Architecture Beyond Code
Knowledge, Policy, and Intent as First-Class Architectural Concerns
The first article in this trilogy argued that the software lifecycle is shifting from code-centric production toward the continuous transformation of intent into verified behavior. If that argument is correct, then software architecture must also be reinterpreted. Classical architecture focused on code-bearing concerns: services, components, interfaces, deployment topologies, and operational infrastructure. Those concerns remain essential, but they are no longer sufficient. AI-native systems rely increasingly on three additional concerns that cannot be treated as peripheral overlays: knowledge, policy, and intent. Knowledge supplies the formalized meaning that generation depends on. Policy constrains what generated behavior may do. Intent expresses what the system is meant to achieve and provides the basis for reconciliation when runtime behavior drifts. This essay traces the maturation of each of these concerns, proposes a Knowledge–Policy–Intent (KPI) reference architecture, and argues that architects must now design the informational and governance substrate through which behavior is generated, bounded, and continuously verified.
May 2025Software in the Age of Generative Systems