by Meredith Delaware
Federal agencies are rapidly adopting artificial intelligence (AI)-enabled tools. A January 2026 Google Public Sector survey of federal government IT leaders and influencers across civilian and defense agencies found 90% of respondents worked for agencies that are planning to or using AI. As agencies increase their AI use, they run into AI readiness challenges due to limited time, workforce resistance, unclear guidance, and fragmented data foundations. Data discovery bridges that gap. Data discovery is one of the lowest-cost, highest‑value investments agencies can make to accelerate AI maturity, enable their workforce, and uncover insights hiding in plain sight.
In Part 1 of this series, I discussed how dedicated time and energy for the workforce to experiment with AI are critical to increasing AI adoption. In Part 2 below, I explain how agencies are using small-scale data discovery projects to give their workforce additional opportunities to improve AI skills while evaluating their data for value and hidden insights.
What is Data Discovery?
Data discovery is the practice of finding, exploring, and understanding existing datasets to reveal patterns, gaps, and opportunities. It’s exploratory, not prescriptive. It differs from traditional analysis, which focuses on answering specific questions rather than independently exploring to uncover meaning in vast datasets.
Data Discovery as the Most Practical AI Classroom
Data discovery often begins as an inventory of information, but in practice, it becomes one of the most effective ways for teams to build practical AI fluency. Instead of learning from generic examples or hypothetical training data, employees engage directly with the information that drives their mission, case notes, logs, forms, reports, and years of program history. This turns AI from an abstract concept into something concrete and familiar. As teams explore real datasets, they naturally start asking the questions that underpin responsible AI use. What does this field actually mean? Can we trust this timestamp? Why does this pattern appear here but not elsewhere? What don’t I see that I expected to see here, and why?
Those conversations build an intuitive understanding of how AI systems interpret data, where the limitations are, and how they can bring their expertise and judgement to improve outputs. Some of the most valuable insights come from simple activities like:
- Brief Data Walkthroughs: Teams spend an hour reviewing one dataset and discussing what insights AI could surface. This exercise builds intuition: “Oh, AI could summarize these narratives!” and makes constraints visible: “We need consistent timestamps before automating this step.”
- “Where’s the Data?” Challenges: Leaders prompt teams to find relevant data for mission questions, highlighting gaps, fragmented sources, and the importance of improved governance.
- Baseline Analytics: Teams work through various baseline analyses to see what output the data yields, including distribution analysis, bottleneck detection, and basic variances for mission-specific use cases.
- Before/After AI Snapshots: Show a raw dataset vs. a simple AI-enhanced version (clusters, summaries, tags). Helps staff see how AI could unlock value from underused data. This reinforces the need for cleaner, better-structured data.
This kind of exploration also brings people with different perspectives into the same conversation. Mission staff, analysts, and IT teams each see something different in the data, and when they explore it together, they begin to develop a shared understanding of AI’s possibilities and limits. These moments often reveal opportunities as well as inconsistencies, contradictions, and gaps that could derail future AI efforts. Instead of treating those issues as obstacles, discovery helps teams acknowledge them early, discuss their impact, and plan realistic pathways forward. In many cases, uncovering messy or incomplete data becomes a productive learning moment: teams see firsthand why data readiness matters and how they can adjust their daily work to address the gaps.
Investing in data discovery for teams helps employees see how AI might help them work faster or understand the mission more deeply, and leaders gain a clearer sense of which use cases are feasible today and which require more foundational work. In this way, data discovery becomes far more than a technical prerequisite. It’s an accessible, low‑risk environment where employees strengthen their understanding of AI, leaders uncover opportunities grounded in real data, and teams collaboratively build the foundation needed for AI to succeed.
The Dual Benefit of Unused Data
Many organizations preparing for an AI future hoard vast amounts of data, expecting that someday this data will be needed to address a specific AI use case. This data often sits untouched indefinitely. If your organization does this, investing in workforce data discovery initiatives can provide the added benefit of facilitating data assessment, analysis, and insight generation.
Agencies sit on free text notes, workflow logs, surveys, and old reports that AI can finally make usable. Unlocking this value doesn’t require major investments, just structured time, tool access, and focus.
For example, the Department of Energy’s AskOEDI tool answers users’ questions about specific Open Energy Data Initiative datasets. As the Government Accountability Office highlighted in a 2025 report, AskOEDI allows users to inquire about the equipment, assumptions, and methodologies used in the origination of a dataset, as well as more abstract questions such as the applicability of data to specific research fields. DOE notes that this project provides access to big data via "data lakes", large cloud-based collections of open-access energy data available to anyone.
Small-Scale Data Discovery Projects Can Uncover Hidden Value
Low-investment experimentation can often uncover powerful insights. Many existing tools can achieve the dual goal of data discovery. Tools like Microsoft Copilot and the recently launched USAi evaluation suite can help agencies make sense of available data sources, test data-discovery techniques, and identify AI projects with the potential to advance agency goals.
Small-scale projects can be effective in returning a high ROI. Through experimentation, sprints limited to ten weeks or less, agencies set clear limits on the time invested in data discovery. PCI-GS has seen organizations find success by assigning “tiger teams”, including junior, intern, or new hire teams, to experiment with underutilized data.
One resource-strapped organization tasked a set of new data scientists and data engineers with using available AI tools to sift through unused data sets. Over the course of eight weeks, the team developed new analytics and insights that led to a new understanding of a pivotal supply chain challenge and changed the approach they were taking to that data collection.
Discover the Value Hiding in Your Data
To meet expectations for increased AI implementation in federal operations, agencies must increase AI fluency, improve data readiness, and mature AI-supported outputs. Incorporating data discovery projects can deliver value against numerous organizational objectives.
PCI-GS can help organizations get more out of their data discovery. We meet organizations where they are, providing the expertise you need to move within project constraints to achieve your AI goals. Whether you’re beginning to assess your existing data ecosystem or are ready to turn this information into action, our experts can help. Contact PCI-GS to move forward.
