Facilities AI agent · OSHA recordkeeping

EHS Incident Recordkeeping AI Agent

Stage every reported injury beside the criteria it was read against under 29 CFR 1904.7, run both § 1904.39 clocks, and hold the annual summary a company executive certifies.

Typical delivery
4–6 weeks
Deployment
Your stack
One certifier
Two clocks
After launch
Agent Care

What this agent does

Assembles the case file, never the certification

In

  1. 01

    An injury is reported, and § 1904.29(b)(3) allows seven calendar days to enter it.

  2. 02

    A treatment is given, and § 1904.7(b)(5) closes the list of what counts as first aid.

Reason

  1. 03

    A worker comes off routine duties, and § 1904.7(b)(4) makes that restricted work.

  2. 04

    A death is learned of, and § 1904.39(a)(1) leaves eight hours to tell OSHA.

  3. 05

    An amputation is learned of, and § 1904.39(a)(2) leaves twenty-four hours.

Decide

  1. 06

    A case meets the closed list at § 1904.29(b)(7), and the name leaves the log.

  2. 07

    A calendar year shuts, and § 1904.41(c) sets 2 March for the electronic filing.

Out

  1. 08

    A summary goes up 1 February, and § 1904.32(b)(6) holds it there until 30 April.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

The agent assembles the cases and runs the clocks. A company executive drawn from the § 1904.32(b)(4) list certifies the annual summary.

Example workflow

One case, report to certification

  1. 1Case evidence receivedIncident reports, treatment records, restriction notes or absence feeds
  2. 2Case context assembledThe injury, the establishment it happened at, the criteria it answers to and the day it was learned of
  3. 3Recordability evidence draftedThe case, the tests it was read against, the day counts and completeness
  4. 4Controls appliedWork-relatedness checks, first-aid checks, day-count checks and completeness confidence

No human action required

Stages 1 to 4 run unaided, and nothing is certified at any of them — the agent is assembling, and the safety lane opens at the completeness gate.

5DecisionSplits at the completeness gate
Evidence sufficient

Goes to the certifying executive to approve.

Anything thin

Adds a safety manager read first.

Safety review

The case is held with its criteria, its day counts and the establishment it sits at.

Certify · Append evidence · Send to safety review

Certified — by a company executive

  1. 6Log and case records updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedRecordability accuracy, clock coverage, reviewer corrections and what the read found

Corrections

Each safety correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Certifying the annual summary under § 1904.32(b)(3).
    • Deciding that a reported case is not recordable.
    • Making the § 1904.39 report to OSHA.
    • Choosing which executive may sign.
  • Automation boundary

    Agent acts unaided

    • ✓Assemble a case beside the § 1904.7 criteria used.
    • ✓Run the seven-day entry clock from the day a report.
    • ✓Track both § 1904.39 clocks from the hour the employer learns.
    • ✓Hold the establishment roster that decides which forms are owed.

    Nothing is certified or reported except by a named person, inside the agreed boundaries.

    • Judging whether a case is work-related.
    • Telling OSHA the log is complete.
    • Setting the restriction a clinician recommends.
  • Changes to the log, the summary or the roster.

Example output

One case, annotated

Our construction safety compliance copilot works safe work method statements for a duty-holder; this record is what one injury carried onto the log.

Log entry · single case

Illustrative example
Case
Restricted work, § 1904.7(b)(4)
Recorded as
Read against the general recording criteria
Test
Restricted work
Evidence of record
Incident report, 3 August 2026
Confidence
Held uncertified
Held for
The certifying executive, by name

As received

Taken from the incident report and the treatment record — it reaches as far as those sources do.

What the record holds

  • Incident report
  • Treatment record
  • Restriction note

Why no certification here

Whether a case is recordable is a § 1904.7 judgement, not a model output.

What the score decides

Below the configured threshold a case picks up a safety read before the executive sees it.

Action

CertifyAppend evidenceSend to safety 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 case

From the site that reports it

03Evidence

Where the evidence is used

An injury log is not a safety programme: this page records what happened and holds the file, and it makes no claim about whether the hazard was ever controlled.

01Approved path

Recordable is a judgement

Recordability is settled against the general recording criteria of § 1904.7, and the agent supplies the rationale rather than the answer.

02Human review

What was checked, and not found

An eCFR banner reporting a 2026 amendment to part 1904 could not be matched to any Federal Register document, the standing of § 1904.12 was left unresolved, the text of directive CPL 02-00-172 was not read, no State Plan variation under § 1904.37 was surveyed, and no injury-rate or recordkeeping-cost figure was confirmed from a primary source.

04

Build an evidence trail

The injury, the test it was read against and the executive who certified stay on the log.

Integrations

Typical integrations

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

  • Incident and report sources

    • Report forms
    • hotline
    • Supervisor incident reports
  • Occupational health

    • Clinic records
    • case notes
    • Treatment and restriction records
  • People and establishments

    • HR system
    • roster feed
    • Hours and headcount records

Agent

EHS incident recordkeeping

  • Reads the cases
  • Assembles the log
  • Holds for the executive
  • Case and records systems

    • EHS platform
    • ticketing
    • Log, summary and retention records
  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

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

Agent controls

Six locks between the model and the executive

Six locks turned in order, the last one stiffest. Whatever opens them all is drawn in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to evidence assembly when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and recording rules; no section of part 1904 was amended in 2025 or 2026, and the electronic filing duty of 88 FR 47254 has bound since 1 January 2024.Track
L4TraceabilityRecord each case, the criteria behind it, the establishment it sits at and every read of the file.Record
L3Executive releaseHold the summary for a company executive named by § 1904.32(b)(4); the hold governs release, not whether an injury should have been logged.Gate
L2Criteria guardrailsTest each case against §§ 1904.4 to 1904.11 as configured; the first-aid list of § 1904.7(b)(5) is closed, and so is the privacy list at § 1904.29(b)(7).Restrict
L1Confidence thresholdsRoute a thin case to a safety read first; § 1904.32(a) puts the year-end log review ahead of the summary that follows it.Require review
Model coreEvidence assembled — the case, the criteria, the day counts and completeness
L1 – L2
Test whether a case may stand
L3
Puts the certification in a person's hands
L4 – L5
Keep the injury and the test behind it
L6
Holds the summary uncertified when signals degrade

How Nestack evaluates it

Evaluate the whole assembly — not only the log entry that comes out.

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

Surface — the summary an inspector reads

  1. E1Final-output evaluationDid the entry record what the case actually was?
  2. E2Step-level evaluationDid the agent read the right case, the right criteria and the live roster?
  3. E3Tool evaluationDid it read and write the correct case record and the correct log?
  4. E4Confidence calibrationDo low-confidence cases actually attract more safety corrections?
  5. E5Slice evaluationHow does performance change across specific establishment types?
  6. E6Business outcomeHow many cases needed a correction before the executive certified?

Floor — the log the employer answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, each fixed at the stage where it first surfaces.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • KR-03

      Stale roster read

      The headcount read is not the one now on site.

    Stage gathers

    The cases, the tests, the dates and the roster

  2. 02 · Reasoning

    2 modes

    • KR-04

      Recordability called early

      A case is logged before the criteria are met.

    • KR-06

      First aid read as treatment

      Care on the closed list is worked as recordable.

    Stage proposes

    The cases, their tests and completeness

  3. 03 · Tool / write

    2 modes

    • KR-02

      Thin case passed forward

      A case moves on without the safety read.

    • KR-05

      Bound to the wrong establishment

      An injury is filed against the wrong site.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • KR-01

      Certified, evidence unrecorded

      The record shows a certification but not what supported it.

    Stage returns

    The summary an executive certifies and OSHA reads

  5. 05 · Change / Version

    1 mode

    • KR-07

      Silent clock regression

      A configuration change moves the clock, not the case.

