Mining and resources
Start with a bounded investigation or evidence task. Mining examples in this guide build on safety, production, and audit work described by the author. The complete agent extensions remain proposals.
Opportunities
- Safety coaching from CCTV, incident and observation data.
- Investigate production discrepancies.
- Collect and check compliance evidence.
- Investigate recurring audit issues across sites.
Each opportunity uses one proposed agent. The autonomy and learning levels and the zones are this guide's recommendation, not observed deployments.
- Agent
- Safety coaching agent
- What it does
- Checks a detected CCTV event against the safety procedure. Drafts coaching for a safety owner to approve
- Autonomy and learning
- Delegated tasks (A2), curated learning (L1)
- Zone or rule it touches
- The rule for decisions about people. The agent never decides discipline or any other matter about a person
- Evidence status
- Proposal. Builds on the author's earlier CCTV risk work. The agent extension has not been deployed
- Agent
- Production discrepancy agent
- What it does
- Checks measurements, sensor health and operating records when production estimates disagree. Prepares a reconciliation case
- Autonomy and learning
- Delegated tasks (A2), curated learning (L1)
- Zone or rule it touches
- Formal regulatory records. A person accepts any change to the production record
- Evidence status
- Proposal. Builds on the author's computer-vision measurement work. The agent extension has not been deployed
- Agent
- Compliance evidence agent
- What it does
- Asks the owning team for a specific record. Checks issuer, date and scope. Prepares a traceable package for the reviewer
- Autonomy and learning
- Delegated tasks (A2), curated learning (L1)
- Zone or rule it touches
- Formal regulatory records. A named person attests
- Evidence status
- Proposal. Builds on the author's earlier standards-to-checklist work
- Agent
- Audit pattern agent
- What it does
- Tests competing explanations for findings that recur across sites. Requests the records that would confirm or rule out each one
- Autonomy and learning
- Delegated tasks (A2), curated learning (L1)
- Zone or rule it touches
- Formal regulatory records, read only. An engineering or assurance owner decides the response
- Evidence status
- Proposal. Builds on the author's topic-modelling work on audit gaps
Delegated tasks (A2) means a person reviews every output. Curated learning (L1) means people update what the agent knows, offline, and review each change before release. Both scales are on the autonomy contract. The zones are set out on the deterministic zones page. None of these agents touches safety actuation. Equipment, isolation and stop decisions stay in the site's approved control systems.
The estate
A miner with several mine sites and a processing plant runs systems in all fourteen layers. The table names the main ones. Vendor names in brackets are examples, not endorsements. Most come from industry knowledge and are typical, not sourced.
- Main systems at a miner with a processing plant
- Site server rooms. Rugged computers on trucks and drills. Private LTE or wireless mesh for autonomous haulage (Nokia, Rajant). Remote operations centre in the city. Cloud
- Layer
- Data platform (R02)
- Main systems at a miner with a processing plant
- Historian (AVEVA PI). Fleet management database. Geology database (acQuire, Maptek). Block models. Lab system (CCLAS, LabWare). Lakehouse (Palantir Foundry at Rio Tinto). Reports and dashboards
- Main systems at a miner with a processing plant
- A buffer network between plant and business networks, with historian copies. Fleet system adapters. Equipment maker telematics (Caterpillar, Komatsu). Rail and port scheduling links. Integration platform
- Main systems at a miner with a processing plant
- ERP. Enterprise asset management (SAP PM, Ellipse, Maximo). Metal accounting (ION Algosys, Caspeo). Production database. HSE and critical control (Cority, Enablon, Intelex, Forwood). Isolation and permit system. Fly-in fly-out rostering. Procurement. Tenement register (Trimble Landfolio). Documents. GIS
- Main systems at a miner with a processing plant
- Fleet management (Modular Mining DISPATCH, Cat MineStar, Wenco, Hexagon). Autonomous haulage (Komatsu FrontRunner, Cat Command). Drill systems. Collision avoidance (Hexagon). Fatigue monitoring (Seeing Machines). Slope monitoring (GroundProbe). Plant control (ABB 800xA, Rockwell PlantPAx, Citect). Expert control on mills. Belt scales. On-stream analysers. Mine planning (Deswik, Maptek, Datamine). Geology modelling (Leapfrog)
- Main systems at a miner with a processing plant
- Expert control on mills and flotation. Fleet dispatch optimisation. Predictive maintenance on haul fleets (BHP). Ore vision. Drilling optimisation (Hexagon Drill Assist). Exploration models
- Layer
- Agent platform (R07)
- Main systems at a miner with a processing plant
- Mostly new. Robotic process automation in finance and payroll. Palantir AIP (Rio Tinto). Cat AI Assistant. Agents sold inside apps (Groundhog)
- Main systems at a miner with a processing plant
- Email, Teams, SharePoint and Excel. Pre-start and toolbox talk material. Tablets for short interval control. Radio
- Main systems at a miner with a processing plant
- Dispatch consoles. In-cab displays. Remote operations centre video walls. Inspection apps. Crib-room screens
- Main systems at a miner with a processing plant
- Corporate directory. A separate OT directory. Privileged access. OT network monitoring. Security event monitoring. Contractor site access. Fleet network security
- Main systems at a miner with a processing plant
- Risk register. Critical control management. Tailings governance (GISTM). Statutory mine safety system. Royalties. Reserve reporting (JORC Code). Emissions reporting (NGER). Land access
- Main systems at a miner with a processing plant
- IT monitoring. Historian interface health. OT asset inventory. Service management. Fleet network health. Equipment maker licence management
- Main systems at a miner with a processing plant
- Dispatchers. Mine technical services: geology, survey and planning. Metallurgists. Maintenance. Remote operations centre (BHP, Rio Tinto, Fortescue in Perth). HSE. Statutory roles such as the site senior executive or registered manager
- Main systems at a miner with a processing plant
- Mostly new. Pipelines from historian to lakehouse. Geology and survey file exchange. Document capture for permits and assays. Equipment master data
Agents run on the business side or in the cloud. A read broker in the buffer network serves them plant data from the historian copy. Nothing an agent does writes to plant control, collision avoidance, fatigue monitoring or the permit and isolation system. The energy and resources page draws this estate as a building, with the boundary to the plant as a wall. It also gives the generator and oil and gas estate beside the miner's.
