Person-Centered AI in Human Services: Letting Technology Assist, Not Decide

Written By: Michael Bostian, Digital Government Solutions, Maximus

---

That balance is especially important for eligibility determination and case management where context, nuance, and downstream impact matter to the individuals and families awaiting benefit decisions.

Letting Humans Do What Humans Do Best

AI excels at recognizing patterns, extracting information, and processing large volumes of data quickly and consistently. What it does not do well yet is interpret meaning across complex circumstances or apply judgment in ambiguous real-world situations.

Consider the eligibility determination process. AI-driven tools can automatically ingest documents, summarize case histories, flag missing information, and compare case data against policy rules at speed. Generative AI–powered policy assistants can also help caseworkers quickly locate and interpret relevant policy sections or procedural guidance, reducing their time spent searching manuals and increasing consistency across staff.

But when a household’s composition is complex, income fluctuates, or a case involves non-standard expenses, good-cause exceptions, or hardship considerations, human judgment remains essential. AI can surface policy language or highlight potential discrepancies, but it cannot assess intent, weigh contextual factors, or determine when discretion is appropriate.

Predictive analytics can further support staff by identifying cases more likely to require additional review or pointing supervisors to emerging workload trends.

Used responsibly, these technology-enabled insights help agency and program leaders prioritize attention and resources. They should not, however, be used to automate final determinations or override professional judgment.

Trust Is Built Through Transparency and Familiarity

Challenges with AI adoption are rarely only technical issues. Staff comfort and confidence play a decisive role in whether new tools are embraced or resisted.

Case workers are more likely to trust AI when they understand how it works, what inputs it uses, and how its outputs should (and should not) be applied. Successful agencies invest heavily in change management for their programs. Early engagement of experienced staff, supervisors, and trainers allows them to translate AI tools into familiar, supportive concepts for their coworkers.

Everyday analogies can help demystify AI. Many people already rely on predictive interactions and automation without hesitation: smartphones that suggest the next word while typing, voice assistants that answer questions or perform simple tasks, or facial recognition that securely unlocks a device. In each case, automation enhances convenience, but users remain in control.

Positioned this way, AI in human services feels less like a black box and more like an extension of tools staff already trust—tools designed to reduce friction, not replace expertise.

Avoiding the “AI Everywhere” Trap

SNAP surge response illustrates this balance in practice. After a federal waiver expired, one county experienced a rapid influx of SNAP employment and training referrals that risked overwhelming staff capacity. The program used staff redeployment, work optimization, and digital outreach to absorb the volume and avoid a backlog. Robotic process automation then reduced thousands of hours of repetitive tasks, while caseworkers retained responsibility for judgment, oversight, and service delivery.

Importantly, the solution did not rely on advanced AI for every task. Instead, leaders applied the right tools to the right problems, using automation to handle repetitive workload spikes and digital outreach to accelerate participant response, while leaving nuanced decisions and program stewardship with experienced staff. This kind of targeted design avoids unnecessary complexity while delivering measurable operational relief during periods of change.

Designing for Accountability and Longevity

Responsible AI use also demands strong governance. Audit logs, role-based access controls, quality assurance processes, and model monitoring are foundational, not optional, for human services programs.

Document formats change. Policies evolve. Models drift. Agencies that plan for continuous feedback, retraining, and human oversight are better positioned to sustain performance and maintain trust over the long term.

A People-First Path Forward

Human services programs exist to serve people, and technology should reinforce that mission. When AI is implemented thoughtfully—supporting staff judgment, improving consistency, and reducing administrative burden—it becomes a force multiplier rather than a disruptor.

The future of AI in human services isn’t about automation everywhere. It’s about applying smart technology where it adds value, preserving human reasoning where it matters most, and building program systems that earn trust over time.

---

---

This post was contributed by an APHSA Strategic Industry Partner. The perspectives and opinions expressed are those of the author(s) and do not necessarily represent the views of APHSA.