Main site

Transportation AI agent · Check calls

Carrier Voice Check-Call AI Agent

Place check calls that open with who is calling, that it is an AI and that the line is recorded — with a consent record on file first, and a person on request.

Typical delivery
4–6 weeks
Deployment
Your stack
Person on ask
Consent first
After launch
Agent Care

What this agent does

Discloses before it asks, on every call

In

  1. 01

    Checking the consent record for the number before anything is dialled, from supported TMS or CRM sources.

  2. 02

    Treating the number as a mobile by default, because in owner-operator trucking the business line usually is one.

Reason

  1. 03

    Treating the synthetic voice as an artificial voice under the TCPA, which is what the FCC ruled in 2024.

  2. 04

    Announcing an AI caller at first interaction — Maine's rule, and EU AI Act Article 50 in force since 2 August 2026.

  3. 05

    Announcing all-party consent to recording and transcription everywhere, with no state-conditional routing.

Decide

  1. 06

    Gating any driver-directed call on duty status, and never asking a driver who may be moving (49 CFR 392.80, 392.82).

  2. 07

    Handing the call to a person the moment one is asked for, and offering that without being asked.

Out

  1. 08

    Retaining the consent state, the disclosures delivered, the duty-status result and the transcript.

  2. 09

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

Product statement

The agent places the call and discloses; a named person takes anything handed over, and your firm holds the consent record.

Example workflow

One check call, consent to record

  1. 1Call reason raisedTMS milestone, visibility gap, appointment change or exception
  2. 2Consent and duty status gatheredThe consent record for the number, its type, the duty status of any driver and the load, each with its source
  3. 3Call preparedScript, disclosures, gate and confidence
  4. 4Controls appliedConsent-record checks, disclosure checks, the duty-status gate and confidence threshold

No human action required

Stages 1 to 4 run unaided and dial nothing — no consent record means no call, and the person's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Places the call with the disclosures.

Low confidence

Adds a desk read first.

A person takes it

The call is held with its consent state, the disclosure text and the confidence.

Place · Amend · Send to the desk

Placed — or handed to a person

  1. 6Load records updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedConsent states, disclosures delivered, duty-status stops and what the call actually established

Hand-offs

Every hand-off to a person is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Placing a call without a consent record for it.
    • Recording or transcribing before the disclosure.
    • Asking a driver anything while they are moving.
    • Deciding that a revocation need not be honoured.

  • Automation boundary

    Agent acts unaided

    • ✓Place the call once a consent record exists for the number for the named owner.
    • ✓Announce the business, the AI identity and the recording.
    • ✓Ask the status questions and write what the answer was.
    • ✓Hand the call to a person on request, and offer that first into the review queue.

    Any write happens inside the boundaries agreed at implementation, never ahead of approval.

    • Committing the load to a new appointment time.
    • Accepting or agreeing a detention or accessorial.
    • Deciding whether a carrier stays on this load.
  • Changes to the disclosure text or consent rules.

Example output

One check call, annotated

Everything the agent says is attached to the consent record behind the call.

Voice output · single call

Illustrative example
Called party
Owner-operator
What came back
Trailer loaded, and the appointment still stands
Call placed
09:12 local
Consent on record
Written consent, mobile number
Confidence
93%
Disclosures given
Business, callback, AI, recording

As received

Taken from the consent record and the load file — nothing on this side is assumed by the agent.

What was checked

  • Consent record for the number
  • Number treated as mobile
  • Duty status of the driver

Why it could be placed

It has a consent record, and the opening follows Maine's standard.

What the score decides

Below the configured threshold the call picks up a desk read before it is placed at all.

Action

PlaceAmendSend to the desk

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 check call

From the TMS or the exception queue

03Calling

Call on a consent record

Draw on the consent state for the number, the load record and the duty status of any driver on it.

01Approved path

Disclose it, then ask

Routine status and appointment calls come back already logged.

02Human review

Send the rest to a person

Missing consent, duty-status stops and low-confidence calls are marked, so the desk's read starts where risk concentrates.

04

Build an evidence trail

The call, the consent recorded for it and the person who took over stay on the load record.

Integrations

Typical integrations

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

  • TMS and load records

    • McLeod
    • MercuryGate
    • Trimble
    • Revenova

  • Visibility and tracking

    • project44
    • FourKites
    • Trucker Tools
    • Macropoint

  • Telephony and voice

    • Twilio
    • SIP trunking
    • Contact-centre platforms

Agent

Carrier voice check calls

  • Checks the consent
  • Places the call
  • Hands over on ask

  • Consent and archives

    • Consent registers
    • CRM
    • Recording archives

  • 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 call

The controls are nested. What survives all of them appears in the map underneath.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeHold calls for a person when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, disclosure-text and consent-rule configuration changes.Track
L4TraceabilityRecord the consent state, the disclosures delivered and the transcript.Record
L3Person on requestHand the call to a named person on request; it governs who speaks, not what was already said.Gate
L2Consent gatesTest each call against the consent record, the disclosure set and the duty-status gate; a failure stops it.Restrict
L1Confidence thresholdsRoute low-confidence calls to a desk read before they are placed at all.Require review
Model coreCall prepared — the script, the disclosures, the gate result and confidence
L1 – L2
Test whether a call may be placed
L3
Puts the line in a person's hands
L4 – L5
Show the call and the consent recorded for it
L6
Holds the call for a person when signals degrade

How Nestack evaluates it

Evaluate the whole call — not only the status it came back with.

