US federal agencies bet on AI and automation to tackle the explosion of digital records
Artificial intelligence has gone from a distant promise to a real necessity inside the American government. With data volumes growing at breakneck speed, US federal agencies are facing a challenge that traditional processes simply can’t handle on their own anymore.
Emails, videos, chats, diplomatic cables, internal records… all of it forms a mountain of unstructured data that only keeps growing. And the pressure is coming from every direction: from the public, which wants quick answers, and from within the agencies themselves, which need to stay compliant with federal information retention and disclosure laws.
During a panel hosted by Federal News Network and sponsored by Casepoint, representatives from agencies including the US Army, the State Department, the FDIC, and Washington Headquarters Services within the Department of Defense came together to talk openly about how they’re using automation and AI to modernize digital records management.
Timothy Kootz, deputy assistant secretary for shared knowledge services in the Bureau of Administration at the State Department, summed up the situation pretty well during the panel: the volume of information is growing faster than the people and traditional records management processes can keep up with. And on top of that, the public now expects information to be accessible, searchable, and available quickly.
The expectation is that data volumes will grow by as much as five times over the coming years, driven by the expansion of AI itself within the government. In other words, the problem isn’t going away. It’s going to scale.
The good news is that solutions are already in motion, and the early results are pretty encouraging. 🚀
The weight of unstructured data at federal agencies
When people talk about unstructured data, a lot of them picture messy files sitting in some forgotten folder. But the reality at US federal agencies is way more complex than that. We’re talking about decades of diplomatic correspondence, meeting recordings, transcripts, field reports, and a huge variety of digital formats that were never designed to be easily indexed or searched. Each file type has its own structure — or rather, its own lack of structure — which makes any attempt at manual organization nearly impossible to scale efficiently.
The State Department, for example, deals daily with diplomatic cables that arrive in varying formats, often without standardized metadata, without classification tags, and without any clear indication of when that content could or should be made public. This creates a massive bottleneck in declassification processes, which traditionally depend on human analysts to review document by document, line by line. With ever-growing volumes arriving every day, that model simply can’t sustain the current demand, let alone future demand.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The FDIC faces a similar challenge, just in the financial world. Richard Huffine, assistant director of information and corporate records management at the FDIC, described a massive explosion of unstructured data that includes emails, chats, videos, and other media, all managed according to federal records retention schedules. Bank inspection records, internal communications, compliance reports, and transaction data pile up in repositories that were built at different times, with different technologies, and without any integrated vision for records management. The result is a fragmented environment where finding a specific piece of information can take days or weeks.
Huffine highlighted that the FDIC has implemented a program called capstone, which allows them to understand, based on each person’s role, what records they’re likely to generate within a broad retention cycle. But the work goes beyond that: the agency is also looking for ways to surface the most critical knowledge, so people aren’t working with 18 different versions of the same document, but rather with the most reliable and relevant version for the long term. The idea is to separate what truly tells the story and documents the decisions made at the FDIC from what can be discarded or prioritized differently.
How automation is changing records management
Automation entered this space not as a magic bullet, but as an intelligent layer capable of handling the volume, speed, and variety of data that agencies produce. Tools powered by artificial intelligence, especially those that use natural language processing and machine learning models, are already being applied to automatically categorize documents, identify sensitive information, and suggest classification levels based on historical patterns. This drastically cuts the time an analyst would need to spend on repetitive, low-value tasks, freeing up human attention for decisions that actually require judgment.
In the US Army, the adoption of automated workflows for digital records management is already delivering concrete results. Carrie McVicker, executive director of the Enterprise Services Agency in the Army, explained that one of the most significant initiatives was embedding a records management specialist directly inside the declassification facility. The reason is pretty practical: during modernization efforts, the team realized that agencies are often their own worst enemies when it comes to pulling records and reviewing them for declassification.
Placing a specialized person inside the operation and strengthening the relationship between records management and declassification was critical for moving modernization forward. McVicker pointed out that this directly impacted the declassification architecture and the way data is received and sent through the facility. Documents that used to take weeks to be sorted and routed now move through more efficient pipelines, with retention policies properly applied and auditable traceability across the entire chain of custody.
Another point that became clear during the panel is that automation isn’t being used to replace human reviewers, but to make their work more focused and efficient. AI systems act as a first filter, separating what’s clearly irrelevant, what can be processed automatically, and what genuinely needs human review. This smart triage transforms a process that used to be linear and slow into something much more dynamic, where human specialists concentrate their energy on the most complex, highest-impact cases.
Declassification at scale: the central role of AI
Declassification of government documents is one of the most delicate and time-consuming processes at any federal agency. It involves reviewing each document to ensure no still-sensitive information is exposed, while at the same time fulfilling the duty of transparency with the public. Historically, this process was done almost exclusively by experienced analysts, who needed deep knowledge of current classification policies and the ability to spot contextual nuances that could easily go unnoticed. The problem is that this model doesn’t scale, especially when the volume of documents eligible for declassification grows exponentially every year.
This is where artificial intelligence shows its greatest potential in the federal context. Kootz pointed to the successful pilot for declassifying diplomatic cables using AI at the State Department as a concrete example. According to him, the AI tools already in use can perform declassification reviews faster and more consistently than a human team. And most importantly: these tools make it possible to scale the process.
The State Department expects the cable declassification workload to grow fivefold in the coming years. Without AI, it would be simply impossible to keep up with that pace. With the current tools, the agency can meet the growing demand without needing to expand its team at the same rate.
One of the biggest wins highlighted during the panel was the reduction of inconsistencies. When humans review documents in large volumes, fatigue and subjectivity inevitably lead to different decisions for similar cases. AI applies the same criteria consistently across all documents, regardless of volume or the time of day the processing happens. This doesn’t eliminate the need for human review in the most complex cases, but it ensures a much more uniform and reliable baseline for the entire records management process.
AI agents and the future of records management at the Department of Defense
J.D. Smith, deputy director of the Executive Services Directorate at Washington Headquarters Services within the Department of Defense, brought a particularly ambitious vision to the panel. According to him, his team is pursuing a fundamental shift: moving away from manual work in many aspects of records management and transitioning to a model powered by AI agents.
These agents, Smith explained, can operationalize policies, file plans, and records disposition schedules. The idea is for the human focus to shift toward oversight and program administration, rather than being buried in heavy, operational tasks like being the person who manually files documents.
Smith described two related problems his team is tackling simultaneously. The first is how to better manage existing records within the Office of the Secretary of Defense. The solution being developed involves deploying AI personas across the organization, mapped to records disposition schedules, that will operationalize the currently manual process. This includes:
- Automatically identifying and categorizing records
- Classifying and filing documents
- Executing preservation holds
- Responding to information requests and searches
- Performing e-discovery operations
- Incorporating missing metadata that matters for both operational needs and data leveraging
The second problem is even more interesting: how to manage records at the moment they’re created, rather than after the fact. Smith explained that the team is developing AI-enabled copilots, configured similarly to a filing assistant, combined with AI bots and AI-enhanced RPA. These systems would be capable of analyzing content as it’s being created.
In practice, this means that while someone is drafting a document, the system could display a prompt like: it looks like you’re working on an international legal agreement. If that’s the case, here are the file numbers to select. With one click, the system files the document, embeds the necessary metadata, and manages the entire lifecycle of that record from its origin. That’s a massive leap from the traditional model, where records management typically happened long after the document was created.
Data challenges that cut across multiple functions
The exponential growth of data doesn’t just affect records management on its own. It has direct implications across multiple mandates within an agency, including processes related to the Freedom of Information Act (FOIA), internal investigations, and litigation. Kelly Swank, vice president of government business development at Casepoint, pointed out that many agencies are starting to look at the full data lifecycle, precisely because the datasets from different functions feed into each other.
According to Swank, it’s essential for agencies to have repeatable, automated, and defensible workflows when dealing with data across the organization. Casepoint, for its part, is integrating AI into its tools to help streamline processes and automate many functions, significantly reducing human review time compared to what was standard before.
This integrated view is crucial because a single document can be relevant to a FOIA request, an internal investigation, and a declassification process all at the same time. If each of those functions operates with siloed systems and independent manual processes, rework is inevitable and the chance of inconsistencies goes up considerably.
What’s coming next
With the expansion of AI use within the federal government itself, the trend is for the volume of data produced to grow even faster over the coming years. Documents generated by automated systems, interaction logs with AI assistants, records of algorithmic decisions… all of this will be added to the already enormous stockpile of unstructured data that agencies need to manage. This reality makes investment in robust digital records management infrastructure even more urgent — infrastructure that can absorb this growth without collapsing under its own weight.
The good news, as the Federal News Network panel made clear, is that agencies aren’t waiting for the problem to arrive before they start acting. The projects already underway at the Army, the State Department, the FDIC, and Washington Headquarters Services show that there’s both the political will and the technical capability to modernize these processes in a structured and responsible way. The path ahead involves real challenges around integrating legacy systems, training teams, and ensuring that the AI models being used are auditable and explainable enough to withstand regulatory scrutiny.
But what’s happening right now is, without a doubt, a turning point in the way the US government views and handles its own data. The combination of artificial intelligence, automation, and modern records management strategies is laying the groundwork for a government that’s more agile, more transparent, and better equipped to meet the demands of a society that increasingly expects fast and reliable access to public information. And that’s definitely something worth keeping an eye on. 👀
