value depends on connection. When information remains siloed, AI cannot see the full picture of operations, risk, or opportunity. The result is slower analysis, duplicated effort, and poor return on AI spending. This article explores how fragmentation manifests in federal environments and why connecting information at the data layer is key to unlocking AI at scale.
DOJ Case Study: Adoption Without Integration
The Department of Justice (DOJ) illustrates both meaningful AI progress and the structural limits of fragmented implementation. DOJ reports 240 AI systems supporting roughly 100–110 distinct use cases, and about 70% of these systems are operational. On the surface, this suggests strong adoption momentum.
A closer look reveals fragmentation along two dimensions.
Fragmentation across components
- Fewer than 5% of DOJ AI systems operate department-wide
- Nearly 70% are confined to a single component
- Multiple systems perform similar functions in different parts of the department, including:
- 12 separate license plate reader systems
- 9 AI systems for audio and video transcription
When components such as the FBI, DEA, ATF, and OIG rely on different tools for similar tasks, insights remain harder to share. AI outputs are shaped by local systems and local data, limiting cross-component visibility and coordinated intelligence.
Fragmentation across workflows
- AI tools often address narrow tasks rather than end-to-end investigative processes
- Analysts may use multiple systems for extraction, translation, transcription, recognition, categorization, and analysis
- Data must be manually exported, reformatted, and re-ingested between tools
This creates friction inside the workflow itself. Even when AI accelerates individual steps, the handoffs between systems slow the overall process and introduce opportunities for error.
The Cost of Fragmented Information
These patterns are not just IT architecture issues. They have operational and strategic consequences.
When systems and data are not connected:
- Investigations and analyses move more slowly due to manual data movement
- Errors and inconsistencies increase as information is reformatted across tools
- Similar capabilities are purchased, maintained, and supported multiple times
- Insights stay localized within components instead of informing department-wide action
- AI remains task-specific rather than supporting full mission workflows
At the root of these issues is data fragmentation. Mission-critical information lives across emails, case files, transcripts, documents, and legacy repositories. When AI systems draw from isolated slices of that environment, they operate with limited context. AI fragmentation is often a symptom of deeper information silos.
Why Centralization Alone Falls Short
A common response to fragmentation is to centralize data into large repositories or data lakes. In federal environments, however, large-scale “lift-and-shift” migration is often:
- Expensive and time-intensive
- Operationally disruptive
- Prohibited by security, privacy, and regulatory requirements
As data volumes grow and environments become more federated, moving or copying everything into one place becomes less feasible. Silos persist, even as AI tools proliferate on top of them.
Connecting Information Without Moving It
An alternative approach focuses on connecting information where it already resides. Instead of replacing existing systems or migrating all content, agencies can create a unified layer of visibility and governance across distributed environments.
This approach is often described as in-place data management. Rather than treating silos as obstacles that must be dismantled, in-place treats them as sources to be connected. The approach extracts and indexes the essence of unstructured data — its metadata and content — while the original files and messages remain in their source systems. Email stays in email platforms. Case files remain in case management systems. Documents stay in file shares or cloud repositories. What changes is the agency’s ability to see, govern, and act across all of it as one unified information environment.
Through in-place management, agencies gain a virtual, unified view of fragmented data without the cost and risk of wholesale migration.
Key capabilities include:
- Records management: Classification, retention, and defensible deletion applied consistently across all agency repositories
- Policy enforcement: Automated, content-based tagging aligned with regulatory and mission requirements
- Risk management: Identification and remediation of sensitive data, including PII, and reduction of redundant, obsolete, or trivial (ROT) information
- Flexibility: Rapid reclassification and policy updates without re-indexing or restructuring underlying systems
Because governance travels across systems, agencies no longer have to manage compliance, retention, and risk one silo at a time. This is especially important for unstructured data, which represents the vast majority of federal information and includes the emails, documents, transcripts, and reports that AI systems increasingly rely on.
Turning Silos into an AI-Ready Information Fabric
By indexing and connecting unstructured data in-place, agencies effectively create a “virtual data lake” without consolidating everything into one repository. Instead of feeding AI models with whatever data is most accessible (which is often incomplete or poorly governed) agencies can deliberately search, curate, and deliver relevant, high-quality information from across the entire agency.
A connected, governed information layer enables:
- More complete data inputs: AI systems can draw from information across components and workflows, not just local sources
- Improved accuracy: Higher-quality, contextual data reduces the likelihood of misleading or incomplete outputs
- Reduced risk: Sensitive or regulated content can be identified and excluded from AI training or analysis as needed
- Cross-component knowledge discovery: Institutional knowledge scattered across documents and messages becomes discoverable beyond organizational boundaries
With in-place management, AI is no longer confined to the data repositories of individual systems, allowing models to operate on a broader, mission-aligned information foundation.
Enabling Workflow Continuity
One of the most important shifts enabled by in-place management is the move from isolated task automation to connected workflows.
When information is discoverable and governed across systems:
- Data does not have to be repeatedly exported and reformatted between tools
- AI services can be applied at multiple points in a process using a shared knowledge base
- Analysts spend less time moving data and more time interpreting results
- Workflows become more continuous, with fewer breaks between systems
This helps address the workflow fragmentation seen in environments like DOJ, where AI may support many individual steps but not the full investigative lifecycle.
A Structural Foundation for Scaling Federal AI
In-place data management offers a way to modernize without dismantling existing architectures. Agencies can preserve system ownership, security boundaries, and operational continuity while still gaining unified visibility and governance. Fragmentation at the system level does not have to mean fragmentation at the information level.
By connecting and governing data where it lives, agencies create the foundation for AI deployments that are more accurate, more secure, and more aligned to mission outcomes. In a federal landscape defined by distributed systems and massive volumes of unstructured data, the path to scalable AI is in-place, not wholesale centralization.
Ready to lay the data foundation for mission-driven, trustworthy AI? See how ZL Tech helps agencies leverage their unstructured data.
