Transposition of Agentic AI Architectures to Mining IROCs

Establish whether the ReAct-, Reflexion-, and orchestrator/sub-agent agentic-AI patterns can be transposed effectively to integrated remote operation centres in mining.

Background

Integrated Remote Operation Centres (IROCs) are presented as a central element of digital transformation in mining because they consolidate monitoring, coordination, and decision-making across geographically dispersed assets. The paper argues that mining IROCs face heterogeneous data sources, real-time constraints, safety-critical decisions, and substantial operator cognitive load.

The paper reviews classical multi-agent systems, LLM-orchestrated agents, ReAct and Reflexion patterns, and orchestrator/sub-agent architectures that have demonstrated utility in manufacturing and offshore oil-and-gas operations. However, it identifies the transfer of these agentic patterns to mining IROCs as unresolved, particularly given the sector’s requirements for verifiability, auditability, safety, and operational integration.

References

The transposition of these patterns to mining IROCs remains, however, an open research question.

— Autonomous intelligence for mining productivity: a multi-agent AI approach to operational value creation  (2610.26369 - Guerin et al., 19 Oct 2026) in Section 1.3, “The promise of agentic AI for IROCs”