Main site

HR AI agent · Employee relations

Employee Relations Case Intake AI Agent

Record the complaint at the hour it lands, put the questions that decide which clock runs in front of a named employee relations officer, and hold the classification for that officer to sign.

Typical delivery
4–6 weeks
Deployment
Your stack
Named officer
Notice recorded
After launch
Agent Care

What this agent does

Records the receipt, never rules on it

In

  1. 01

    A complaint arrives, and the hour it was received is written down before anything else.

  2. 02

    Notice is what creates the duty, and Faragher, 524 U.S. 775, turns on what happened next.

Reason

  1. 03

    A class is proposed, and the period under 42 U.S.C. 2000e-5(e)(1) is named beside it.

  2. 04

    A discharge or demotion surfaces, and under Ellerth, 524 U.S. 742, the defence is gone.

  3. 05

    A report reaches someone able to investigate, and 18 U.S.C. 1514A(a)(1)(C) is in play.

Decide

  1. 06

    An acknowledgement falls due within seven days under Directive (EU) 2019/1937 Article 9.

  2. 07

    A file holds special-category data, and GDPR Article 9(2)(f) is the condition relied on.

Out

  1. 08

    A prior report names the same respondent, and the earlier file travels with this one.

  2. 09

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

Product statement

Receipt, the hour and the questions that decide the clock belong to the agent. The classification, the closure and any instruction belong to a named employee relations officer.

Example workflow

One complaint, receipt to determination

  1. 1Complaint receivedHotline call, web form, manager referral, union channel or a message sent straight to a person
  2. 2Receipt fixed and timestampedThe channel, the hour, the words as written, the respondent named and the day the clock starts
  3. 3Class proposed, clocks namedThe proposed class, each deadline it opens, the questions still unanswered and confidence
  4. 4Controls appliedTaxonomy checks, clock-start checks, prior-matter checks and classification confidence

No human action required

Stages 1 to 4 run unaided, and nothing is determined at any of them — the agent is recording, and the officer lane opens at the confidence gate.

5DecisionSplits at the confidence gate
Class well evidenced

Goes to the named officer to sign.

Anything ambiguous

Adds a counsel read first.

Officer determination

The matter is held with its account, its proposed class and the deadlines already running.

Confirm · Reclassify · Send to counsel review

Determined — by the named officer

  1. 6Case and records systems updatedOnly where write access and retention policy allow it
  2. 7Outcome evaluatedClassification stability, acknowledgement timing, officer reclassifications and what review found

Reclassifications

Each officer reclassification is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Deciding conduct was severe or pervasive.
    • Closing a matter, or calling it resolved.
    • Issuing a confidentiality instruction.
    • Ruling a report out of scope.

  • Automation boundary

    Agent acts unaided

    • ✓Timestamp the complaint at the hour it was received.
    • ✓Propose the classification, and name the acknowledgement clock it starts.
    • ✓Carry the acknowledgement deadline where the assigned officer sees it.
    • ✓Hold the matter open for the officer who signs it.

    Nothing closes and no confidentiality instruction issues except by the named officer.

    • Judging whether an affirmative defence survives.
    • Telling a reporter what the outcome will be.
    • Deciding that a file is privileged.
  • Changes to the taxonomy, the routing or the clocks.

Example output

One complaint, annotated

This serves an employee relations team that may have to show, years later, what was received and when — and after Murray v. UBS Securities the employer justifies a later step by clear and convincing evidence; below is one intake exactly as the agent leaves it.

Intake record · single complaint

Illustrative example
Complaint
Conduct report, respondent named
Proposed class
Possible protected-category report, unconfirmed
Received
Logged at 09:14, 3 August 2026
Evidence attached
Reporter account, one attachment
Confidence
Held unclassified
To be signed by
The named officer, by name

As received

Taken from the reporter account and the channel log, and it asserts nothing beyond what actually arrived.

What the intake holds

  • Reporter account
  • Channel and hour
  • Earlier related file

Why no classification here

Saying what a complaint is in law is a call a named officer makes.

What the score decides

Below the configured threshold an intake picks up a counsel read before the officer.