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

Surface — the call the carrier hears

  1. E1Final-output evaluationWere the disclosures delivered before anything was asked?
  2. E2Step-level evaluationDid the agent use the right consent record, load and duty status?
  3. E3Tool evaluationDid it dial the correct number and write the correct load?
  4. E4Confidence calibrationDo low-confidence calls actually attract more hand-offs?
  5. E5Slice evaluationHow does performance change across specific call reasons?
  6. E6Business outcomeHow many calls were handed over, and how many should have been?

Floor — the person who answers the phone

Failure modes

Where each failure originates in the agent

Seven ways a call goes wrong, by the stage it starts.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • AN-03

      Stale consent record

      The number is dialled on a consent state since revoked.

    Stage gathers

    Consent state, number type, duty status and the load

  2. 02 · Reasoning

    2 modes

    • AN-04

      Disclosure out of order

      The reason for the call arrives before the disclosures do.

    • AN-06

      Mobile read as a desk line

      A business number is assumed not to be a cellular one.

    Stage proposes

    The script, the disclosures and confidence

  3. 03 · Tool / write

    2 modes

    • AN-02

      Call placed without a person

      A dial goes out before anyone has cleared the number.

    • AN-05

      Moving driver asked anyway

      The duty-status gate passes a driver who is still rolling.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • AN-01

      Recording before disclosure

      Capture starts before the all-party announcement lands.

    Stage returns

    The call the carrier hears and answers

  5. 05 · Change / Version

    1 mode

    • AN-07

      Silent disclosure regression

      A model or script change drops a disclosure from the opening.

    Stage tracks

    Model, prompt, disclosure text and consent rules

  • Sev-1 · no consent record
  • Sev-2 · recording before disclosure
  • Sev-3 · consent degrades, call routes to review
See the complete Transportation failure-mode catalogue

Affected slices

A total across call reasons is the wrong number to read

Break the log down by call reason before reading the total — first contact and in-transit calls carry most of the hand-offs, and a blended figure hides both. Nestack reports the hand-off rate by call reason, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
First contact with a new carrier3.9%3.5×Review
In-transit status on a moving load2.6%2.4×Review
Detention and appointment changes1.8%1.6×Watch
Scheduled pickup confirmations0.8%0.7×Normal

Bar: hand-off-rate lift vs. scheduled-pickup baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

If there is no case, the cycle is open

Explaining a bad call does not close the cycle. A standing regression case does, and that suite is what the next check call placed is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Hand-off rate rises in a call reason.

  2. 02

    Diagnose

    Every call is replayed against the consent state it was placed under, and one cause is kept.

  3. 03

    Improve

    The fix is versioned against the calls that produced it.

  4. 04

    Verify

    The release stops until every affected case is passing.

  5. 05

    Learn

    The case stays in the suite, and the calling rules are amended.

  6. Learn → Detect

    The return edge. What comes next is tested against a suite that grew.

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

Workstream

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

  1. 01Check-call workflow discovery and boundary definition.Week 1
  2. 02Telephony and visibility assessment.Week 1, Week 2
  3. 03Consent record, disclosure and duty-status rule mapping.Week 1, Week 2
  4. 04Consent-state and load ingestion.Week 2
  5. 05Call scripting and disclosure binding.Week 2, Week 3
  6. 06Confidence scoring and hand-off routing.Week 3
  7. 07Person-on-request workflow.Week 3, Week 4
  8. 08Telephony and TMS integration.Week 5
  9. 09Disclosure and recording cases.Week 4, Week 5
  10. 10Guardrails and consent controls.Week 4
  11. 11Call-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 lane set, one number poolProductionProduction telephony and TMSAdvancedMultiple entities / regions
Introduced at Pilot
Calling on your own consent records✓✓✓
Person on request✓✓✓
Call-quality baseline✓✓✓
Introduced at Production
Reporting by call reason—✓✓
Hand-off workflow in your systems—✓✓
Approved write-back—✓✓
Visibility-platform integration—✓✓
Introduced at Advanced
Multi-entity calling rules——✓
Multi-stage desk approvals——✓
High call volume——✓
Multi-state calling controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, transaction 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 consent records and calling policy

    Consent-state and load ingestion and disclosure mapping

    Week 1

  2. 02Representative past check calls

    Calling baseline, disclosure binding and the duty-status gate

    Week 2

  3. 03Your disclosure text and revocation route

    Consent record, disclosure and duty-status rule mapping

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    Telephony and visibility assessment, then integration setup

    Week 2

  5. 05Calls you would not want placed

    Consent cases and the evaluation suite

    Week 4

  6. 06What a call may never ask a moving driver

    Confidence scoring, hand-off routing, guardrails and consent controls

    Week 3

  7. 07Named people to take handed-over calls

    Person-on-request 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

Phases occupy the weeks the work really needs, so week 5 runs evaluation and launch side by side.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Check-call workflow discovery, consent mapping and the boundary

W2Telephony and visibility integration and the calling baseline

W3Calling workflow, confidence logic and hand-off controls

W4Evaluation suite, consent cases and failure-mode testing

W5TMS integration, a pilot lane set and targeted corrections

W6One week of check calls run under the desk, then handover

Reading the band
Each bar covers only the weeks its work is named in. The week 5 overlap is real, not padding.
At the end of W6
The week closes validation and Agent Care picks up the monitoring.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Transportation AI agent

Build a check-call agent that discloses before it asks anything.

Show us your numbers, your consent records and your call reasons. Name the person who signs off the disclosure text and owns the consent register — the calls go out in your name, and that is who we build around.

Typical delivery
4–6 weeks
Deployment
Your calling stack
Person on ask
Consent first
After launch
Agent Care available

Nestack Agents · Carrier voice check calls

AGT-TR-06 · Agent Care available after launch

Build a check-call agent

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

About you
Your calls

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