DiscvrAI

Growth & Portfolio Intelligence

Mine Cost Reconciliation Agent

Reconciled cost per tonne, BCM, and machine-hour with no variance plug.

Target outcome: Reconciled $/tonne

The business problem

1 to 2% revenue, over 2% EBITDA from SKU focus

Mining COP is equipment-hour-dominated with joint overburden-and-mineral output. Monthly MIS hides tyre life, overhaul spikes, stripping ratio drift, and contractor claim gaps.

What we deploy

Mine Cost Reconciliation Agent

Equipment cost-per-hour, four cost objects held simultaneously, and a residual-free variance bridge extended for stripping ratio, lead creep, and contractor claims on opencast and underground mines.

How this agent runs

Every agent follows the same operating loop on your stack: audit the evidence, decide with confidence scores, execute with human approval, and record results back to your systems.

Audit

Full coverage on the transactions and documents you choose: contracts, invoices, dispatches, disputes.

Decide

Confidence-scored recommendations with counterfactuals, tax feasibility, and policy rules applied first.

Execute

Human-in-the-loop for consequential actions. Agents prepare; your team approves.

Record

Clean results posted back to your ERP, WMS, or operational systems with a full audit trail.

What we need from you

Data and systems

  • ERP and WMS sales, margin, and inventory history
  • SKU master, plant/BU hierarchy, and planning calendars
  • Existing BI or spreadsheet definitions leadership already trusts

Typical timeline

Portfolio or forecast workbench in 8 to 12 weeks. Executive consolidation views follow once definitions are aligned.

8 to 12 weeks on your exports and operational data. A defensible outcome number before you commit to rollout. We agree the baseline and success metric before deployment starts.

Key capabilities

Rolling cost per SMU-hour for every machine with component amortisation

Four cost objects ($/BCM, $/ROM t, $/saleable t, $/hour) with explicit cascade

Residual-free variance bridge and contractor or levy evidence packs

How we deploy it

8 to 12 weeks on your exports and operational data. A defensible outcome number before you commit to rollout. Same operating model on every agent: scope, deploy, measure, then scale or hand off.

Step 1

Scope

Agree one outcome, baseline KPI, and the exports or systems the agent will read.

Step 2

Deploy

Agent live on your ERP, WMS, or operational files within the engagement window.

Step 3

Measure

Track results against the baseline with named transactions and audit trails.

Step 4

Scale

Expand to adjacent modules or run independently. You own the code from day one.

Good fit if

  • Mine cost per tonne is asserted, not reconciled to machine hours
  • Stripping ratio or lead creep moves the number without a clean bridge
  • Contractor and royalty claims are reconciled manually under time pressure

Production module in the field

COP Intelligence for Mining

Case studies, module depth, and technical FAQs for the platform this agent runs on.

View COP Intelligence for Mining

Frequently asked questions

What cost objects does mining COP cover?+

Cost per BCM of overburden, per ROM tonne, per saleable tonne, and per machine operating hour, with explicit conversion between them.

How is equipment cost handled?+

Each machine gets a rolling cost per SMU-hour from fuel, tyres, GET, spares, and amortised overhauls, not spiked into a single month.

Does the variance bridge have a plug?+

No. Price, usage, stripping ratio, yield, grade, lead, availability, volume, and mix isolate exactly and sum to total variance on your data.

Can it reconcile contractor and levy claims?+

Yes. Surveyed vs claimed volume, owner-supplied inputs, escalation clauses, and statutory levies produce evidence packs with source lineage.

What is the typical pilot path?+

Scope one pit or section, run a historical replay on 3 to 6 months of exports, then parallel-run against the mine cost sheet with under 1% variance as the gate.

Ready to scope this agent?

8 to 12 weeks on your exports and operational data. A defensible outcome number before you commit to rollout.