Incorporating artificial intelligence (AI) into public sector services presents a transformative opportunity to enhance service delivery, increase access, improve efficiency, and support decision-making. Initiatives involving AI capabilities such as generative AI, machine learning and natural language processing offer significant advancements in automation and modernization of government operations.
However, realizing these benefits requires a well-structured, risk-aware and human-centered approach. This is especially true in health and human services (HHS), given the vulnerable populations served and the critical importance of access to essential services such as housing, food assistance and more.
Cúram has created a framework for a comprehensive, step-by-step process to ensure that AI applications are effectively and ethically integrated into human services, prioritizing the needs and experiences of users.
Our main goal is to provide a data-driven guide for the systematic identification, evaluation, and implementation of public sector AI use cases that are aligned with business objectives and maintain a strong focus on human-centered principles. By engaging a diverse group of stakeholders—including end users, policymakers, legal representatives, and technologists—this approach ensures that multiple perspectives are considered, and issues are identified quickly through real world experience while fostering solutions that are both practical and impactful for public services.
Start with well-defined business goals and objectives. Rather than shoehorning technology into plans, focus on the outcome: improving efficiency, enhancing service delivery, streamlining access, or supporting decision-making.
Evaluate each use case to determine if the technology will deliver the expected benefits. Consider:
Evaluate each use case to determine whether your organization is willing to accept the risk. Consider:
Develop a scoring system to rank each use case including at a minimum the following three dimensions: perceived business value (outcomes), technical feasibility, and risk. Starting with the smaller, lower-risk use cases, where a human is kept in the loop, is advised.
Implement selected pilot AI projects for the top-ranked use cases to test their effectiveness in a controlled environment. Ensure that pilots include mechanisms for users to easily report functionality issues and provide input. Also consider metrics and feedback to ensure validation of transparency, data security and privacy. In this phase the goal is to test, learn, and iterate.
Include extensive user testing phases and ongoing user engagement in implementation. Monitor, evaluate, and continually improve - conduct continual oversight and regular audits to ensure ongoing impact, user satisfaction, accuracy, adherence to principles and quality standards, and minimization of bias.
By leveraging this practical framework, public sector organizations can harness the potential of AI towards their digital transformation goals in delivering benefits that are both meaningful and sustainable. Whether assessing existing or emerging technologies around AI, considerations around outcomes, risk and feasibility are key to responsible use in the adoption of artificial intelligence in health and human services and are in the interest of public safety when it comes to protecting the needs of the communities that governments serve.
For deeper insights on how AI can be used in the public sector, specifically in health and human services, read the whitepaper from Cúram. The paper explores case studies on uses of AI that can help support government caseworkers deliver crucial services to constituents faster, help young people secure a better future, and help developers to efficiently support government modernization.