Public relations AI agent · Internal communications

Internal Communications AI Agent

Send in the lawful order: draft the announcement, record which group was told and on what date, and leave the release to the named signatory whose letter goes to the Commissioner.

Typical delivery
4–6 weeks
Deployment
Your stack
Named signatory
Sequence first
After launch
Agent Care

What this agent does

Drafts the notice, never sends it

In

  1. 01

    An announcement is drafted, and the group it would reach is written down beside it.

  2. 02

    A date is set on the draft, because 12 NYCRR 921-2.1 reads notice as ninety calendar days.

Reason

  1. 03

    A federal notice falls due, and 29 U.S.C. 2102(a) puts sixty days ahead of the closing.

  2. 04

    A works council is owed a hearing, and § 111 BetrVG places it ahead of the send.

  3. 05

    An opinion is owed before adoption, and Directive (EU) 2025/2450 adds a written response.

Decide

  1. 06

    A public authority is notified, and Article 4 of Directive 98/59/EC adds thirty days.

  2. 07

    A read receipt is switched on, and BAG 1 ABR 16/23 makes that a co-determined device.

Out

  1. 08

    A concern channel is described, and Article 9 of Directive (EU) 2019/1937 sets its clocks.

  2. 09

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

Product statement

Drafting, sequencing and the record belong to the agent. The release, and the signature on the letter that goes to a public official, belong to a named signatory.

Example workflow

One announcement, decision to release

  1. 1Decision recordedA board commitment, a restructuring plan, a site order or a line from the works council file
  2. 2Audience and dates assembledWhich groups the measure touches, which bodies are owed consultation in good time under Article 2, and what is still open
  3. 3Announcement draftedThe draft, the groups named on it, the order it sits in and confidence
  4. 4Controls appliedSequence checks, consultation checks, register checks and drafting confidence

No human action required

Stages 1 to 4 run unaided, and nothing is sent at any of them — the agent is drafting, and the signatory lane opens at the release gate.

5DecisionSplits at the release gate
Consultation closed

Goes to the named signatory to release.

Anything ahead of a body

Adds an employment-counsel read first.

Signatory review

The draft is held with its audience list, the bodies owed notice and the date it sits under.

Release · Amend · Send to counsel review

Released — by the named signatory

  1. 6Notice and consultation records updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedSequence accuracy, groups missed, signatory corrections and what the later review turned up

Corrections

Each correction the signatory makes counts in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Signing the letter to the Commissioner.
    • Deciding when a consultation has closed.
    • Fixing the date an order takes effect.
    • Composing sentences a named leader will own.
  • Automation boundary

    Agent acts unaided

    • ✓Draft the announcement from the decision record.
    • ✓Record which group is told, on what date, and by which channel.
    • ✓Hold the send where a consultation body has not yet been heard.
    • ✓Keep each draft for the signatory who releases it.

    Nothing is sent except by a named signatory, inside the agreed boundaries.

    • Judging whether a measure is transnational.
    • Deciding which staff a notice must reach.
    • Choosing which works council is competent.
  • Changes to the sequencing rules or the gate.

Example output

One announcement, annotated

Severance design and manager support sit with human resources; this page holds the notice, the order it goes out in and the register under it. Below is one draft exactly as the agent leaves it.

Draft output · single announcement

Illustrative example
Measure
Site closure, one New York site
Recorded as
Consultation open, opinion not yet given
Audience
Draft only
Evidence of record
Workforce register, 3 August 2026
Confidence
Held unsent
Held for
The named signatory, by name

As received

Assembled from the register, payroll, timekeeping and the plan, and it names no group those four do not.

What the record holds

  • Workforce register
  • Consultation file
  • Notice filing trail

Why no release here

Releasing a notice to the staff is an act a named signatory owns.

What the score decides

Below the configured threshold a draft gets a counsel read before the signatory sees it.

Action

ReleaseAmendSend 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

Each announcement

From the decision that dates it

03Announcement

Where the announcement is used

Federal notice under 20 CFR 639.7 names a company official and takes no signature; 12 NYCRR 921-2.2 wants a natural person to sign, for the same layoff.

