Worker Exposure & Health Surveillance AI Agent

What this agent does

Assembles the record, not the determination

A shift is sampled, and the pump, the person who wore it and the task actually worked come with the sample.

A pump is fitted, and its calibration record and pre-shift examination time travel beside that sample.

Reason

A sample is despatched, and the custody trail is tracked against the transmission window the rule sets.

A result comes back, and the accreditation in force on the analysis date is checked, not the one current now.

A result reaches a limit, and the written window opens on the record beside the person named to send it.

Decide

A plan is read against each similar-exposure group, and the roles nobody sampled this quarter are named.

A certification nears its re-examination date, and the samples it would cover are named before it lapses.

Out

A surveillance offer falls due, the offer and the answer are recorded, and the record outlives the job.

A write runs only inside the approval boundaries agreed at implementation, and never ahead of a signature.

→ Product statement

The agent assembles and tracks; a person certified under 30 CFR 70.202 signs the sample, and a physician makes the medical determination.

Example workflow

One sampling round, pump to record

AgentHuman

1Shift sampledSampling round, contractor field notes or laboratory portal

2Record assembledCalibration logs, custody forms, rosters and accreditation certificates, each with its source

3Coverage comparedGroup coverage, validity flags, clock status and confidence

4Controls appliedCalibration and custody checks, certification-currency checks, clock checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is signed at any of them — the agent is assembling, and the hygienist's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold

High confidence

Goes to the occupational hygienist to read.

Low confidence

Adds a certified-person read first.

Hygienist review

The record is held with its evidence, its named gaps and the confidence.

Accept · Correct · Send for second read

Accepted — record filed for the period▼

6Exposure register updatedOnly where write access and approval policy allow it

7Outcome evaluatedCoverage completeness, invalid-sample causes, notification timeliness and inspection findings

Corrections

A gap found later is refilled by sampling again, not by a late entry.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Certifying the sample under 30 CFR 70.202.

Deciding which limit binds, or that one is exceeded.

Notifying a miner that a result reached a limit.

Reading an audiogram, a spirometry trace or a film.

Automation boundaryAgent acts unaided

✓Assemble the record from calibration, custody and roster sources.

✓Compare coverage against each similar-exposure group.

✓Run the notification and retention clocks from a recorded time.

✓Name what is missing and hold the record for the hygienist.

Writes stay inside the boundaries agreed at implementation, and never ahead of a signature.

Certifying the corrective-action record after an exceedance.

Deciding whether a worker may continue the work.

Circulating a health monitoring report to anyone.

Changes to sampling rules, clocks or approval thresholds.

Example output

One sampling round, annotated

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

Exposure record output · single sampling roundIllustrative example

Exposure group

Assembled line

Rule cited

Source of record

Confidence

Clock

Contractor maintenance

Two roles in this group were not sampled in the quarter just closed

30 CFR 70.208

Roster and plan

86%

Runs to the quarter end

As receivedTaken from the sampling plan, the roster and the calibration log — nothing on this side is written by the agent.

Evidence used Pump calibration log Chain-of-custody form Certification register

Why this is openThe quarter counts valid representative samples, and a person decides what follows.

ActionAcceptCorrectSend for second read

What the score decidesBelow the configured threshold the record goes to the certified person before the hygienist.

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value landsValue 01 – 04

Every sampling roundFrom the sample record

03Assembly

Take the part the safety

The safety assistant works a shift — SWMS, isolation, permit packs. This one works a person over years.

01Approved path

The sample has to survive

Routine rounds arrive already read against the plan, with their evidence attached.

02Human review

Send review to the gaps

A lapsed certification, a pump examined outside its window or an unsampled role is marked before the hygienist reads.

04Build an evidence trail

The result, the sampling plan behind it and the person who notified stay on the record.

Integrations

Typical integrations

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

Sampling and hygiene dataGravimetric and CPDM results
Noise dosimetry records

Laboratory and accreditationAnalytical laboratory reports
Accreditation certificates

People and rostersHR roster · contractor lists
Certification registers

Agent

Exposure record assembly

Reads the sample
Checks the coverage
Holds for review

Scheduling and recordsSurveillance appointment diary
Records archive · retention

Observability & evaluationOpenTelemetry · Langfuse
Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six checks between the model and the record

Six checks, outermost first. What the whole set still misses is drawn in the map below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modeHold assembly at draft-only when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, clock-rule and sampling-plan configuration changes.Track

L4Record traceabilityRecord the sources, the comparison, the corrections and what stayed open.Record

L3Hygienist reviewHold records for the named hygienist; the gate governs release, not whether the accepted record is complete.Gate

L2Policy guardrailsTest records against calibration, custody, certification and clock rules; a failure returns the record.Restrict

L1Confidence thresholdsRoute low-confidence records to the certified person before the hygienist sees them.Require review

Model coreRecord assembled — group coverage, validity flags, clock status and confidence

L1 – L2Test whether a record may stand

L3Puts the read in a hygienist's hands

L4 – L5Keep the result and the plan behind it

L6Reverts to record assembly when signals degrade

How Nestack evaluates it

Evaluate the assembly workflow — not only the finished record.

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

Surface — the record an inspector reads

Depth of coverage ▼

E1Final-output evaluationDid each item in the record match the source it came from?

E2Step-level evaluationDid the agent use the right group, the right plan and the right period?

E3Tool evaluationDid it read and write the correct sampling record and field?

E4Confidence calibrationDo low-confidence records actually attract more hygienist corrections?

E5Slice evaluationHow does performance change across specific exposure groups?

E6Business outcomeHow many records reached the hygienist with a gap, and how many were refilled?

Floor — the outcome the operator answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, placed at the stage where each one begins.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

DV-03

Superseded sampling plan

The group was read from a plan the revision replaced.

Stage gathersCalibration logs, custody forms and roster records

02 · Reasoning2 modes

DV-04

Coverage read as complete

A job title places a crew in the wrong exposure group.

DV-06

Accreditation checked late

Currency is read at report date, not analysis date.

Stage proposesGroup coverage, validity flags and confidence

03 · Tool / write2 modes

DV-02

Clock started from the file

The window is dated from when the report arrived.

DV-05

Refusal held off-record

An offer declined by email never reaches the record.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

DV-01

Notification not raised

The written window under 30 CFR 62.110 closes unflagged.

Stage returnsThe record the hygienist reads and an inspector reads

05 · Change / Version1 mode

DV-07

Silent clock-rule drift

A model or rule change narrows what is flagged.

Stage tracksModel, prompt, clock rules and plan configuration

Sev-1 · assembled outside the boundary Sev-2 · a gap reaches the filed record Sev-3 · lab data degrades, record holds

See the complete Mining failure-mode catalogue →

Affected slices

One exposure group can carry most of it

A programme-level sampling-coverage figure can read complete while a few exposure groups carry nearly all of the unusable samples. Nestack reports the unusable-sample rate by group, not only in total.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Contractor maintenance crews	8.2%	3.7×	Review
Blast crews and drillers	5.9%	2.7×	Review
Pre-strip and haul operators	3.2%	1.5×	Watch
Designated-area sampling	1.1%	0.5×	Normal

Bar: unusable-sample lift vs. the designated-area baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

The loop ends in a regression case

The loop shuts when the voided sample is a regression case. That suite is what the next result returned is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Unusable samples rise in one exposure group.

02Diagnose

The pump the worker took off at crib and nobody recorded is read back until one cause remains.

03Improve

Any change ships numbered, with the results that caused it attached.

04Verify

One exposure case still failing is enough to hold the release.

05Learn

One case joins the suite, one line joins the exposure record.

Learn → DetectThe return edge. Detection next time starts from a longer suite than this one.

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

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

01Exposure workflow discovery and boundary setting.

02Sampling and laboratory source assessment.

03Sampling plan, clock and retention rule mapping.

04Sample and result ingestion, normalised.

05Coverage comparison and source binding.

06Confidence scoring and exception routing.

07Hygienist review workflow.

08Laboratory and roster integration.

09Sampling and notification cases.

10Guardrails and privacy controls.

11Exposure-trail instrumentation.

12Deployment, documentation and Agent Care handover.

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 exposure group ProductionProduction sampling systems AdvancedMultiple sites / schemes

Introduced at Pilot

Assembly to your sampling plan✓✓✓

Hygienist review gate✓✓✓

Sampling-coverage baseline✓✓✓

Introduced at Production

Reporting by exposure group—✓✓

Review routing in your systems—✓✓

Approved write-back—✓✓

Laboratory-result integration—✓✓

Introduced at Advanced

Multi-scheme sampling rules——✓

Multi-stage hygiene approvals——✓

High sampling volume——✓

Multi-site retention controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.

Separate from buildBuild 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 bringWe build with it

01Your sampling plan and exposure-group structure → Sample and result ingestion and evidence mappingWeek 1

02Representative laboratory reports and calibration logs → Assembly baseline, coverage extraction and source bindingWeek 2

03Your clocks, retention periods and certification register → Clock, retention and certification mapping, and the boundaryWeek 1

04Access to relevant APIs, feeds or exports → Sampling and laboratory assessment, then integration setupWeek 2

05Records you would not want subpoenaed → Pump-calibration cases and failure testingWeek 4

06What no notification may withhold → Confidence scoring, exception routing, guardrails and privacy controlsWeek 3

07Named hygienists to read the records → Hygienist review workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Every band is exactly as wide as the weeks its phase consumes, so the fifth legitimately carries two.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Exposure workflow discovery, plan mapping and the automation boundary W2Laboratory integration and the assembly baseline W3Assembly workflow, confidence logic and review controls W4Evaluation suite, privacy controls and failure-mode testing W5Roster integration, pilot exposure groups and targeted corrections W6One monitoring cycle run under the occupational hygienist, then Agent Care handover

Reading the bandA bar covers only the weeks its own work is named in. The week 5 doubling is two phases at once.

At the end of W6Once the sampling round validates, Agent Care assumes the agent.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Mining AI agent

Build an exposure-record agent around your sampling plan and your hygienist.

Show us your sampling plan, your calibration logs and your clocks. Not a system that reads a medical result — one that assembles the record around it and dates what is contested, including the 2024 silica rule, whose compliance dates were stayed as at 18 August 2026.

Discuss this agent →

Nestack Agents · Exposure recordsAGT-MN-13 · Agent Care available after launch