Main site

Mining AI agent · Emergency notification

Emergency Notification & Readiness AI Agent

Run the statutory notification clock from when a reportable event is known, draft the notification against the regulator form, and keep the readiness record true — a named mine official makes the call.

Typical delivery
4–6 weeks
Deployment
Your stack
Named officials
Clock-driven
After launch
Agent Care

What this agent does

Runs the clock, not the emergency

In

  1. 01

    An event occurs, and the record is pulled from supported control-room, dispatch, shift-log and roster sources.

  2. 02

    A clock starts when the operator knew or should have known, not when somebody finished deciding.

Reason

  1. 03

    A notification is drafted against the fields the regulator form asks for, and held for the official who makes it.

  2. 04

    A scene is preserved, and the duty that attached at the event is surfaced with the record it has to leave behind.

  3. 05

    A filing window opens, and the accident report and the quarterly return are tracked against the event that began them.

Decide

  1. 06

    A drill falls due, and coverage is counted against the roster that actually worked, not the roster on the plan.

  2. 07

    A physical expires, and the agent tracks the date on the form and never what the physician wrote inside it.

Out

  1. 08

    A plan is approved, and the training window is counted from the approval date, not the date it was submitted.

  2. 09

    A write action runs only inside the approval boundaries agreed during implementation.

Product statement

The agent runs the clock and keeps the record; a named mine official notifies, and a record found wrong is corrected after it was relied on.

Example workflow

One event, first report to filing

  1. 1First report receivedControl room, dispatch log, shift record and medical transport note
  2. 2Clock startedTime of first knowledge, candidate event classes and the duties that attach, each with its named source
  3. 3Notification draftedDraft notification, form fields, preservation checklist, open items and confidence
  4. 4Controls appliedForm-field completeness, clock checks, preservation and retention tests and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is notified at any of them — the agent is drafting, and the official lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the mine official.

Low confidence

Adds a safety manager read first.

Official review

The draft is held with its clock, its open items and the confidence.

Approve · Amend · Send to safety manager review

Approved — released to the record

  1. 6Notification record updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedAmendment distance, open-item outcomes, regulator queries and post-filing corrections

Amendments

Every official amendment is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Declaring an emergency, or ordering an evacuation.
    • Deciding whether an event is reportable.
    • Making the notification to the regulator.
    • Taking charge on shift during a mine emergency.

  • Automation boundary

    Agent acts unaided

    • ✓Start the clock at the first record of the event, and show it running.
    • ✓Draft the notification against the fields the form requires.
    • ✓Surface the preservation and record duties that attach at the event.
    • ✓Name what has expired in the readiness record, and hold the pack.

    Writes run only inside the boundaries agreed at implementation, never ahead of a notification.

    • Certifying a drill, a team or an apparatus test.
    • Clearing a rescue team member as medically fit.
    • Judging a scene-preservation exception in real time.
  • Changes to event-class, clock or approval rules.

Example output

One notification, annotated

Everything the agent drafts is attached to the record it was drawn from.

Notification output · single event

Illustrative example
Event class
Underground fire, no injury
Notification drafted
Fire underground not extinguished inside ten minutes
Time to notification
22 minutes
Source of record
Control-room log on file
Confidence
86%
Notifier of record
Shift foreman, named on the roster

As received

Taken from the control-room log and the shift record — nothing on this side is written by the agent.

Source facts used

  • Control-room event log
  • Shift and dispatch record
  • Suppression report

Why this is open

The log shows when the fire was called in, and nothing shows the official being told.

What the score decides

Below the configured threshold the item picks up a safety manager read before the official sees it.

Action

ApproveAmendSend to safety manager review

Value

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 reportable event

From the first record

03Clock and pack

Start the clock on the record

The window opens when the operator knew or should have known, so the time spent deciding is already time elapsed.

01Approved path

Doubt resolves toward calling

In Canyon Fuel, decided 30 January 2026, the Commission held the window opened as soon as the foreman was told.

02Human review

Not the alarm, and not the response

Our maintenance planning agent carries equipment currency; our safety compliance assistant carries training and document currency. This page owns the clock that starts at an event.

04

Build an evidence trail

The event, the clock it started and the official who notified stay on the notification.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Control room and dispatch

    • SCADA
    • dispatch systems
    • Shift logs and rosters

  • Regulator forms

    • MSHA 7000-1
    • 7000-2
    • Form 5000-3 records

  • Training and competency

    • Pegasus
    • Onsite Track Easy
    • Drill and training ledgers

Agent

Emergency notification

  • Reads the record
  • Runs the clock
  • Holds for the official

  • Rescue and equipment records

    • Apparatus inspection logs
    • Self-rescuer test records

  • 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 notification

Six layers, outermost first. What the whole set still lets by is drawn in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow to clock tracking when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, event-class and form-version changes.Track
L4TraceabilityRecord the event time, the clock, the open items and the approval time.Record
L3Official approvalHold the pack for the named mine official; it governs release of the pack, not whether the call went in on time.Gate
L2Policy guardrailsTest the pack against the form, preservation and retention rules on file; a failure returns the pack.Restrict
L1Confidence thresholdsRoute low-confidence items to a safety manager read before the official sees them.Require review
Model corePack assembled — clock, draft notification, open items and confidence
L1 – L2
Test whether the pack may stand
L3
Puts the release in the hands of an official
L4 – L5
Keep the event and the clock behind it
L6
Narrows to clock tracking when signals degrade

How Nestack evaluates it

Evaluate the notification workflow — not only the finished form.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the notification the regulator reads

  1. E1Final-output evaluationDid every field trace to the record behind it?
  2. E2Step-level evaluationDid the agent use the event-class list and form version in force?
  3. E3Tool evaluationDid it read and write the correct event and the correct mine?
  4. E4Confidence calibrationDo low-confidence items actually attract more official amendments?
  5. E5Slice evaluationHow does performance change across specific event classes?
  6. E6Business outcomeHow many items were amended, or corrected after the form had gone in?

