Why AI-Driven Development Exposes a Critical Gap in Change Governance

Modern platform engineering in 2026 relies on two main control layers: internal developer platforms (IDPs) that catalog services and ownership, and AI agent governance gateways that monitor runtime tool calls. However, neither layer can definitively answer which projects a pending change affects, who approved it, or where the evidence of that approval is stored. IDPs track what exists but not change intent, while agent gateways log individual actions without linking them to broader portfolio-level decisions. The author argues that AI-native delivery — where agents produce most code at machine speed — makes this missing "change-intent governance" layer a critical requirement rather than an optional feature. A minimal implementation would need deterministic impact claims tied to immutable snapshots and append-only approval records that neither the platform nor the requester can alter.