Main site

Transportation AI agent · Freight audit

Freight Audit, Pay & Dispute AI Agent

Audit invoices against the contract, the tariff and the documents, evidence each finding, and leave the dispute and any deduction to a named analyst who checks the contract first.

Typical delivery
4–6 weeks
Deployment
Your stack
Analyst signs
Evidence-first
After launch
Agent Care

What this agent does

Finds the discrepancy, never takes the money

In

  1. 01

    Ingest invoices, rate confirmations, contracts and accessorial schedules from supported sources.

  2. 02

    Normalise charge codes and units, and carry each charge forward with the document it was billed from.

Reason

  1. 03

    Compare the billed charge against the rate on the confirmation, the contract and the governing tariff.

  2. 04

    Separate the audit finding, which is analysis, from the dispute, which is an act with a deadline on it.

  3. 05

    Read an ocean demurrage or detention invoice against the elements the FMC billing rule requires, one at a time.

Decide

  1. 06

    Track the contest window, which runs from receipt of the bill and binds the shipper and the carrier alike.

  2. 07

    Route the finding, with the documents behind it, to the analyst who decides whether anything is paid or contested.

Out

  1. 08

    Retain the finding, the documents, the analyst's decision and the correspondence against the invoice.

  2. 09

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

Product statement

The agent finds and evidences; a named analyst decides whether money moves, and a deduction is checked against the contract before it is taken.

Example workflow

One invoice, receipt to decision

  1. 1Invoice receivedCarrier invoice, ocean demurrage bill, detention bill or accessorial charge
  2. 2Authority gatheredThe rate confirmation, the contract, the governing tariff and the documents the charge relies on
  3. 3Finding preparedCharge, authority, variance and confidence
  4. 4Controls appliedRate and tariff checks, invoice-element checks, contest-window checks and the confidence threshold

No human action required

Stages 1 to 4 run unaided and move no money — the finding is prepared, and the analyst's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the named analyst to decide.

Low confidence

Adds an audit-desk read first.

Analyst decides

The finding is held with its documents, its authority and the confidence.

Dispute · Accept · Send to the audit desk

Decided — disputed or accepted

  1. 6Payables records updatedOnly where write access and approval policy allow it
  2. 7Outcome evaluatedRecovery outcomes, carrier responses, reversed findings and contests prepared late

Overrides

Every analyst reversal is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Deducting or short-paying any part of an invoice.
    • Deciding who the proper billed party is here.
    • Releasing payment or approving an invoice to pay.
    • Issuing a written notice of disallowance to a carrier.

  • Automation boundary

    Agent acts unaided

    • ✓Compare billed charges against the contract and.
    • ✓Carry each charge forward with the document it was billed from.
    • ✓Check a demurrage invoice against its required elements.
    • ✓Flag the windows in play and hold the finding for the analyst.

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

    • Waiving, settling or writing off a disputed charge.
    • Reading a contract clause as advice on your rights.
    • Committing the company in writing to a carrier.
  • Changes to audit rules, thresholds or approval limits.

Example output

One charge line, annotated

Everything the agent finds is attached to the document it was read from.

Audit output · single invoice

Illustrative example
Invoice
Ocean demurrage
Charge line
Demurrage billed past the free time stated on the invoice
Billed amount
$2,340.00
Authority read
Service contract on file
Confidence
91%
Status
Finding raised, not deducted

As received

Taken from the carrier's invoice and the service contract on file — nothing on this side is inferred by the agent.

Authority used

  • Service contract rate
  • Free-time terms on the bill
  • Container availability record

Why it stops here

A finding is analysis, not a right to withhold and a person still decides.

What the score decides

Below the configured threshold the finding picks up an audit-desk read before it.

Action

DisputeAcceptSend to the audit 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 invoice

From the payables feed or the TMS

03Audit

Test the charge against its authority

Draw on the rate confirmation, the contract, the governing tariff and the documents on file.

01Approved path

Find it, and evidence it

Routine invoices come back matched, with the authority attached.

02Human review

Pass the rest to the desk

Variances, missing invoice elements and low-confidence findings are marked, so the analyst's read starts where money concentrates.

04

Build an evidence trail

The discrepancy, the documents behind it and the analyst who released it stay with the invoice.