Floor — the outcome the operator answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed where the clock first goes wrong.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • EE-03

      Clock read from the finding

      The first record of the event was not the one read.

    Stage gathers

    Event records, rosters, forms and certifications

  2. 02 · Reasoning

    2 modes

    • EE-04

      Injury reflex applied

      An event with nobody hurt is read as no event.

    • EE-06

      Window run from submission

      Training is counted from the wrong date.

    Stage proposes

    Clock, notification draft, open items, confidence

  3. 03 · Tool / write

    2 modes

    • EE-02

      Back-dated certification

      A late test is signed as though it ran on time.

    • EE-05

      Roster counted, not coverage

      A drill certificate hides who never attended.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • EE-01

      Investigators not named

      The form carries a department, not people.

    Stage returns

    The notification pack the mine official reads

  5. 05 · Change / Version

    1 mode

    • EE-07

      Silent event-class drift

      A model or rule change narrows what reads as reportable.

    Stage tracks

    Model, prompt, event classes and form versions

  • Sev-1 · a wrong pack reaches the regulator
  • Sev-2 · a wrong basis reaches the record
  • Sev-3 · source degrades, item goes to review
See the complete Mining failure-mode catalogue

Affected slices

One event class can carry most of the returns

A site-level notification-timeliness figure can read as clean while a few event classes carry most of the returned items. Nestack reports the return rate by event class, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Underground fire and outburst8.8%3.7×Review
Entrapment and hoist damage6.1%2.5×Review
Metal and nonmetal surface4.4%1.8×Watch
Reported injury events2.0%0.8×Normal

Bar: return-rate lift vs. injury-event baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

A cycle ends when the suite grows by one

A cycle is done when the missed window has become a case the next release must pass. That suite is what the next notification drafted is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Return rate rises in one event class.

  2. 02

    Diagnose

    The fifteen minutes that started while somebody drove out to look are read back until the cause narrows to one.

  3. 03

    Improve

    Stamp the change; the events behind it are filed under that number.

  4. 04

    Verify

    Each touched notification case runs once more, and one red holds it.

  5. 05

    Learn

    It stays as a standing test, and the reporting rules move with it.

  6. Learn → Detect

    The return edge. Detection next time runs against one more case than last.

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, notification workflow, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Notification workflow discovery and boundary mapping.Week 1
  2. 02Event-source and record assessment.Week 1, Week 2
  3. 03Event-class, clock and record-retention mapping.Week 1, Week 2
  4. 04Event ingestion and clock start.Week 2
  5. 05Notification drafting and form binding.Week 2, Week 3
  6. 06Confidence scoring and exception routing.Week 3
  7. 07Official notification workflow.Week 3, Week 4
  8. 08Control-room and forms integration.Week 5
  9. 09Clock and readiness cases.Week 4, Week 5
  10. 10Guardrails and notification controls.Week 4
  11. 11Event-trail instrumentation.Week 5, Week 6
  12. 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 tierPilotOne mine, one event classProductionProduction control-room systemsAdvancedMultiple mines / regulators
Introduced at Pilot
Clock tracking and drafting✓✓✓
Mine-official approval✓✓✓
Readiness-currency baseline✓✓✓
Introduced at Production
Reporting by event class—✓✓
Exception workflow in your systems—✓✓
Approved write-back—✓✓
Control-room integration—✓✓
Introduced at Advanced
Multi-regulator form rules——✓
Multi-stage notification approvals——✓
High event volume——✓
Multi-regulator notification controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, event 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.

  1. 01Your event classes and notification paths

    Event-class, clock and record-retention rule mapping

    Week 1

  2. 02Representative prior notifications and forms

    Event ingestion, clock start and the drafting baseline

    Week 2

  3. 03Your emergency plan, drills and rescue records

    Readiness-record mapping and automation-boundary definition

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Event-source and record assessment, then integration setup

    Week 2

  5. 05Notifications you would not want timed

    Window cases and failure-mode testing

    Week 4

  6. 06What no readiness record may skip

    Confidence scoring, exception routing, guardrails and notification controls

    Week 3

  7. 07Named mine officials to make the call

    Official notification 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 schedule reflects the weeks actually worked, which is why one band in week five is a doubled one.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Notification workflow discovery, rule mapping and the automation boundary

W2Source integration and the drafting baseline

W3Clock logic, drafting workflow and notification controls

W4Evaluation suite, readiness-currency checks and failure-mode testing

W5Control-room integration, a pilot event class and targeted corrections

W6One drill cycle run under the emergency response coordinator, then Agent Care handover

Reading the band
A bar covers only the weeks its own work is named in. The week 5 overlap is real, not padding.
At the end of W6
The final checks clear on live drills and monitoring moves to Agent Care.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Mining AI agent

Build the notification clock before the fifteen minutes start.

Show us your event classes, your control-room records and the readiness ledger you keep. The record has to survive being read back years later by someone who already knows what happened and wants to know what you knew and when — so we build the clock, the pack and the ledger against that reading. During the event itself the agent is not in the loop, and the official keeps the call.

Typical delivery
4–6 weeks
Deployment
Your event records
Named officials
Clock-driven
After launch
Agent Care available

Nestack Agents · Emergency notification

AGT-MN-17 · Agent Care available after launch

Build an emergency notification agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your notification

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.