01Approved path

Told before it was allowed

The clocks run from an event nobody in the drafting seat witnesses, and a tool still reading Directive 2009/38/EC as unamended has missed Directive (EU) 2025/2450 of 26 November 2025, transposed by 1 January 2028. Standing policy text is reached separately, by 17 CFR 240.21F-17(a).

02Human review

What was checked, and not found

No rule consulted prescribes the words an internal announcement must use, and a staff-wide briefing is not by itself a Regulation FD event, because employees sit outside the covered recipients listed at 17 CFR 243.100(b)(1). The Board standard for facially neutral workplace rules is unsettled.

04

Build an evidence trail

The announcement, the group it reached and the officer who signed it stay together.

Integrations

Typical integrations

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

  • Workforce and payroll

    • HRIS
    • payroll
    • timekeeping
    • Site, wage and separation records
  • Consultation records

    • Works council
    • EWC registers
    • Opinion and response trails
  • Distribution and intranet

    • Email
    • intranet
    • digital signage
    • Audience groups and send records

Agent

Internal announcement and notice

  • Reads the register
  • Drafts the notice
  • Holds the send
  • Filings and approvals

    • State portals
    • legal hold
    • Signed notices and dated approval trails
  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

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

Agent controls

Six latches between the model and the send

Six catches on one latch, the last the surest. What releases is drawn in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to draft-only output when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt and sequencing rules, and note the version each draft was written under.Track
L4TraceabilityRecord each draft, the groups under it, the date it was released and each read of the file.Record
L3Signatory releaseHold the draft for a named signatory; the hold governs release, not whether the date is right.Gate
L2Sequence guardrailsTest each draft against the consultation record, and return one that would run ahead of a body.Restrict
L1Confidence thresholdsRoute a draft with an unresolved audience to a counsel read before the signatory sees it.Require review
Model coreDraft written — the announcement, its audience, the bodies owed notice and confidence
L1 – L2
Test whether a send may go
L3
Leaves the release to a named signatory
L4 – L5
Keep the announcement and the consultation behind it
L6
Holds the send when signals degrade

How Nestack evaluates it

Evaluate the whole sequence — not only the message that goes out.

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

Surface — the announcement staff open

  1. E1Final-output evaluationDid the draft name the groups the measure actually touched?
  2. E2Step-level evaluationDid the agent read the live register, the right dates and the consultation file?
  3. E3Tool evaluationDid it read and write the correct measure and the correct notice record?
  4. E4Confidence calibrationDo low-confidence sends actually attract more signatory corrections?
  5. E5Slice evaluationHow does performance change from one audience class to another?
  6. E6Business outcomeHow many drafts needed a correction before the signatory released one?

Floor — the record filed with the state

Failure modes

Where each failure originates in the agent

Seven failure modes, each named at the point in the run it first shows.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • UJ-03

      Stale group list read

      The list read is not the one the org now has.

    Stage gathers

    The groups, the dates, the sites and the bodies

  2. 02 · Reasoning

    2 modes

    • UJ-04

      Clock run from the draft date

      A send date is counted from the briefing, not the order.

    • UJ-06

      Consultation treated as closed

      An opinion still owed is worked as given.

    Stage proposes

    The draft, its audience and the bodies owed notice

  3. 03 · Tool / write

    2 modes

    • UJ-02

      Notice put to the wrong body

      The local council is told where the group one sits.

    • UJ-05

      Staff send placed ahead of a filing

      A staff briefing runs before the filing it triggers.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • UJ-01

      Published wording narrows a channel

      A policy line tells staff whom they may tell.

    Stage returns

    The notice staff open and a state office files

  5. 05 · Change / Version

    1 mode

    • UJ-07

      Read receipts on by default

      Open tracking arrives with an upgrade nobody bargained.

