“You can build the best early warning detection capability in the world, but if the people in your organization don’t act on the warnings, what’s the point?”

For Elizabeth Puchek, Chief Data Officer (CDO) at the U.S. Consumer Product Safety Commission (CPSC), that observation captures a reality many public-sector leaders face as they pursue data-driven transformation. Technology alone does not prevent harm, instead, prevention happens when high-quality data, effective governance, responsible AI, and operational decision-making come together to drive action.

As agencies confront complex public safety challenges, the focus is shifting from retrospective reporting to proactive risk detection. The ability to identify emerging hazards earlier, connect information across traditionally siloed systems, and enable faster decision-making has become a strategic priority across government. 

Yet achieving that vision requires more than advanced analytics.

It demands trusted data-sharing frameworks, modern infrastructure, clear accountability, and a workforce prepared to work alongside AI-enabled systems.

In this conversation, Puchek shares how CPSC is building the foundations for preventive analytics, including efforts to improve interoperability through health data exchange frameworks, modernize data architecture, and explore AI-powered signal detection capabilities.

Edited Excerpts 

Q: Cross-agency collaboration is often essential for preventive action. How are you thinking about improving interoperability and data-sharing across agencies while still maintaining governance, privacy, and security standards?

I am fortunate to have entered a domain that has been tackling this problem for years. The U.S. Department of Health and Human Services (HHS) Office of the National Coordinator for Health Information Technology (ONC) leads nationwide interoperability efforts and has developed the Trusted Exchange Framework Common Agreement (TEFCA).

It aims to establish a unified governance, policy, and technical framework for secure, standardized health information exchange across federal agencies and healthcare organizations. This framework provides common rules of the road and a trust architecture that allows organizations to reliably share data while maintaining strict privacy and security protections.

In 2024, ONC and the Recognized Coordinating Entity (RCE) released the first version of the Exchange Purpose Implementation Standard Operating Procedures (SOP) for Public Health. This SOP allows access to health information to support core public health services. By participating in TEFCA and partnering with hospitals, I hope to improve efficiency in investigating and addressing health hazards and root causes to improve public safety.

Q: Real-time analytics is becoming increasingly important in public safety environments. What challenges do agencies face in building real-time decision capabilities, and how can leaders realistically begin modernizing toward that future?

Public safety relies on diverse, voluminous data: from health records to public reporting to the news. AI can streamline data normalization and quick signal detection far more efficiently than manual methods.

Leaders can and should examine the guardrails required to responsibly use AI to accelerate this process. We also need to recognize that diving into AI requires investment, not just in acquiring models or tools or services but also in the current workforce. 

We don’t need to transform staff into AI engineers or developers, but we do need folks to understand how AI works because they are integral to its implementation. Whether there is a human in the loop to perform quality control, to train the models, or to express requirements; we are responsible for leading our workforce through this transformation. This requires training for our current staff and recruitment for targeted positions throughout the organization.

Q: AI is rapidly changing how organizations detect risk and prioritize action. What use cases at the CPSC do you believe hold the greatest potential for improving preventive public safety outcomes over the next few years?

I am excited about the potential that lies in the data here. We ingest data from a great variety of sources, but evaluation of this data is still largely siloed and manual. I am so excited to begin the hard work of engineering the infrastructure that will allow us to access this data all at once and at the speed of mission.

Not only do we have to perform a lot of data imputation and extraction from unstructured sources, but we also have to perform a lot of schema standardization between sources. I also look forward to building a graph database and resolving all of this data around a few entities like product type, manufacturer, and more, so that we can gain a more holistic view of these actors.

With that infrastructure in place, natural language processing models will allow officers and inspectors easier access to this data and drive faster, more informed decision-making.

Q: Are there any emerging initiatives, pilots, or technology directions you’re particularly excited about that could significantly improve how agencies anticipate and respond to public safety risks?

Agencies are beginning to move from reactive reporting toward truly anticipatory public safety capabilities, and several emerging directions make that shift realistic. Advances in AI‑driven signal detection are especially promising.

Tools that normalize and analyze large, messy data inputs (from clinical narratives to consumer submissions) can reveal emerging hazards far earlier than manual review ever could. Pilots under way across the federal landscape demonstrate this potential:

• Early‑warning models that integrate injury data, e‑commerce trends, and import information
• Real‑time analytics architectures that combine cloud‑based ingestion pipelines with machine‑learning‑ready datasets
• Predictive modeling proof‑of‑concepts designed to flag emerging hazards before they escalate

What excites me most are the technologies that directly support prevention:

• Integrated data environments that cut across historically siloed systems and structured and unstructured data
• Predictive and NLP‑based models that accelerate early hazard recognition
• Modernized architectures that enable real‑time decision‑support

Together, these create the foundation for agencies not only to detect risk sooner but also to respond with greater precision, clarity, and public trust.

Q: For agency CDOs looking to strengthen preventive analytics capabilities, what foundational investments or organizational priorities would you recommend focusing on over the next 12 to 24 months?

As with any transformational initiative, you need a PLAN. It may sound like a luxury, but you are going to save so much time and money in the long run if you’ve taken the time to consider the requirements, constraints, and risks that you’re dealing with.

As you’re planning, don’t forget to build in time for the following: 

• Data quality: As CDOs, we are constantly having to remind leaders how important data quality is. Hopefully that message is starting to sink in because it’s critically important for preventative analytics to be successful.
  Everyone wants to invest in analytics, but it takes investment in time and expertise (technical and business) to integrate messy, unstandardized, and multi-source data and build pipelines that deliver clean, consistent, and well-governed metadata.

• Signal detection: Teams want to go straight to model selection and implementation, but you can’t do that unless you have mapped out what an actionable early detection looks like for your mission. What are your indicators and leading/lagging relationships?
• Responsibility: You can build the best early warning detection capability in the world, but if the people in your organization don’t take action on those warnings, then what was the point? The process to identify, elevate, and act on warnings needs to be thoughtfully designed and implemented to yield intended results.

About Elizabeth Puchek:

Elizabeth Puchek serves as the Chief Data Officer for the U.S. Consumer Product Safety Commission, where she leads enterprise data modernization and innovation efforts that strengthen the agency’s mission of keeping consumers safe. Elizabeth oversees the development of trusted data assets, advanced analytics capabilities, and forward‑looking strategies that support evidence‑driven decision‑making across the organization.

Before joining CPSC, Elizabeth was the CDO at the U.S. Citizenship and Immigration Services (USCIS) and held several key leadership positions at U.S. Customs and Border Protection (CBP). At USCIS, she advanced enterprise data governance, analytics modernization, and streamlined business processes in support of critical immigration services and policymaking.

At CBP, Elizabeth led initiatives that enhanced operational intelligence and mission‑readiness across complex border and trade environments. These roles shaped Elizabeth’s expertise in transforming large federal data ecosystems and delivering actionable insights for high‑stakes, public‑sector missions.