    Stage tracks

    Model, prompt, recording rules and log fields

  • Sev-1 · a fatality report past eight hours
  • Sev-2 · wrong evidence reaches the log
  • Sev-3 · source degrades, case held uncertified
See the complete Facilities failure-mode catalogue

Affected slices

Warehouse sites absorb the corrections

An establishment-level recordability figure can read clean while warehouse and distribution sites carry most of the rework. Nestack reports the correction rate by establishment type, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Warehouse and distribution sites7.5%3.6×Review
Manufacturing establishments5.4%2.6×Review
Field and service establishments3.3%1.6×Watch
Office establishments1.8%0.9×Normal

Bar: correction-rate lift vs. office establishment baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What a missed case costs

A cycle shuts when the late fatality report is a regression case. That suite is what the next case logged is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on warehouse and distribution sites.

  2. 02

    Diagnose

    The strain that was first aid on Friday and restricted duty on Monday is read back until one cause remains.

  3. 03

    Improve

    The change ships numbered, and the cases that forced it ride with it.

  4. 04

    Verify

    Nothing releases while one touched case entry is still red.

  5. 05

    Learn

    It is retained for good, and the recording rules are amended in that same commit.

  6. Learn → Detect

    The return edge. The next log is measured 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, case assembly, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Recordability-criteria and automation-boundary work.Week 1
  2. 02Incident, treatment and roster sources.Week 1, Week 2
  3. 03Case-to-criteria and clock-coverage rule mapping.Week 1, Week 2
  4. 04Case and treatment ingestion.Week 2
  5. 05Case, criteria and record binding.Week 2, Week 3
  6. 06Completeness scoring and review routing.Week 3
  7. 07Executive certification workflow.Week 3, Week 4
  8. 08Occupational-health system integration.Week 5
  9. 09Recordability and clock cases.Week 4, Week 5
  10. 10Guardrails and executive controls.Week 4
  11. 11Case-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 establishment, one yearProductionProduction recordkeeping workflowAdvancedMultiple sites / entities
Introduced at Pilot
Case assembly to your criteria✓✓✓
Certifying executive release✓✓✓
Establishment-roster baseline✓✓✓
Introduced at Production
Reporting by establishment—✓✓
Certification workflow in your systems—✓✓
Approved write-back—✓✓
Occupational-health integration—✓✓
Introduced at Advanced
Multi-site rosters——✓
Cross-year evidence packs——✓
Large case loads——✓
Multi-clock notification controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, case 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 establishments and the roster behind each one

    Establishment mapping and case capture

    Week 1
  2. 02Representative incident, treatment and restriction records

    Record binding, criteria logic and the log baseline

    Week 2
  3. 03Your reporting procedure under § 1904.35(b)(1)(i)

    Criteria mapping, clock binding and the automation boundary

    Week 1
  4. 04Access to relevant APIs, feeds or exports

    Incident, health and roster-source assessment, then integration setup

    Week 2
  5. 05Logs you would not want certified

    Recordability cases and failure testing

    Week 4
  6. 06What no injury log may establish

    Completeness scoring, review routing, guardrails and release controls

    Week 3
  7. 07A company executive who satisfies § 1904.32(b)(4)

    Certification 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

Each width is the time a phase honestly takes and not a layout choice, so week five holds a pair.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1EHS workflow discovery, criteria mapping and the automation boundary

W2Source integration and the recordability baseline

W3Case assembly, clock logic and release controls

W4Evaluation suite, recordability cases and failure-mode testing

W5Health-record integration, pilot cases and targeted corrections

W6One recordkeeping year run under the certifying executive, then Agent Care handover

Reading the band
Each bar covers only the weeks its own work is named for. The fifth carries a pair because the work does.
At the end of W6
Validation closes on live logs, and Agent Care picks up the watch.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Facilities AI agent

Build a recordkeeping agent around the case that changes character between Friday and Monday.

Show us one case and the criteria it was read against. Only § 1904.32(b)(4) says who may certify: an owner, an officer of the corporation, the highest-ranking official at that establishment, or that official's immediate supervisor.

Typical delivery
4–6 weeks
Deployment
Your establishments
One certifier
Two clocks
After launch
Agent Care available

Nestack Agents · OSHA recordkeeping

AGT-AF-09 · Agent Care available after launch

Build a recordkeeping agent

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

About you
Your establishments

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