CS-12
JSE Mining Group — AI Safety Intelligence in Deep-Level Mining
A JSE-listed South African mining group operating 8 deep-level gold and platinum mines across Gauteng, North West, and Limpopo faced an existential safety challenge: 14 fatalities and 890 lost-time injuries annually were exposing the group to Section 54 work stoppages that cost $28M per year in production loss, while regulatory scrutiny from the Department of Mineral Resources threatened operating licences. Anicalls' AI Safety Intelligence Agent transformed the group's safety culture and operational safety management — reducing incidents by 71% and delivering $44M in combined annual saving.
71%Safety Incident Reduction
$44MAnnual Saving
ZeroFatalities in Year 2
94%High-Risk Event Prediction
Business Challenge
The Safety and Business Crisis in Deep-Level South African Mining
14 Fatalities & 890 LTIs Annually
The group's 28,000 mineworkers operating at depths of 2,500–3,800 metres faced extraordinary hazard exposure — seismicity, heat stress, methane, fall-of-ground, and machinery interactions. The 14 annual fatalities were not only a human tragedy but triggered mandatory Section 54 production stoppages (averaging 3.5 days per fatality) and individual Section 55 incident investigations that collectively cost $14M in direct production loss annually.
Section 54 Stoppages: $28M Annual Impact
The Department of Mineral Resources' Section 54 stoppage powers were exercised 47 times across the group's 8 mines in the preceding year — costing $28M in production losses. The reactive nature of safety management meant stoppages came without warning: a fatality or serious near-miss on a Monday morning could shut a R800M/month operation for 5 days. The unpredictability of Section 54 stoppages was the group's largest single operational risk.
Heat Stress at Extreme Depth
At depths below 3,000 metres, virgin rock temperature exceeds 60°C — requiring extensive refrigeration systems to maintain working temperatures below 28°C wet bulb (the MHSA regulatory limit). Refrigeration system failures, personnel overcrowding in stopes, and individual physiological variation created undetected heat stress conditions. 28% of lost-time injuries were heat-stress related — entirely preventable with real-time physiological monitoring and environmental sensing.
Reactive Safety Management Culture
Safety management was entirely reactive — incident reports, investigation committees, and corrective action registers that captured failures after they occurred. Leading indicator monitoring was manual and inconsistent: supervisors self-reported on safety observations, creating reporting bias (near-misses were under-reported to avoid adverse performance assessments). The safety data estate was fragmented across 8 mines with no cross-mine pattern analysis capability.
Solution Delivered
Anicalls AI Safety Intelligence Agent — 8-Mine Deployment
Sensing
Real-Time Underground Sensing Network
Anicalls deployed an integrated underground sensing network across all 8 mines — covering seismic monitoring (1,200 geophones), environmental sensors (temperature, humidity, gas levels, dust), and personnel tracking (UWB positioning for all 28,000 workers and 4,200 pieces of trackable equipment). Real-time data streams from 48,000+ sensors flow to the AI safety platform continuously, providing complete environmental and personnel situational awareness underground.
- 1,200 seismic geophones across 8 mines
- UWB positioning for 28,000 workers + 4,200 equipment
- Real-time gas, temperature, humidity, dust monitoring
- 48,000+ sensor data streams processed continuously
Prediction
AI Hazard Prediction Engine
Machine learning models predict high-risk events — seismic bursts, fall-of-ground risk, heat stress conditions, and gas accumulation — with 94% accuracy and sufficient lead time for preventive action. Seismic prediction models identify areas of elevated rockburst risk 4–8 hours before events, enabling personnel evacuation of at-risk panels. Heat stress prediction identifies individual workers approaching physiological limits before symptoms manifest, triggering rest rotation.
- Seismic burst risk prediction (4–8 hour lead)
- Individual heat stress physiological prediction
- Fall-of-ground risk zone identification
- Gas accumulation trend detection and alert
Response
AI Emergency Response Coordination
AI-powered emergency response coordinates evacuation routing, rescue team dispatch, surface support mobilisation, and family notification automatically within seconds of a detected emergency event — versus the previous 20–45 minute manual escalation cycle. Integration with the mine's refuge chambers enables real-time tracking of workers sheltering during seismic events, providing surface control with accurate casualty location data for rescue team briefing.
- AI evacuation route optimisation (per-person)
- Rescue team auto-dispatch with location briefing
- Refuge chamber occupancy tracking
- Family notification within 3 minutes of event
AI Workforce Deployment
The Anicalls AI Safety Operations Team
AI Environmental Monitoring Agents
24 AI Environmental Monitoring Agents — 3 per mine — process continuous sensor data streams from all 48,000+ underground sensors, detecting anomaly signatures, predicting hazard escalation, and generating real-time alerts for the safety control room and underground supervisors. Each agent monitors its assigned mine section with sub-10-second response times — compared to the previous 30-minute manual round-based inspection cycle.
AI Safety Compliance Agents
16 AI Safety Compliance Agents automate the MHSA regulatory compliance reporting burden — generating Section 11 statutory safety reports, tracking corrective action closure rates, maintaining the legal appointment register, and producing DMRE inspection-ready compliance packages. The 60% reduction in safety administrative burden freed safety officers to spend time underground — where their expertise creates the most value — rather than at office desks completing paperwork.
AI Behavioural Safety Analysts
Dedicated AI Behavioural Safety Analyst agents process supervisor safety observation records, near-miss reports, and leading indicator data to identify behavioural safety trends — detecting production pressure-safety trade-offs, supervisor observation frequency patterns, and work area safety culture scores. Cross-mine pattern analysis identifies systemic issues that manifest at multiple operations, enabling targeted safety culture interventions before they result in incidents.
Technologies Used
The AI Technology Stack Deployed
Platform
MineGuard AI Platform™
Purpose-built underground mining safety AI platform — hardened for deep-level mining operating conditions including RF-challenged environments, high EMI, extreme temperature and humidity, and intrinsic safety requirements for gas-hazardous environments. Integrated with standard mine communication systems (Leaky Feeder RF, fibre optic backbone) and compatible with all major underground personnel tracking systems (Booyco, Siemens, Johnson Controls PLC systems).
- Intrinsically safe sensor hardware (ATEX certified)
- Leaky Feeder RF + fibre optic integration
- Booyco / Siemens personnel tracking integration
- MHSA Section 11 compliance automation
Seismic AI
Seismic Hazard Prediction Models
Recurrent neural networks trained on 12 years of seismic data from the group's 8 mines — identifying precursor micro-seismic patterns that precede major seismic events. The models incorporate geological structure data, mining-induced stress field modelling, and historical seismic catalogue analysis to produce spatially resolved seismic hazard maps updated hourly. Prediction accuracy of 94% for magnitude 2.0+ events with 4-hour lead time.
- 12-year seismic training data per mine
- Spatially resolved hourly hazard maps
- 94% accuracy for M2.0+ events (4-hour lead)
- Geological structure and stress field integration
Physiology AI
Wearable Physiological Monitoring
Smart safety helmets with integrated heart rate, core temperature, and hydration sensors — deployed to all 28,000 workers — stream physiological data to the AI platform for individual heat stress prediction. The AI predicts each worker's time-to-limit state (when core temperature will reach MHSA threshold of 38.5°C) and triggers proactive rest rotation 30 minutes before predicted limit — preventing heat stress incidents before symptoms manifest.
- Smart helmet biometric sensors (28,000 units)
- Individual heat stress time-to-limit prediction
- 30-minute early rest rotation trigger
- MHSA 38.5°C threshold compliance automation
Quantified ROI
The Financial and Human Impact at 24 Months
$44M Annual Combined Saving
The $44M annual saving comprised: $20M in Section 54 stoppage elimination (from 47 stoppages to 4 in Year 2 — a 91% reduction); $12M in insurance premium reduction (following 3-year incident trend improvement); $8M in reduced workers' compensation costs (71% fewer LTIs); and $4M in legal and regulatory defence costs eliminated. The total financial case (3-year NPV at 12% discount): R1.8 billion ($96M) versus a R420M ($22M) 3-year platform investment.
Zero Fatalities in Year 2
The most significant outcome: zero mineworker fatalities in Year 2 — the first fatality-free year in the group's 42-year operating history. From 14 fatalities in the baseline year, Year 1 delivered 5 fatalities (64% reduction) and Year 2 delivered zero. The human value of this outcome is immeasurable — but the associated regulatory relief, community trust restoration, and workforce morale improvement contributed meaningfully to recruitment, retention, and productivity.
Productivity Improvement: +8.2% Ore Yield
The elimination of Section 54 stoppages and the reduction in LTI-related absenteeism increased effective underground working hours by 12.4% annually. Combined with reduced production disruption from seismic events (AI-predicted evacuation versus uncontrolled mass evacuation), ore milled per month increased 8.2% with no additional capital investment. The productivity improvement generated an additional R620M ($33M) in annual gold and platinum revenue.
Business Outcomes
Safety, Regulatory, and Social Outcomes
Regulatory
DMRE Principal Inspector Commendation
The Department of Mineral Resources and Energy Principal Inspector for Mines issued a formal commendation for the group's safety performance improvement — citing it as "the most significant safety transformation achieved by any South African mining group in a single 24-month period." Three mines previously on regulatory watch lists were removed from enhanced monitoring. The group's social licence to operate — previously at risk — was restored across all 8 operations.
Community
Mining Community Trust Restored
The communities surrounding the group's mines — who had previously staged protests over mining fatalities and demanded mine closures — became advocates for the group's operations following the zero-fatality Year 2. The group's social and labour plan commitments (housing, education, healthcare) were rated "fully compliant" for the first time, enabling licence renewals without condition. Community trust surveys showed net positive sentiment for the first time in 8 years.
Talent
Workforce Recruitment & Retention
Deep-level mining recruitment had become difficult as safety reputation deteriorated — with skilled miners increasingly preferring surface operations or open-cast mines. Following the safety transformation, the group's voluntary turnover rate fell from 18% to 7%, and time-to-fill for critical mining roles improved from 68 days to 24 days. The group was named South Africa's top employer in the mining sector by MEIBC for the first time in 2025 — directly attributed to the AI safety programme.
Executive Testimonial
"When we achieved our first fatality-free year in 42 years of mining, our board had to pause for a moment to truly understand what that meant — for the families of our workers, for the communities around our mines, and for the future of deep-level mining in South Africa. The AI gave us the predictive capability we had always needed but never had. The $44M saving matters to our shareholders. But zero fatalities is what matters to us as a company."
Metrics Dashboard
24-Month Performance Scorecard
0Fatalities Year 2 (was 14)
-71%Lost-Time Injuries
4Sec. 54 Stoppages Year 2 (was 47)
$44MAnnual Saving
94%Seismic Event Prediction Accuracy
+8.2%Ore Yield Improvement
7%Staff Turnover (was 18%)
4.4×3-Year Platform ROI
Protect Your Workforce. Protect Your Licence.
See how the Anicalls MineGuard AI Platform can predict and prevent mining hazards before they result in injury — delivering zero-harm performance and eliminating the Section 54 stoppage risk that threatens your operation.