Action

ConfirmReclassifySend to counsel 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 complaint

From the channel it arrived on

03Receipt

Where the wording itself counts

In SEC Release 34-74619 the wording of a confidentiality paragraph read to interviewees was itself the violation of 17 CFR 240.21F-17(a), and nobody had been silenced.

01Approved path

Receiving it is the act

The duty attaches on receipt, and Burlington Northern, 548 U.S. 53, then asks whether the process would deter a reasonable worker from complaining.

02Human review

Checked for, and not found

No published benchmark was found for the accuracy of employee-relations complaint classification, and no authority was found on whether receipt by an automated system is notice. The Board standard for workplace rules is unsettled, so no wording is treated here as safe.

04

Build an evidence trail

The complaint, the hour it landed and the officer who classified it stay together.

Integrations

Typical integrations

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

  • Case management systems

    • NAVEX One
    • Convercent
    • Ethico
    • Case IQ
    • HR Acuity
    • Vault

  • Reporting channels

    • Hotline
    • web form
    • email
    • Manager and steward referrals

  • Core HR records

    • Workday
    • SAP SuccessFactors
    • Reporting lines and the prior matters

Agent

Employee relations case intake

  • Records the receipt
  • Proposes the class
  • Holds for the officer

  • Legal hold and records

    • Matter management
    • e-discovery
    • Retention and legal-hold schedules

  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

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

Agent controls

Six bolts between the model and the file

Six bolts across one door, the last the heaviest. What opens is set out in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to recording receipt when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and taxonomy rules, and note the version each intake was classified under.Track
L4TraceabilityRecord the hour of receipt, the class proposed and every read of the file, since 42 U.S.C. 2000e-3(a) protects the reporter.Record
L3Officer releaseHold the matter for the designated representative 2 CCR 11023(b)(5)(A) contemplates; the hold governs release, not whether the class is right.Gate
L2Taxonomy guardrailsTest any standard wording against 17 CFR 240.21F-17(a); Release 34-105542 shows the rule still drawing orders in 2026.Restrict
L1Confidence thresholdsRoute an ambiguous intake to a counsel read before the officer is asked to sign it.Require review
Model coreIntake recorded — the account, the hour, the proposed class and the clocks it opens
L1 – L2
Test whether a classification may stand
L3
Leaves the determination to a named officer
L4 – L5
Keep the complaint and the classification behind it
L6
Escalates to a named officer when signals degrade

How Nestack evaluates it

Evaluate the whole intake — not only the class that comes out.

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

Surface — the class the case file carries

  1. E1Final-output evaluationDid the intake record the hour of receipt and the clock that started with it?
  2. E2Step-level evaluationDid the agent read the right channel, the right history and the live taxonomy?
  3. E3Tool evaluationDid it read and write the correct matter and the correct respondent?
  4. E4Confidence calibrationDo low-confidence intakes actually attract more officer reclassifications?
  5. E5Slice evaluationHow does performance change across specific complaint classes?
  6. E6Business outcomeHow many intakes were reclassified before the officer signed?

Floor — the duty the employer answers for

Failure modes

Where each failure originates in the agent

Seven failure modes, set at the stage each one first becomes visible.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • TY-03

      Prior notice not retrieved

      Earlier reports on the same respondent stay unseen.

    Stage gathers

    The account, the channel, the hour and the history

  2. 02 · Reasoning

    2 modes

    • TY-04

      Protected-category content missed

      A report reads as an ordinary workplace grievance.

    • TY-06

      Tangible action never raised

      A discharge or demotion is left out of the account.

    Stage proposes

    The classification, the clocks started and confidence

  3. 03 · Tool / write

    2 modes

    • TY-02

      Confidentiality text issued by default

      A standard paragraph goes out untailored to the matter.

    • TY-05

      Acknowledgement lapses, clock runs

      No receipt is sent and the feedback window opens anyway.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • TY-01

      Closed on a rule, not a determination

      A matter is shut with no reasoned ground behind it.

    Stage returns

    The class a file carries and a tribunal reads

  5. 05 · Change / Version

    1 mode

    • TY-07

      Analytics accrue with no privilege

      Voluntary self-review carries no privilege of its own.

