Building Maryland’s Data Stewardship Capacity in the Age of AI: From Data to Decisions That Matter

What does it take to turn data into better public decisions? And how can governments build the institutional capacity to responsibly reuse data in an increasingly AI-driven world?

Over the past few weeks, we at The GovLab have had the pleasure of working with chief data officers, chief information officer,s and other public servants from across Maryland state government to explore questions about data’s role in public decision-making. We examined these questions through an intensive, hands-on data stewardship course—what we call the Data Stewardship in the Age of AI bootcamp.

Organized in partnership with the Maryland Department of Information Technology (DoIT) and delivered through the Data Stewardship Academy, an initiative of The GovLab and The Data Tank, the program brought together participants from across state agencies and functions, including chief data officers, analysts, chief privacy officers, chief information officers and others involved in data governance, policy, and service delivery.

Held over four intensive days the program was designed around the premise tha Governments do not just need more data, better technology, or more sophisticated AI. They need people and institutions capable of translating public needs into responsible data use—and data insights into meaningful action.

That is the role of the strategic data steward. And that is what we helped cultivate among participants over the course of the bootcamp through presentations, guest lectures, and tailored exercises. Using our Data Stewardship Canvas as guidance, we led the cohort through modules focused on defining the demand and building data-to-action capability. We led this training using materials meant to make data stewardship practical and emphasize the importance of institutionalization.

Starting with Questions: Defining the Demand

Government agencies are surrounded by large and growing volumes of data—albeit fragmented across departments, subject to different access conditions, and rarely used for purposes where it could be most useful. Many governments have adopted efforts like Maryland’s Data Exchange to make data easier to find and reuse.

One of the central messages of the program was that effective use of data begins not with data collection but problem definition—the demand for data, not its supply. Too often, organizations start by asking what data they possess or what technologies they could deploy. We encouraged participants to reverse that logic:

• What public problem are we actually trying to solve?

• What questions would help us understand that problem?

• What decisions would better evidence enable?

• What is the minimum viable data needed to answer those questions?

• Who holds that data, and under what conditions could it be responsibly reused?

Through this sequencing, participants were encouraged to focus on not on questions of supply (“Where can I get data?”) but on the ways that data can make a real, meaningful difference for decision-makers. We encouraged participants to understand the root cause of the problem they hoped to address and the possible outcome—that being the decision they hoped to inform.

The emphasis on formulating better questions before deciding what data to collect, share, or analyze is critical in the age of AI. As generating answers becomes easier, the ability to formulate questions that matter becomes an even more valuable public-sector capability.

Building the Full Data-to-Action Capability

Formulating questions was the first part of several days of presentations, guest lectures, and exercises. Drawing on our past experiences supporting data stewards, we helped bootcamp participants explore different dimensions of strategic data stewardship:

1. Defining the problem and the questions that matter. We explored how to move from broad policy ambitions to specific, decision-relevant questions, identifying stakeholders and distinguishing between descriptive, diagnostic, predictive, and prescriptive inquiry.

2. Understanding data supply and AI readiness. Participants learned to map data ecosystems, identify minimum viable datasets, and assess data against FAIR-R principles—extending traditional FAIR considerations to include responsible, AI-ready reuse.

3. Making the business case for collaboration. Access to data is rarely just a technical matter. It requires understanding incentives, demonstrating mutual value, and making the case for why a data holder should collaborate. We used the 9Rs Framework to explore these questions.

4. Designing fit-for-purpose governance. Participants examined data collaboratives, commons, trusts, and other operating models, alongside the 4Ps of data governance: Purpose, Principles, Processes, and Practices.

5. Managing risks across the lifecycle. We considered privacy, bias, re-identification, model misuse, provenance, and the importance of safeguards that extend beyond a one-time compliance assessment.

6. Moving from insight to action. The final sessions focused on decision intelligence, impact measurement, institutionalization, and the often-neglected last mile: ensuring that data actually informs decisions and produces public value.

Making Data Stewardship Practical: Maryland’s Housing Challenge

One of the most interactive aspects of the bootcamp was the opportunity to bring together professionals from across state government to solve real, practical problems facing their agencies. Data stewardship cannot succeed if it remains confined to individual agencies, isolated technical teams, or a handful of designated data officers. Many of the most pressing public problems—and the data required to address them—cross organizational boundaries.

To foster this concept, we decided to develop an exercise that focused the training on a real problem facing Maryland: the state’s ambition to drive toward functional zero chronic homelessness, aligned with Priority 6 of the Moore–Miller Administration's State Plan: making Maryland a desirable and affordable home for all residents.

The challenge illustrated why data stewardship requires more than simply identifying a dataset.

Understanding and addressing chronic homelessness may require combining information from housing programs, social services, health systems, eviction records, and other sources—often held by different organizations and collected under different conditions. Participants were challenged to consider:

• Which individuals and households face the greatest risk of experiencing chronic homelessness?

• Where could preventive interventions have the greatest impact?

• What data could help assess whether housing assistance programs are working?

• How could agencies responsibly combine data while protecting privacy and preventing discrimination?

The exercise asked participants to develop a Data Reuse Brief: identifying priority questions, mapping potential data sources, examining metadata, and surfacing the unresolved issues that would need to be addressed before reuse could begin. Importantly, the objective was not to produce a predictive model or another dashboard. It was to understand the institutional, operational, and governance conditions necessary to make better evidence-informed decisions possible.

Through this work, participants were able to see what it means to be a data steward. The strategic data steward is not simply a custodian responsible for protecting or documenting information. Nor is the role limited to ensuring compliance. A strategic data steward connects demand with supply, enables responsible collaboration, anticipates risks, and helps translate data into decisions and public value. That capability becomes particularly important as governments increasingly explore AI-assisted decision-making.

A Broader Lesson for Government in the Age of AI

Our experience in Maryland reinforced a conviction that has guided much of our work at The GovLab and The Data Tank: Data governance is foundational to AI governance. But data governance itself must also become ready for AI.

That transformation is not just about standards, platforms, or new technologies. It is about developing the human and institutional capacity to ask the right questions, mobilize the right data, establish legitimate governance arrangements, and act on evidence responsibly. In other words, the future of data-driven government depends as much on strategic stewardship as on technical infrastructure.

The ultimate measure of success is not how much data is shared, how many dashboards are produced, or how many AI systems are deployed. It is whether governments become better at identifying the questions that matter, making informed decisions, and improving people's lives.

It is about moving from one-off exercises to sustained institutional practices that can inform decision-making. That kind of lasting change requires organizational mandates, incentives, workflows, leadership support, and communities of practice that enable stewards to act.

***

The GovLab thanks the Maryland Department of Information Technology for the partnership and to all participants for their engagement, questions, and willingness to apply new approaches to real government challenges. We are also grateful to our guest faculty and practitioners—Dave Gottesman, Oliver Wise, Casey Weston, Brennan Lake, Nick Hart, and Nancy Potok—for bringing valuable perspectives on public-sector data use, AI readiness, risk, and institutionalizing stewardship.

We look forward to seeing how this cohort continues to advance responsible data reuse across Maryland.

For more information on the Data Stewardship Academy, its tools, training, and programs, visit DataStewards.net.