    Stage tracks

    Model, prompt, send rules and group lists

  • Sev-1 · a send ahead of consultation
  • Sev-2 · a notice misses an affected site
  • Sev-3 · register thin, send held back
See the complete Internal-comms failure-mode catalogue

Affected slices

Statutory notices absorb the corrections

A class-level sequence figure can read clean while statutory redundancy notices carry most of the signatory corrections. Nestack reports the correction rate by audience class, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Statutory redundancy notices7.4%3.7×Review
Works council and EWC sends5.2%2.6×Review
Cross-border policy rollouts3.2%1.6×Watch
Routine intranet updates1.7%0.8×Normal

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

2 of 4 slices over threshold

Evidence-linked improvement

What an early send costs

A loop closes when a group told before consultation opened is a standing case. That suite is what the next announcement is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Correction rate rises on statutory redundancy notices.

  2. 02

    Diagnose

    The group that heard about the site order before the council did is worked backwards until one cause is left standing.

  3. 03

    Improve

    Number the announcement; the approvals that cleared it are filed beneath it.

  4. 04

    Verify

    A single unconsulted group case stops the whole send.

  5. 05

    Learn

    One case joins the suite, one line joins the sequencing rules.

  6. Learn → Detect

    The return edge. The next announcement meets a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope set beside the delivery timeline. The week structure follows the six-week plan — discovery, registers, sequencing workflow, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Internal-communications and automation-boundary scope.Week 1
  2. 02Register, consultation and notice sources.Week 1, Week 2
  3. 03Consultation-sequencing and named-signatory mapping.Week 1, Week 2
  4. 04Register and notice intake.Week 2
  5. 05Audience binding and draft logic.Week 2, Week 3
  6. 06Confidence scoring and counsel routing.Week 3
  7. 07Signatory release workflow.Week 3, Week 4
  8. 08Directory-system integration.Week 5
  9. 09Sequencing and consultation cases.Week 4, Week 5
  10. 10Guardrails and distribution controls.Week 4
  11. 11Distribution-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 audience class, one cycleProductionProduction internal-comms workflowAdvancedMultiple jurisdictions / sites
Introduced at Pilot
Drafting to your sequencing rules✓✓✓
Named signatory release✓✓✓
Workforce-register baseline✓✓✓
Introduced at Production
Reporting by audience class—✓✓
Counsel review workflow in your systems—✓✓
Approved register write-back—✓✓
Directory-and-intranet integration—✓✓
Introduced at Advanced
Multi-regime sequencing rules——✓
Cross-border notice packs——✓
Large workforce registers——✓
Multi-jurisdiction sequencing controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, register 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 live audience classes and the notice each one owes

    Sequencing-rule capture and clock versioning

    Week 1
  2. 02Representative announcements, registers and consultation files

    Source binding, sequencing logic and the register baseline

    Week 2
  3. 03Your approval chain and the signatories it names

    Sequencing-rule mapping, clock binding and the automation boundary

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

    Register, consultation and directory source assessment, then integration setup

    Week 2
  5. 05Notices you would not want timestamped

    Consultation cases and the failure round

    Week 4
  6. 06What no announcement may pre-empt

    Confidence scoring, review routing, guardrails and release controls

    Week 3
  7. 07A named signatory who releases the notice

    Handover to the named signatory, 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

Two of these phases share week five in fact, and no band was widened to make the column read neatly.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Sequencing-rule discovery, clock versioning and the automation boundary

W2Register and consultation integration and the workforce baseline

W3Audience binding, drafting logic and release controls

W4Evaluation suite, sequencing cases and failure-mode testing

W5Directory-system integration, pilot sends and targeted corrections

W6One announcement cycle run under the internal-comms lead, then Agent Care handover

Reading the band
A bar covers only the weeks its own work is named for, and week five carries a pair.
At the end of W6
When the approval 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 · Public relations AI agent

Build an internal comms agent around the announcement your last restructuring sent in the wrong order.

Show us one restructuring and the order its messages went out in. Not how well they read. Which group was told first, which body had not yet been consulted, and whose name went on the letter to the Commissioner. A send made out of order comes back as a case.

Typical delivery
4–6 weeks
Deployment
Your announcements
Named signatory
Sequence first
After launch
Agent Care available

Nestack Agents · Internal communications

AGT-COM-11 · Agent Care available after launch

Build an internal comms agent

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

About you
Your workforce

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