Policy & Practice | Fall 2026

ai in focus

By Heidi Reed and Lauren Aaronson

A New Path Forward for Eligibility Modernization: How AI-Assisted Development, Applied with Deep Expertise, Can Make Modular Modernization a Reality

E very day, state eligibility systems decide who gets health coverage, food assistance, cash assistance, child care benefits, and more. They process millions of cases across dozens of programs, applying layers of federal and state rules at a scale that few other government systems match. Keeping them current has always been difficult, and the pressure to modernize has never been greater. States have long faced a difficult choice: keep patching aging systems that can barely keep up with policy change or pursue full replacements that carry enormous risk and cost. A third option is gaining traction. AI-assisted development tools, used by teams with the right expertise, are making modular modernization both practical and affordable. Complexity does not go away, but skilled teams now have better instruments to work through it. A typical state manages hundreds of eligibility rules across Medicaid, the Supplemental Food Assistance Program (SNAP), Temporary Assistance for Needy Families (TANF), and other programs, each governed by its own federal framework with distinct income definitions and verifi cation requirements. Those rules shift by program, state, county, and some times, household type. The complexity is an accurate reflection of the under lying policy, not a design flaw that modernization will solve. The Real Complexity States Are Managing

Teams also face a constant stream of policy changes (federal rule updates, state mandates, court orders) each requiring system modifications while the current platform continues serving people every day. Carving out space for strategic modernization inside that workload is one of the central chal lenges. The aim is to build systems that manage the complexity better, working through it incrementally rather than pretending it can be engi neered away. Why Modular Modernization, and Why Now The ability to divide a large system into smaller, independently deploy able parts has long been appealing in theory. Executing it is a different matter. You cannot simply separate one part of a legacy system and call it a module. Data move between

components, and business rules are often shared across the system in ways that are not well documented or fully understood. Getting the boundaries right requires people who understand both the technical architecture and the policy it encodes, and that combi nation is genuinely hard to find. Data migration makes this even harder. Moving a capability into a new module often means restructuring data, preserving years of historical records for audits and appeals, and keeping both old and new systems in sync during the transition. Errors in this process have real consequences for real people, so the work demands exceptional care. The approach follows a clear pattern: build new capabilities along side the existing system, route work to new modules gradually, and retire legacy components only after replace ments are proven in production. The

14

Policy & Practice Fall 2026