Jeffery Marino, Director of the Office of Data and Innovation (ODI), delivered a keynote at the Forum on Emerging Technologies, discussing the importance of a product management mindset in government service delivery. His address highlighted the lessons learned from past IT failures, the role of generative AI, and the need for iterative development approaches to improve public sector technology projects.

Marino began by referencing the well-documented launch failure of healthcare.gov in 2013, describing it as a case study in execution challenges rather than technological shortcomings. He explained that the website, designed to serve as the entry point for millions of Americans to access health insurance under the Affordable Care Act, was built under a rigid political deadline, with little consideration for the user experience. “The site wasn’t designed for Americans to use; it was designed to fulfill a requirement,” he noted, emphasizing that the project prioritized completion over usability. The fragmented approach, with over 50 contracts involved, further complicated the development process, leading to a flawed launch and eroded public trust.

Despite the initial failure, the response to the crisis laid the foundation for change. Marino pointed out that the remediation efforts, which involved a combination of government leaders and private sector experts, led to the creation of the U.S. Digital Service and other initiatives focused on improving government technology delivery. He drew a direct connection between that effort and the work being done at ODI, CDT, and by industry partners today, stating that “teams like ours have been working to introduce new tools and methods to ensure that users are involved early and often.”

Transitioning to the topic of Gen AI, he described it as a tool with “incredible transformational potential” for government services, particularly in its ability to simplify complex processes. He illustrated this by suggesting that an AI assistant could have improved the healthcare.gov experience by guiding users through the enrollment process, rather than requiring them to navigate a static and complicated interface. He also highlighted the benefits of “digital twins,” which allow governments to simulate system launches in a virtual environment to identify potential challenges before deployment. This approach, he argued, could address the risk aversion that often slows government innovation.

However, Marino stressed that the success of Gen AI in government is dependent not just on its capabilities but on how it is implemented. “Gen AI offers limitless potential, but its success depends much less on what it can do and almost everything on how we use it,” he said. He called for the adoption of a structured product management approach to ensure responsible and effective deployment. This includes engaging users early, building cross-functional teams, and iterating frequently rather than waiting for perfection. He warned that “waiting for perfection looks like disaster” and emphasized that smaller, incremental failures are preferable to large-scale breakdowns.

He also emphasized the importance of measuring outcomes rather than outputs, explaining that success should not be based on the number of applications processed but rather on whether people are actually receiving the services they need. He called for investment in talent before tools, arguing that “we have to build the infrastructure that supports innovation, attract and train the best people to lead this transformation.”

Marino concluded by reinforcing that generative AI is not simply a new technology but a catalyst for broader transformation in government service delivery. He urged the audience to embrace an approach that prioritizes user needs, fosters collaboration, and values incremental progress over large-scale launches. “This moment of transformation demands that we champion incremental success over big bang releases and that we have the grace and the risk tolerance to have those hopefully incremental failures along the way as well,” he said.

Watch the keynote speech below: