McKinsey on harnessing AI to make mining safer

A recent report from McKinsey & Co’s Metals & Mining Practice discusses harnessing AI to make mining safer. The McKinsey authors – Henrique Fagundes, Vishal Agarwal, and Yanto Yanto – argue that AI could improve both safety and productivity by anticipating problems before they arise, detecting worker fatigue and automating dangerous tasks.

The insight report was shared by QuantumBlack, AI – which is the AI-focused arm of McKinsey that focuses on how organisations can most effectively and responsibly use AI to create business value.

The authors argue that mining remains one of the most hazardous industries, with fatality rates that exceed those in transportation, construction, and manufacturing. “Meaningful safety transformation remains uneven across the mining sector. Several structural barriers continue to impede progress.”

The report cites these as follows – first organisational fragmentation and unclear accountability persist. Recurring gaps exist between corporate safety frameworks and site-level execution. Operators may be anchored in a compliance-driven and lagging-indicator mindset. “Even in countries with extensive regulations, operators may be focused on preparing for audits and managing documentation instead of proactively identifying and mitigating hazards.”

Data silos and legacy infrastructure may undermine the impact of digital investments. “In some cases, data from advanced sensors installed to monitor slope stability and equipment health is not integrated into company-wide safety dashboards that could allow safety experts to see compounded risk factors.”

The global push to scale production of critical minerals has created capability and contractor management gaps, particularly in emerging markets. “Rapid workforce growth and increased reliance on contractors in some regions have introduced variability in safety standards and supervision, including inconsistent training, inadequate management of worker fatigue, and limited frontline oversight.

Finally, the culture and change management required to embed new technologies is substantial. “Pilot programs in autonomous haulage, fatigue monitoring, and AI-enabled hazard detection have shown safety benefits, but mining digital transformations typically require sustained leadership alignment, frontline adoption, workforce upskilling, and change management to scale beyond pilots and deliver durable impact.”

McKinsey argues that these obstacles persist against a backdrop of technological changes that have made safety inroads. “Beginning in about 2010, automation (including autonomous trucks and other equipment) and remote monitoring began to spread throughout the industry. Companies accelerated their use of these technologies over time, with increased deployment of autonomous haulage and equipment incorporating Internet of Things (IoT) technology in recent years, with human oversight in or on the loop.”

It says these technological solutions may offer additional safety improvements across four
complementary levers:

  • To control hazards, autonomous drones map and inspect underground shafts, tailings dams,
    and blasting zones in real time, allowing humans to monitor from a safe distance. IoT
    enabled sensors embedded in vehicles, ventilation systems, conveyor belts, crushers, and
    other equipment continuously monitor air composition, ground stability, and equipment
    health. Connected through edge computing and 5G networks, these systems provide instant
    analytics and predictive maintenance information to trained operators, allowing them to
    make decisions that systematically reduce the likelihood of catastrophic failure.
  • To improve health, safety, and environmental conditions, trained operators in centralised
    control rooms aggregate live operational data into a continuous safety dashboard that can
    strengthen oversight, standardize governance, and accelerate responses to irregularities.
    Meanwhile, autonomous trucks and remote-monitoring centers can reduce human exposure
    to risky situations, and sensor-driven predictive analytics can reduce maintenance-related
    incidents.
  • To change behaviours, smart watches or dash cameras, used with appropriate employee
    consent and privacy guardrails in place, can allow real-time monitoring of worker fatigue,
    PPE (personal protective equipment) compliance, and proximity interactions that provide
    immediate feedback to supervisors and operators, reinforcing safe operating discipline
    before unsafe acts escalate.
  • To educate and train, digital twins replicate entire mines, allowing operators to simulate
    stress scenarios, maintenance schedules, and safety drills in a virtual environment before
    implementing changes on-site.

Safety transformations could be redefined by AI

The authors argue that Generative AI and agentic AI systems enhance and may even redefine these levers. “Gen AI copilots in control centres summarise safety data, identify root causes of near misses, and draft incident reports in seconds. Although human intervention and monitoring is always required,
agentic AI systems that can act autonomously within predefined safety parameters are beginning to orchestrate preventive responses, from adjusting ventilation to rerouting autonomous vehicles around unstable ground.”

Miners are experimenting with large language model (LLM) – based hazard and incident analytics
that can analyse thousands of hazard reports, near-miss narratives, and investigation summaries
to identify recurring precursors (such as disabling or overriding alarms or contractor interface
breakdowns) and propose targeted monitoring questions for future supervisors that could avoid
incidents.

These approaches McKinsey says have the potential to remove the structural barriers described above, as illustrated by the following examples:

  • AI for hazard identification and cross-company coordination. Gen-AI-style foundation
    models are beginning to reshape frontline safety management, in conjunction with human
    oversight. Rather than deploying isolated AI tools, one company implemented a domain
    trained large AI model designed to support multiple safety and operational applications
    across its coal operations. The model integrates video feeds, sensor data, and text-based
    reports to detect hazardous behaviors (such as personnel entering restricted zones), identify
    abnormal conditions, and enhance centralised remote oversight of underground activities – all with high accuracy – delivered to trained supervisors.
  • AI for behavioral change and proactive safety management. An operator of bulk transshipment services equipped its vessels with an AI-enabled video analytics platform that continuously monitors existing CCTV feeds to identify deviations from standard operating conditions, such as crew entering restricted zones, slips or falls on deck, or smoke appearing in the engine room. These alerts are routed to trained safety analysts who verify what the analytics has flagged, check it against other operational data, and rapidly inform the relevant captain with precise instructions, if needed, to act before a minor hazard escalates. Once an event is closed, an AI agent generates a structured incident report that includes root causes of the incident and actionable lessons learned. Finally, a safety specialist evaluates whether the recommendations should be considered for implementation across the fleet.
  • AI for better contractor safety management. Mining companies are beginning to use AI and digital tools to strengthen contractor safety management, particularly in high-risk activities where contractor execution, site controls, and real-time monitoring are critical. One global mining company’s AI-backed contractor management framework better integrates contractors and third parties into its systems, processes, and performance management routines

The report concludes: “Intelligent safety is becoming a foundation of competitive advantage in mining. Safety technology and operational excellence are converging into a single, data-driven ecosystem.
Combining automation with AI-powered safety platforms and appropriate human oversight can contribute to improved performance and worker safety, reducing unplanned downtime, improving asset utilisation, and protecting workers. The result is mines that are not only safer, but also more productive, efficient, and sustainable.”

 

The post McKinsey on harnessing AI to make mining safer appeared first on International Mining.