AI-Assisted Decision Support in Housing Policy Implementation: Distinguishing Deterministic Rules, Manual Review, and Heuristic Suggestion

U.S. housing policy directly shapes the housing options available to homeowners and families, but easing zoning rules and broadening legal permission do not, on their own, translate into completed homes. Implementation difficulties surface at two distinct stages: many eligible homeowners never formally apply once they encounter the cost, complexity, and uncertainty of pre-application analysis, and even among those who do apply, a substantial share of permitted projects never reach completion. The principal site of friction is the implementation phase, not the legalization phase-and addressing it requires mechanisms that translate policy rules into parcel-level feasibility judgments that property owners, designers, and local governments can rely on. This article examines how artificial intelligence and related computational methods should be applied to residential buildability assessment. The central question is where AI should sit-and where it should not-within decisions that involve regulatory compliance. The article proposes a three-layer architecture distinguishing rule sets, deterministic calculation, and heuristic suggestion. Matters involving regulatory boundaries-setbacks, lot coverage, building separation-should be handled through deterministic rule sets and calculation; AI may legitimately assist in building and maintaining such systems, but should not perform runtime compliance adjudication. Site conditions that cannot be reliably confirmed from public data should enter manual review. Heuristic methods are appropriate for ranking, placement suggestion, and scenario comparison within the feasible space defined by deterministic rules. Accessory dwelling units (ADUs) serve as the principal scenario for discussion and validation, but the architecture is not specific to ADUs-it applies more broadly to small-scale residential development, adaptive reuse, accessory-structure approval, and land-use change review. The article is presented as a normative contribution to the literatures on AI in public-rule systems, housing policy implementation, and digital governance of the built environment.