Mining AI agent · Closure provisioning
Mine Closure & Rehabilitation Provisioning AI Agent
Carry the disturbance ledger, unit rates and assumptions under a closure cost estimate, and reconcile it against the provision and the assurance lodged — a named person signs the plan, officers sign the accounts.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Named officers
- Pre-lodgement
- After launch
- Agent Care
What this agent does
Assembles the estimate, not the provision
In
- 01
Ground is disturbed, and the survey behind it is read from supported mine-planning, GIS and survey sources.
- 02
A hectare is classified by domain and status, and the reported rehabilitation figure is derived from that ledger.
Reason
- 03
An area stays disturbed until completion criteria are evidenced against it, rather than when seeding finishes.
- 04
A cost line is built on a dated unit rate, and the source and the date of that rate travel with the line.
- 05
A non-use management area keeps its management cost, so dropping a rehabilitation cost does not zero the domain.
Decide
- 06
A revision falls due on a trigger or a due date, and the mine-plan change that fired it is named on the estimate.
- 07
A levy basis stays apart from the closure cost estimate; WA guideline v6.0 rules the MRF estimate out for it.
Out
- 08
A regulatory estimate, a provision and the lodged instrument are bridged off one ledger, each difference attributed.
- 09
A write action runs only inside the approval boundaries agreed during implementation.
Product statement
The agent assembles the support and the bridge; a named person signs the closure plan, and the officers who sign the accounts carry the provision.
Example workflow
One estimate, ledger to lodgement
- 1Disturbance data receivedSurvey pickups, mine plan, GIS layers and the prior annual return
- 2Ledger assembledHectares by domain, status and year, each carrying the survey it came from
- 3Estimate builtCost lines, unit rates, assumption register and confidence
- 4Controls appliedLedger reconciliation, rate-currency checks, accuracy-range tests and confidence threshold
No human action required
Stages 1 to 4 run unaided, and nothing is lodged at any of them — the agent is assembling support, and the technical services lane opens at the confidence gate.
Goes to the technical services head.
Adds a cost-engineering read first.
Technical services approval
The estimate is held with its assumption register, its open variances and the confidence.
Approve · Amend · Send to cost review
Approved — released for lodgement
- 6Closure systems updatedOnly where write access and approval policy allow it
- 7Outcome evaluatedAmendment distance, variance outcomes, regulator queries and post-lodgement revisions
Amendments
Every reviewer amendment is counted in the evaluation.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Lodging the closure plan — WA guideline v6.0 names no signatory.
- Certifying a Code Part under O. Reg. 35/24 — a qualified person.
- Setting the rehabilitation provision in the accounts.
- Determining the assurance or the security held.
Automation boundary
Agent acts unaided
- ✓Maintain the disturbance ledger by domain, status and year.
- ✓Carry each cost line with its dated unit rate and the source it was drawn from.
- ✓Run the regulatory and accounting builds off the one ledger.
- ✓Bridge the estimate to the provision and attribute each difference.
Any write happens inside the boundaries agreed at implementation, never ahead of approval.
- Nominating an ERC period or a risk category.
- Approving a unit rate that departs from the tool.
- Drafting a sustainability or framework disclosure.
- Changes to rate, domain or approval rules.
Example output
One cost line, annotated
Everything the agent assembles is attached to the record it was drawn from.
Closure estimate output · single cost line
Illustrative example- Domain
- Waste rock dump
- Cost line
- Topsoil respread and deep ripping across the dump batters
- Disturbed area
- 126 ha
- Source of record
- Survey pickup, dated
- Confidence
- 89%
- Rate basis
- Contract schedule rate, dated
As received
Taken from the survey pickup and the contract schedule — nothing on this side is written by the agent.
Source facts used
- Survey pickup
- Contract rate schedule
- Prior annual return
Why this rate
It is the dated contract rate on file, carried with its date — not a rate the agent chose.
What the score decides
Below the configured threshold the line picks up a cost-engineering read first.
Action
ApproveAmendSend to cost reviewValue
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value lands
Value 01 – 04
Every disturbed area
From the disturbance ledger
Build off one ledger
One ledger feeds the regulatory estimate, the levy basis and the accounting provision.