    Stage tracks

    Model, prompt, taxonomy rules and clock config

  • Sev-1 · a matter closed by the agent alone
  • Sev-2 · a duty attaches and nobody is told
  • Sev-3 · signal degrades, intake held back
See the complete Human Resources failure-mode catalogue

Affected slices

Unlabelled reports absorb the reclassifying

A class-level accuracy figure can read clean while unlabelled conduct reports carry most of the reclassifying. Nestack reports the officer-reclassification rate by class, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Unlabelled conduct reports12.6%3.7×Review
Cross-border reports8.9%2.6×Review
Retaliation follow-ons5.6%1.6×Watch
Named-policy complaints2.5%0.7×Normal

Bar: reclassification-rate lift vs. named-policy baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What an unread complaint costs

A cycle shuts when the account closed before anybody read it is a standing case. That suite is what the next intake window is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Reclassification rate rises on unlabelled conduct reports.

  2. 02

    Diagnose

    The matter that sat for a year because the first reader called it a personality clash is worked backwards until one cause is left standing.

  3. 03

    Improve

    Cases ship numbered, and the accounts behind them travel attached.

  4. 04

    Verify

    Each touched classification case is run again, and one red holds it back.

  5. 05

    Learn

    The case is kept, and the intake rules are amended in that same commit.

  6. Learn → Detect

    The return edge. The next intake 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, channels, intake classification, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Intake workflow and automation-boundary definition.Week 1
  2. 02Channel, hotline and case-system review.Week 1, Week 2
  3. 03Complaint routing and employee-relations rule mapping.Week 1, Week 2
  4. 04Complaint intake capture.Week 2
  5. 05Classification logic and clock binding.Week 2, Week 3
  6. 06Confidence scoring and counsel routing.Week 3
  7. 07Officer determination workflow.Week 3, Week 4
  8. 08Case-system and records integration.Week 5
  9. 09Classification and notice cases.Week 4, Week 5
  10. 10Guardrails and escalation controls.Week 4
  11. 11Intake-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 channel, one populationProductionProduction case workflowAdvancedMultiple channels / jurisdictions
Introduced at Pilot
Classification to your taxonomy✓✓✓
Named officer determination✓✓✓
Channel-inventory baseline✓✓✓
Introduced at Production
Reporting by complaint class—✓✓
Officer review workflow in your systems—✓✓
Approved write-back—✓✓
Case-and-rota integration—✓✓
Introduced at Advanced
Multi-jurisdiction notice rules——✓
Multi-stage officer sign-off——✓
Heavy channel traffic——✓
Multi-channel escalation controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, channel coverage, 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 live intake channels and the officer each one names

    Channel capture and receipt timestamping

    Week 1

  2. 02Representative past complaints and how each was classified

    Channel binding, classification logic and the intake baseline

    Week 2

  3. 03Your notice deadlines and the rota that answers them

    Taxonomy mapping, clock binding and the automation boundary

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Channel, hotline and case-system assessment, then integration setup

    Week 2

  5. 05Files you would not want produced

    Notice cases and the evaluation round

    Week 4

  6. 06What no intake may resolve

    Confidence scoring, counsel routing, guardrails and release controls

    Week 3

  7. 07A named officer who signs the determination

    Officer sign-off 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 band is as wide as its own phase costs, so the fifth week shows a pair where one would look neater.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Channel discovery, taxonomy mapping and the automation boundary

W2Channel integration and the intake-classification baseline

W3Classification logic, clock binding and release controls

W4Evaluation suite, notice cases and failure-mode testing

W5Case-system integration, pilot intakes and targeted corrections

W6One intake cycle run under the employee-relations lead, then Agent Care handover

Reading the band
A band sits only on the weeks its own work is named in, and the fifth week is genuinely doubled.
At the end of W6
Once the intake record validates, Agent Care adopts the agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · HR AI agent

Build a case intake agent around the receipt that started the duty.

Show us one complaint channel and the last matter that came through it. If nobody can say what hour it arrived, who was told and which clock started, the record you would rely on is the record you do not have. Receiving it was the act.

Typical delivery
4–6 weeks
Deployment
Your intake channels
Named officer
Notice recorded
After launch
Agent Care available

Nestack Agents · Employee relations case intake

AGT-HR-04 · Agent Care available after launch

Build a case intake agent

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

About you
Your complaints

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