Presented by EY
As federal agencies adopt artificial intelligence and pursue more ambitious modernization initiatives, the role of the chief data officer is rapidly expanding. Data governance remains important, but today’s CDO must also connect information across organizational boundaries, improve decision-making, support new technology and help employees understand their individual responsibilities within an enterprise data environment.
That changing mission is the focus of “Beyond Governance: The New CDO Agenda,” an episode of Modernizing Government: The EY Insight. Marseta Dill, Acting Chief Data Officer at the Federal Aviation Administration, and Kamna Bohra, Senior Manager at EY, join Fed Gov Today host Francis Rose to discuss what it takes to build a data-first federal organization—and why the work begins with people and processes, not technology.
For Dill, a data-first organization considers information from the beginning of every modernization effort. When an agency develops a new application or system, its leaders should not focus exclusively on functions and technical requirements. They must also understand what data the system will collect, how that information will move through its workflows and how it will support the agency’s mission.
Dill compares this approach to the evolution of cybersecurity. Cybersecurity can no longer be added after a system is built; it must be incorporated during planning and design. Data should be treated the same way. Considering it early changes the questions development teams ask and helps them understand how the system will operate with the information it uses.
That responsibility also extends beyond employees with “data” in their titles. Every person who creates, analyzes or uses information has a part to play in maintaining its value.
Bohra says this has become even more important as AI tools become widely available across civilian and defense agencies. Longstanding problems with data can now spread or intensify much faster. That places additional pressure on chief data officers to establish the standards, guardrails and interoperability necessary for employees to use information successfully and safely.
Common challenges include determining whether data is visible, accessible, understandable, interoperable and secure. Much of the government’s valuable information also remains stored in spreadsheets, PDFs, databases, presentations and other disconnected sources.
Those files should not simply be dismissed as evidence of poor data management. Bohra argues that they frequently demonstrate healthy problem-solving behavior. Someone encountered an operational need, used the tools available and produced an analysis valuable enough to influence decisions. In some cases, that analysis may become more trusted by decision-makers than the formal system of record.
The challenge is to bring that value into an enterprise environment without losing the knowledge and business context that made it useful.
Dill says the FAA works with employees to understand the functions their locally maintained information supports while helping them recognize how their data connects to the rest of the agency. Enterprise platforms can make that information more accessible and discoverable, but the larger objective is to create connections among data, people and missions.
That requires a shift from data ownership to data stewardship.
“Data is an enterprise asset,” Dill says. The goal is not to take information away from the teams that create and understand it. Instead, the CDO can serve as a partner who sees relationships among projects and connects data stewards with counterparts elsewhere in the organization. Those conversations help teams identify shared benefits and align around common mission outcomes.
This is particularly important in a federated organization such as the FAA. Dill says the agency has spent several years breaking down silos and helping employees take advantage of enterprise tools. Although technology supports that effort, the largest challenges are often cultural.
Bohra says a chief data office must avoid becoming known primarily for compliance requirements, reporting requests and additional administrative work. The office should be additive—providing tools, playbooks and guardrails that allow teams to operate safely while creating forums for substantive conversations about mission use cases.
Deploying another platform will not accomplish that on its own. Agencies must also redesign their business processes and operating models to account for what current and emerging technologies make possible. Because those technologies are evolving so quickly, organizations need processes flexible enough to accommodate continued change.
Federal agencies also have new opportunities to extract value from unstructured data. Technological advances have made it easier to work with information contained in documents, meeting notes, presentations and multimedia. That material often preserves something transactional systems do not: the institutional memory of how and why decisions were made.
Technology, however, cannot interpret that information effectively without human expertise. Bohra notes that an agency can place a document repository into a large language model, but only a domain expert can define what meaningful information should be extracted and determine whether the resulting output is useful. Without that expertise, the volume of unstructured material can generate noise rather than insight.
Dill similarly looks for the people with the strongest operational knowledge when discussing a new initiative. Their contributions often reveal valuable underlying data and analysis that may not have been identified formally. The resulting “data products”—information that has been refined, contextualized or analyzed for a particular purpose—can be more valuable than the raw data alone and should be shared where they can support other teams.
This combination of information, analysis and institutional context becomes especially important when agencies prepare for AI.
Dill recommends that organizations begin by asking whether their people—not only their data—are ready. Employees need a realistic understanding of what AI can and cannot do. Leaders must determine whether the workforce trusts the technology, whether teams are open to exploring its potential and whether an initiative has the appropriate mix of data scientists, technologists and business experts.
Most importantly, agencies should begin with the mission problem.
The question should not be where an agency can add AI to an existing process. Leaders should define what they are trying to accomplish and then reimagine what the process could become with AI serving as an enabler.
Bohra identifies “decision intelligence” as the next major opportunity. Decisions incorporate data, policy, risk, stakeholder considerations, outputs and, most importantly, human rationale. That rationale is rarely captured alongside the underlying data. It may instead remain in a meeting, presentation or memorandum.
Consistently documenting those elements as metadata could reveal patterns in how an organization makes decisions. It would also help employees understand where particular data has contributed to mission outcomes, making that information more valuable for future modernization and AI initiatives.
This work ultimately becomes a business-process exercise as much as a technology project. Agencies must understand how information flows among applications and organizational functions. By rationalizing those connections, they can reduce complexity, simplify workflows and improve how data is curated, shared and protected.
Cross-functional collaboration is essential. Bohra says strong use cases emerge when developers, data professionals and mission experts work together. Agencies may also need apprenticeship models that allow employees with different technical and operational skills to learn from one another.
Looking ahead, Dill says the FAA is examining how successful solutions can be reused across the enterprise. As the number of AI use cases grows, agencies can identify common challenges and enable teams to build on work already completed elsewhere rather than creating isolated solutions repeatedly.
Bohra believes agencies must also continually reconsider the design of their organizations. Hiring technical talent or automating individual roles will not be enough. Leaders must imagine the organization of the future with data, technology and evolving workforce requirements built into its operating model.
That model may not remain appropriate for years—or even months. The new CDO agenda is therefore not a single governance program or modernization project. It is an ongoing effort to connect information, preserve institutional knowledge, redesign operations and prepare the workforce for continuous change.
