Mining companies are moving from isolated analytics toward tightly coupled cyber-physical systems in which industrial artificial intelligence (AI), operational digital twins, edge computing, and software agents influence physical production. The transition creates material opportunities in maintenance, dispatch, energy management, ore control, processing, and hazardous-area inspection, but it also introduces new failure modes where erroneous data, models, or autonomous actions may propagate across the mine-to-mill value chain. This paper develops a practical engineering framework for implementing these technologies without transferring accountable engineering authority to AI. The study uses a conceptual and systems-engineering methodology combining structured literature synthesis, functional decomposition of a mining enterprise, risk-based governance, and a theoretical integrated decision scenario. The proposed architecture contains eight layers: data sources, a unified industrial data layer, governed engineering models, an operational digital twin, a bounded agent layer, an independent verification layer, human-in-the-loop authorization, and the execution environment. A four-level K1–K4 criticality scheme links technical consequence to mandatory controls, escalation, and rollback. The framework also defines agent rights, stage-gates, value-realization criteria, and a 2026–2030 implementation roadmap. The principal contribution is an implementation-oriented bridge between digital-twin research, industrial AI, autonomous mining, and corporate governance. The paper does not claim that artificial general intelligence exists or that the framework has been industrially validated; it presents a testable architecture for future pilots and controlled deployment.
Durmishkhan Gakharia (Fri,) studied this question.