Assumptions carry the estimate
A rate that was current when the plan was written stays in the estimate until somebody dates it.
Substance, not the clock
The tenement and permit agent tracks when a lodgement falls due; this one assembles what discharges it.
Build an evidence trail
The cost line, the unit rate under it and the officer who approved stay on the provision.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Mine planning and survey
- Deswik
- Vulcan
- Surpac
- Micromine
Spatial and disturbance
- ArcGIS
- QGIS
- Drone and satellite imagery
Environmental management
- Cority
- Enablon
- IsoMetrix
- SafetyCulture
Agent
Closure cost & provisioning
- Reads the ledger
- Builds the estimate
- Holds for approval
Finance and reporting
- SAP
- Oracle ERP
- Workiva
- close tooling
Observability & evaluation
- OpenTelemetry
- Langfuse
- Supported monitoring/evaluation sources
Integration availability depends on the client's existing systems and API access.
Agent controls
Six layers between the model and the estimate
Six layers narrowing as they run. What survives the innermost is named in the map below.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Test whether the estimate may stand
- L3
- Puts the release in the approver's hands
- L4 – L5
- Keep the cost line and the rate behind it
- L6
- Narrows to assumption tracking when signals degrade
How Nestack evaluates it
Evaluate the estimating workflow — not only the finished number.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the estimate the regulator reads
- E1Final-output evaluationDid every cost line trace to a hectare and a dated rate?
- E2Step-level evaluationDid the agent use the right ledger, rate set and jurisdiction configuration?
- E3Tool evaluationDid it read and write the correct domain and the correct period?
- E4Confidence calibrationDo low-confidence cost lines actually attract more reviewer amendments?
- E5Slice evaluationHow does performance change across specific rehabilitation domains?
- E6Business outcomeHow many lines were amended, or corrected after the plan was lodged?
Floor — the outcome the balance sheet carries
Failure modes
Where each failure originates in the agent
Seven failure modes, each set at the stage where it originates.
Agent lifecycle
Direction of processing →
01 · Retrieval
1 mode
- DK-03
Stale disturbance extent
Ledger built on last year's survey shapefile.
Stage gathers
Survey pickups, mine plan, GIS layers and returns
02 · Reasoning
2 modes
- DK-04
Undated unit rate
A rate is carried forward without its date.
- DK-06
Levy basis reused
MRF rates stand in for the closure estimate.
Stage proposes
Cost lines, unit rates, assumptions and confidence
03 · Tool / write
2 modes
- DK-02
Provision left behind
Accounts stay on the superseded estimate.
- DK-05
Status counted twice
Seeded ground is read back as rehabilitated.
Stage writes
Only where write access and approval policy allow it
04 · Output
1 mode
- DK-01
Range and method omitted
Estimate goes out with no stated accuracy range.
Stage returns
The estimate lodged and the provision recognised
05 · Change / Version
1 mode
- DK-07
Silent rate-set drift
A model or rate-table change moves cost lines quietly.
Stage tracks
Model, prompt, rate tables and jurisdiction config
- Sev-1 · a wrong estimate reaches a lodgement
- Sev-2 · a wrong line reaches the provision
- Sev-3 · survey degrades, line goes to review
Affected slices
One domain can sit under a clean total
A site-wide disturbance-reconciliation figure can read as clean while a few rehabilitation domains absorb most of the amendments. Nestack reports the amendment rate by domain, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Non-use management areas | 7.7% | 3.7× | Review | |
| Tailings storage facilities | 5.4% | 2.6× | Review | |
| Waste rock dumps | 3.3% | 1.6× | Watch | |
| Cleared laydown and hardstand | 2.2% | 1.0× | Normal |
Bar: amendment-rate lift vs. cleared-laydown baseline · scale 0–4.0× · tick marks the 2.0× review threshold
2 of 4 slices over threshold
Evidence-linked improvement
No cycle closes without a new test
A cycle is done when the stale assumption has become a case the next release must pass. That suite is what the next estimate rebuilt is measured against.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
Amendment rate rises in a rehabilitation domain.
