Governance-Integrated Agentic Automation for Enterprise Innovation Engineering: A Productivity-Governance Control Cycle and Risk-Adjusted Evaluation Model

Agentic artificial intelligence (AI) is shifting enterprise automation from passive content generation to software entities that can plan, use tools, access data and execute multi-step work. The engineering problem is no longer only whether agents can increase productivity, but how organisations can scale them without creating opaque accountability chains, uncontrolled permissions, unreliable outputs or workforce displacement without adaptation. This paper develops an original design-science framework for governance-integrated agentic automation in enterprise innovation engineering. Drawing on AI risk management, agentic AI governance, human-centred automation, algorithmic auditing and the author’s prior work on AI stakeholders, dynamic ethical equilibrium and the digital productivity paradox, the study proposes the Productivity-Governance Control Cycle (PGCC) and a Productivity-Governance Coherence Index (PGCI). The framework links use-case selection, autonomy tiering, human approval gates, AgentOps telemetry, audit evidence and workforce redesign into a single engineering cycle. An illustrative four-use-case evaluation shows that apparently high-value automation can fall below the safe-scaling frontier when residual risk and weak human-control adequacy are included. The contribution is a practical, standards-aligned architecture for converting agentic AI from experimental productivity tools into governable enterprise systems. The paper argues that sustainable AI value creation depends on engineering governance into the workflow itself rather than adding compliance after deployment.