Process maps
Each worked example maps the work on this estate, step by step. It shows the process today across these systems, then the same work with an agent. It names who approves, where, and what the agent reads and writes.
- Investigate production discrepancies: truck tonnes from the pit against belt-scale tonnes at the plant. Fleet system, historian, lab system, EAM calibration records and metal accounting.
- Safety coaching from CCTV, incident and observation data: a detected event becomes coaching the safety owner approves. Video system, HSE system, permits, roster and learning system.
- Investigate recurring audit issues across sites: the same finding at several sites. Audit system, EAM history, inspection records and procedures.
- Investigate an asset before planning work: a mill bearing or conveyor drive runs outside its normal range. Historian, asset performance management, EAM and engineering tools.
Enterprise capabilities
Use verified asset identity, dated observations, governed access to records, and reliable work-order interfaces. Preserve measurement uncertainty and field evidence. An agent can gather and explain evidence; the responsible operating process controls physical actions and formal assurance.
Start and measure
Choose one site, zone, or process with a clear owner and representative historical cases. Compare with existing alarms, rules, and expert review. Measure confirmed outcomes, false conclusions, review time, and total cost. Test interrupted work and conflicting information before expanding.
People and controls
Assign an operational owner, an engineering reviewer, data owners, and a support route. Do not count generated work items as completed field work. Use the same access boundaries during evaluation as during production.
Score each agent on the risk profile and register it with a named sponsor through the agent lifecycle.
Explore the design
Evidence and limits
These are proposed applications. Their value and safety require evaluation on local data and within the organization's operating procedures. The linked cases distinguish practitioner experience from external evidence and proposed extensions.
A research round on 15 October 2026 searched for mining agent deployments. None was found at BHP, Rio Tinto or Fortescue. They show machine learning in production, platform deals and job advertisements that mention agents. No agent was found from Deswik, Komatsu, Modular Mining, Maptek, Datamine or Wenco for these processes. Absence in twenty searches is not proof of absence. Each agent above is a proposal built on work the author described. None is evidence that the full agent works.
The estate table is typical practice, not a survey of one company. The Global Mining Guidelines Group (GMG) has published guidelines for the industry. Three came out in 2019: autonomous systems, short interval control and interoperability. A system safety guideline followed in December 2023. No GMG guideline on AI agents was found.
Sources:
- BHP Insights, AI in mining, August 2024.
- Business Wire, Rio Tinto renews Palantir, 12 November 2024 [vendor].
- Global Mining Review, GMG guideline on autonomous systems in mining, 24 April 2019.
- ABMEC copy of the GMG guideline for applying system safety in mining, December 2023.
- AusIMM, how AI is changing metal accounting, 23 April 2026.
- Proactivo, GMG gathering in Lima on why mining AI pilots stall, September 2026.
Cite this page
Written by Murali Sid, Founder and AI Systems Architect, last updated 15 October 2026. Licensed CC BY-SA 4.0, so quote it freely with attribution.
Murali Sid (2026). Mining and resources. Agentic Architecture Skills. https://www.agenticarchitectureskills.com/library/blueprints/verticals/mining-and-resourcesMachine-readable version of this page: /library/blueprints/verticals/mining-and-resources.md
Healthcare
A proposed design for a public hospital network in Australia or New Zealand. Four agents carry referrals, follow-up, meeting preparation and accreditation evidence around the patient. Clinicians keep every clinical de...
Manufacturing and supply chain
A proposed workflow for manufacturing and supply chain. Revised 17 September 2026. The roles below are design options, not claims that these agents are deployed or that multiple agents are necessary.