- 02
Diagnose
The unit rate that was current when the plan was written, and never since, is read back to one cause.
- 03
Improve
Stamp the change; the cost lines behind it are filed against that number.
- 04
Verify
Each touched estimate case is run once more, and one red holds it back.
- 05
Learn
It stays as a standing test, and the rate rules move with it.
Learn → Detect
The return edge. The next estimate is detected against a suite one case longer.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, sources, estimating workflow, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Closure workflow discovery and boundary mapping.Week 1
- 02Survey, GIS and ledger source assessment.Week 1, Week 2
- 03Unit-rate, method and jurisdiction rule mapping.Week 1, Week 2
- 04Disturbance ingestion and reconciliation.Week 2
- 05Estimate assembly and source binding.Week 2, Week 3
- 06Confidence scoring and variance routing.Week 3
- 07Technical services approval workflow.Week 3, Week 4
- 08Survey and finance-system integration.Week 5
- 09Disturbance and unit-rate cases.Week 4, Week 5
- 10Guardrails and approval controls.Week 4
- 11Provision-trail instrumentation.Week 5, Week 6
- 12Deployment, documentation and Agent Care handover.Week 6
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel
Final scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
| Capability✓ in scope · — not at this tier | PilotOne site, one domain set | ProductionProduction survey systems | AdvancedMultiple sites / jurisdictions |
|---|---|---|---|
| Introduced at Pilot | |||
| Estimate assembly and reconciliation | ✓ | ✓ | ✓ |
| Technical services approval | ✓ | ✓ | ✓ |
| Disturbance-reconciliation baseline | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Reporting by domain | — | ✓ | ✓ |
| Variance workflow in your systems | — | ✓ | ✓ |
| Approved write-back | — | ✓ | ✓ |
| Survey-and-ledger integration | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-jurisdiction estimating rules | — | — | ✓ |
| Multi-stage technical approvals | — | — | ✓ |
| High domain volume | — | — | ✓ |
| Multi-jurisdiction provisioning controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on integrations, workflow complexity, disturbance volume, approval controls and deployment requirements.
- Separate from build
- Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bring
We build with it
01Your disturbance ledger and domain list
→Unit-rate, method and jurisdiction rule mapping
Week 102Representative prior closure estimates
→Disturbance ingestion, reconciliation and the assembly baseline
Week 203Your documented unit rates and assumptions
→Estimating-method mapping and automation-boundary definition
Week 104Access to relevant APIs, feeds or exports
→Survey, GIS and ledger assessment, then integration setup
Week 205Estimates you would not want re-costed
→Assumption cases and failure-mode testing
Week 406What no closure plan may leave undated
→Confidence scoring, variance routing, guardrails and approval controls
Week 307Named technical services reviewers
→Technical services approval workflow, then pilot and production validation
Weeks 5–6
Nothing else is required
Deployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
The plan is drawn against elapsed weeks rather than slide space, and so week five doubles up here.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Closure workflow discovery, rule mapping and the automation boundary
W2Source integration and the reconciliation baseline
W3Estimating workflow, confidence logic and approval controls
W4Evaluation suite, assumption checks and failure-mode testing
W5Survey integration, a pilot domain set and targeted corrections
W6One revision cycle run under the technical services head, then Agent Care handover
- Reading the band
- Each bar covers only the weeks its work is named in. The week 5 overlap is real, not padding.
- At the end of W6
- When the estimate validates, Agent Care takes the agent on.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Mining AI agent
Build the reconciliation the estimate, the provision and the bond all rest on.
The regulatory estimate, the levy basis and the balance-sheet provision are built on different premises, and making them equal breaks one of them. Show us your disturbance ledger, your rate schedule and who signs — we will map the bridge and set the automation boundary.
- Typical delivery
- 4–6 weeks
- Deployment
- Your survey and ledger data
- Named officers
- Pre-lodgement
- After launch
- Agent Care available
Nestack Agents · Mine closure and provisioning
AGT-MN-11 · Agent Care available after launch