Using AI to Improve Child Welfare

Across the country, child welfare agencies carry out a vital public mission amid growing complexity—supporting increasing caseloads, navigating evolving policy frameworks, and continually developing a workforce with skills to make consequential decisions for children and families.

At the same time, these agencies must document every action, comply with mandates from multiple levels of government, and deliver high-stakes outcomes. The gap between what the child welfare workforce has the responsibility to do, relative to what existing systems enable them to do, has widened in recent years. States have begun to adopt AI as a means to close that gap.

This report, Using AI to Improve Child Welfare: Navigating Rules in Real Time by David R. Schwartz, published by the IBM Center for The Business of Government in collaboration with the University of Michigan School of Social Work, addresses this issue directly. Drawing on insights from a national roundtable of state and local child welfare leaders, researchers, and practitioners, the author explores how responsible use of AI can support frontline social workers, supervisors, and agency leaders—by reducing administrative burden, improving access to policy and case information, and strengthening professional judgment and accountability.

This report reflects a growing recognition across government that successful innovation in human services must support the frontline workforce that serves families and children in need. The report presents insights about what AI can do, cautions about ethical risks, and discusses how technology can serve people.

The report makes clear that the promise of AI in child welfare lies not in automation of decisions about child safety, but rather in removing administrative burdens that have made this work increasingly challenging. The AI tools described in this report focus on answering policy questions in realtime, synthesizing complex case histories, assisting with documentation, and supporting training—all while keeping humans in the loop. These advances are grounded in early, practical implementations already underway in multiple states.

This report integrates historical perspective, applied research, and practitioner experience to frame a set of principles for responsible adoption. These principles can support relief rather than replacement, transparency over opacity, low-risk use cases before high-stakes ones, and governance structures that support emerging technology. From this foundation emerge clear lessons that can shape key imperatives, including local control of sensitive data and collaboration across agencies and disciplines.Importantly, the author underscores that outcomes in child welfare cannot emerge only from automation, but require informed design, governance, evaluation, and stewardship. AI systems can enhance progress toward these goals, but only with sufficient oversight, feedback loops, and continuous learning. When implemented thoughtfully, AI can enable the frontline workforce to spend less time navigating systems and processes and more time doing what humans do best: assess risk, support families, build trust, and keep children safe.

This report builds on the IBM Center’s longstanding body of work examining how effective use of technology, data, and management practices can strengthen government performance, including Responsible AI for Public Evaluation, AI in State Government, GenAI and the Future of Government Work, and AI and the Modern Tax Agency.

As governments continue to explore the responsible use of AI for service delivery, Using AI to Improve Child Welfare offers a grounded and pragmatic path forward. By showing a path to innovation that supports practitioners and the needs of vulnerable populations, this report contributes findings and recommendations to ensure that technology enhances public value at the heart of child welfare.