Decision Intelligence for AI-Driven Agile Lifecycle Governance: Linking Architecture-Centered Management to Defect Risk Forecasting

Abstract

Agile delivery improves responsiveness but can amplify architectural drift, defect leakage, and governance ambiguity when quality controls are not explicitly linked to measurable risk signals. In practice, many teams implement defect prediction or operational analytics as standalone dashboards rather than as decision drivers for sprint planning, release gating, and architecture investment. This paper presents DI-AIGLG, a decision intelligence-based governance framework that links architecture-centered management with defect risk forecasting and observability evidence to produce repeatable, auditable, and low-overhead governance actions across the agile lifecycle. DI-AIGLG formalizes governance as a closed-loop system consisting of evidence capture, predictive risk modeling, decision policies, and outcome learning. We define an architecture exposure score, propose a calibrated risk score that combines predicted defect likelihood with architectural blast radius, and map risk to actionable controls spanning backlog selection, CI/CD gating, and release readiness. A worked example demonstrates risk budgeted sprint planning under capacity constraints. The framework is designed to be implementable using CI/CD telemetry, automated testing signals, and AI-enhanced observability pipelines.