MINING TRANSFORMATION
Not in Theory.
In Actual Mines.
Most transformation advisors have never sat in a mine control room, read a Wenco report, or diagnosed why Pronto and SAP are telling two different stories. We have. BHP Billiton. First Quantum Minerals. Live AI models in active mining environments. That is the foundation we bring to every engagement.
Why It Matters
$10M
Savings delivered in 12 months
FIRST QUANTUM - ZAMBIA
20+
Years in senior advisory across mining environments
3 CONTINENTS
0.90
ROC AUC engine failure prediction accuracy
LIVE IN DATA SENTINELS
70%
Live AI models built for mining, deployed in phases
PREDICTIVE · HAUL · GRADE
BW
Locally registered in Botswana. HQ Austin, TX. Partners across Zambia, South Africa and Zimbabwe.
$10M
Savings. First Quantum Minerals. 12 months.
BHP
Billiton Software Implementation Lead. London.
5
Live AI models. Predictive maintenance to grade prediction.
4
Countries with registered presence or active partners.
THE REAL PROBLEM
Why $4 Billion in Savings Is Harder Than It Looks.
You have had the reports. You have seen the roadmaps. The recommendations sit on drives, referenced in meetings, implemented in pieces. Three things keep getting in the way.
THE REAL PROBLEM
Why $4 Billion in Savings Is Harder Than It Looks.
You have had the reports. You have seen the roadmaps. The recommendations sit on drives, referenced in meetings, implemented in pieces. Three things keep getting in the way.
DIAMOND
Botswana · Southern Africa
Primary KPI
Grade maintained · ROM integrity · Recovery rate
Typical Data Problem
Multiple ore bodies with high grade variation. Decisions on blend and recovery made without a trusted single source of truth across geological models and plant feed data.
Our Approach
Grade-aware predictive models. Custody data integration. Dilution dashboards. ROM grade prediction before ore reaches the plant.
COPPER
Zambia · DRC · Chile
Primary KPI
Tonnes processed · Fleet cycle time · Energy cost per tonne
Typical Data Problem
Fragmented OT systems with poor fleet utilisation visibility. High volume operations where 1% cycle time improvement is worth millions but the data to find it does not exist in a usable form.
Our Approach
Fleet utilisation analytics. Haul road intelligence. Predictive maintenance across high-volume fleets. OT and ERP integration to build one source of truth.
GOLD
South Africa · Ghana · West Africa
Primary KPI
Grade control · Reagent consumption · Leach circuit recovery
Typical Data Problem
The gap between the geological model and what actually comes out of the plant. Reagent decisions made on schedule rather than real-time circuit data.
Our Approach
Grade control data integration. Leach circuit optimisation models. Reagent consumption analytics tied to actual recovery outcomes rather than fixed schedules.
IRON ORE
Southern Africa · West Africa
Primary KPI
Tonnes · Fe grade at ship · Rail and port throughput
Typical Data Problem
Operations optimise the pit but ignore the bottleneck downstream. Rail, port, and blending data live in separate systems with no connected view from pit to ship.
Our Approach
Full value chain integration from pit to ship. Blending optimisation for Fe grade targets. Rail and port logistics connected to mine production data.
AI THAT ACTUALLY WORKS
Five Models. Deployed in Phases.
Each Tied to a Savings Mechanism.
40% of AI projects will fail by 2027: not because the technology does not work, but because organisations automated broken processes. We only deploy AI when three conditions are met.
THE RULE
We only deploy AI when:
01
The data is clean
02
The process is documented and measurable
03
The ROI is signed off before build starts
LIVE
PHASE 1
Predictive Maintenance AI
ROC AUC 0.87 to 0.90: engine failure accuracy
Engine failure prediction on Wenco and Readyline sensor data. Three windows: Urgent, Short, Long. Already built. Deploys as soon as data is clean.
Saving: Eliminates unplanned engine failure. One unplanned failure on a major haul truck costs $500K to $2M in lost production and repair.
LIVE
PHASE 2
Oil Lab Classification AI
Good · Fair · Watch · Action: 4-class condition grading
GradientBoosting classifier grades each oil sample per compartment. Mechanical maintenance triggered by data, not schedule or gut feel.
Saving: Proactive maintenance costs 3 to 5 times less than reactive. Across a major mining fleet, this is $10M to $30M annually.
PHASE 3
Grade Prediction Assist
Applicable across diamond, gold and copper operations
Drill hole data, geological models, and plant feed data combined to predict ore grade before it reaches the plant. Grade-aware blend optimisation.
Saving: 1% improvement in grade management equals hundreds of millions in recovered value.
BUILT MODELS
PHASE 2
Haul Road Intelligence
Up to 60% of productivity tied to time management
AI upgrades the Haul Road Explorer into an intelligent system: flags which segments to prioritise based on predicted tonnes lost.
Saving : 1% cycle time improvement across a 50-truck fleet equals millions annually. At major scale, ~$50M annual opportunity.
PHASE 2
Fleet Anomaly Detection
14-day early failure warning horizon
Real-time alerts when a machine deviates from baseline. Not “the engine failed.” — “this engine is trending toward failure in 14 days.”
Saving: Planned maintenance vs emergency shutdowns. Scheduled 8-hour stop vs 3-week production loss.
“We are not an AI company selling AI. We are implementers who use AI when it has earned the right to be used.”
THE TWO-PRONGED APPROACH
One Prong Stabilises. The Other Builds the People Who Sustain It.
Every phase has a fixed deliverables list, signed off before work begins. Any request outside the list triggers a formal change order with pricing. No change order, no work. This protects both parties.
PHASE 1
MONTHS 1–6
Data Foundation Audit and Integration
Full ETL audit across Wenco, SAP, MineStar and Pronto. Data governance framework: one source of truth per KPI. Real-time dashboards: utilisation, OEE, haul road performance. Predictive maintenance models. Budget modelling and scenario planning tools.
Deliverable: Data Health Report + Prioritised Opportunity Heat Map
PHASE 2
MONTHS 7–18
Quick Wins and Systems
Deploy predictive maintenance models. Haul road explorer live. Real-time fleet utilisation dashboards. MTS budgeting and forecasting tool. Component lifecycle managemen
Target: $500M to $1B in cost identification in first 12 active months
PHASE 3
MONTHS 19–30
Workforce Transformation and Capability Building
ICT and Developer Framework Training. Management Interpretation Training. Engineering Investigation Training. Machine Learning model training managed by internal team. Coordinated savings pipeline: each team owns a savings number.
Target: $2B to $3B cumulative savings identified and tracked
PHASE 4
MONTHS 31–36
Embed and Handover
Full documentation of all models, dashboards, and governance. Internal centre of excellence. Data Sentinels in advisory role only: you are self-sufficient.
Outcome: Sustainable change. Zero consultant dependency.
PROOF OF WORK
We Have Done This Before.
Not in theory. In actual mines. With actual machines. And actual savings.
“You don’t need another consultant. You need someone who will get stuck in, deliver the savings, and hand over the capability.”
WHERE WE START
The First 90 Days.
No strategy documents. No steering committee workshops. We go straight to the data.
DAYS 1–14
System Access and Stakeholder Map
We need to understand the current state before we change anything. Most failed transformations skip this step.
DAYS 15–45
Data Health Assessment
You cannot reduce costs with bad data. Fixing the foundation creates immediate credibility.
DAYS 46–90
Opportunity Heat Map and Savings Pipeline