Integrations

Typical integrations

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

  • Freight audit and pay

    • Cass
    • Trax
    • nVision Global
    • enVista

  • TMS and rating

    • McLeod
    • MercuryGate
    • Oracle OTM
    • e2open
    • Blue Yonder

  • Payables and ERP

    • SAP
    • Oracle
    • NetSuite
    • Coupa
    • Workday

Agent

Freight audit, pay and dispute

  • Reads the invoice
  • Evidences the finding
  • Holds for an analyst

  • Documents and EDI

    • EDI 210
    • rate confirmations
    • Document management

  • 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 payment

The layers nest. What each one misses is named in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeRevert the agent to reporting-only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, audit-rule and tariff-configuration changes.Track
L4TraceabilityRecord the invoice, the authority read, the finding, the documents and the decision.Record
L3Analyst approvalHold findings for a named analyst; it governs whether money moves, not whether the finding is right.Gate
L2Policy guardrailsTest findings against the configured contract, tariff and invoice-element rules; a failure returns the finding.Restrict
L1Confidence thresholdsRoute low-confidence findings to an audit-desk read before the analyst sees them.Require review
Model coreFinding produced — charge, authority, variance and confidence
L1 – L2
Test whether a finding may stand
L3
Puts the money in an analyst's hands
L4 – L5
Keep the discrepancy and the documents behind it
L6
Reverts to reporting-only when signals degrade

How Nestack evaluates it

Evaluate the audit workflow — not only the closing variance.

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

Surface — the finding the carrier sees

  1. E1Final-output evaluationDid the finding hold against the authority it was tested on?
  2. E2Step-level evaluationDid the agent read the right contract, tariff and accessorial schedule?
  3. E3Tool evaluationDid it read and write the correct invoice and the correct charge line?
  4. E4Confidence calibrationDo low-confidence findings actually attract more analyst reversals?
  5. E5Slice evaluationHow does performance change across specific charge types?
  6. E6Business outcomeHow many findings were reversed, or contested and then dropped?

Floor — the outcome the payables team answers for

Failure modes

Where each failure originates in the agent

Seven ways an audit goes wrong, at the stage it starts.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • BB-03

      Superseded rate document

      Charge tested against an expired contract or tariff.

    Stage gathers

    Invoice, rate confirmation, contract, tariff and documents

  2. 02 · Reasoning

    2 modes

    • BB-04

      Authority misread

      The wrong schedule is treated as governing the charge.

    • BB-06

      Finding read as entitlement

      Analysis is presented as a right to withhold money.

    Stage proposes

    Charge, authority, variance and confidence

  3. 03 · Tool / write

    2 modes

    • BB-02

      Contest window missed

      A contest is raised after the right to contest has run.

    • BB-05

      Duplicate finding

      One charge is raised twice across two invoices.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • BB-01

      Evidence not attached

      A finding leaves without the document behind it.

    Stage returns

    The finding the analyst reads and the carrier sees

  5. 05 · Change / Version

    1 mode

    • BB-07

      Silent rule regression

      A model or rule change widens what the agent will call a discrepancy.

    Stage tracks

    Model, prompt, audit rules and tariff config

  • Sev-1 · money moves outside the boundary
  • Sev-2 · a wrong finding reaches the carrier
  • Sev-3 · source degrades, finding routes to review
See the complete Transportation failure-mode catalogue

Affected slices

Overall recovery can hide one charge type

Ocean demurrage is the slice that hides. Aggregate recovery reads steady while that cohort absorbs most of the reversed findings. Nestack reports the reversal rate by charge type, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Ocean demurrage invoices8.3%3.7×Review
Accessorial and detention5.0%2.2×Review
Reweigh and reclass charges3.6%1.6×Watch
Contracted lane invoices2.0%0.9×Normal

Bar: reversed-finding-rate lift vs. contracted-lane baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

The loop closes on a test, not a credit

A cycle is done when the miss has become a test the next release has to survive. That suite is what the next invoice audited is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Reversed findings rise in one charge type.

  2. 02

    Diagnose

    The analyst who reversed it walks the invoice, the authority and the documents back until one cause is left.

  3. 03

    Improve

    Version-stamp the change and attach the invoices that exposed it.

  4. 04

    Verify

    The affected cases run again, and a fail stops the release.

  5. 05

    Learn

    It becomes a standing test, and the audit rules change with it.

  6. Learn → Detect

    The return edge. The next audit cycle 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, audit workflow, evaluation, integration, then production validation and handover.

Workstream

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

  1. 01Freight audit workflow discovery and boundary definition.Week 1
  2. 02TMS, payables and EDI source assessment.Week 1, Week 2
  3. 03Contract, rules-tariff and accessorial-schedule mapping.Week 1, Week 2
  4. 04Invoice ingestion and charge normalisation.Week 2
  5. 05Authority matching and variance logic.Week 2, Week 3
  6. 06Confidence scoring and finding routing.Week 3
  7. 07Analyst decision workflow.Week 3, Week 4
  8. 08Payables and TMS integration.Week 5
  9. 09Evidence and window cases.Week 4, Week 5
  10. 10Guardrails and payment controls.Week 4
  11. 11Invoice-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 carrier, one entityProductionProduction payables systemsAdvancedMultiple entities / modes
Introduced at Pilot
Auditing to your contracts and tariffs✓✓✓
Analyst approval✓✓✓
Audit-quality baseline✓✓✓
Introduced at Production
Reporting by charge type—✓✓
Decision workflow in your systems—✓✓
Approved write-back—✓✓
Payables-system integration—✓✓
Introduced at Advanced
Multi-carrier contract and tariff rules——✓
Multi-stage payables approvals——✓
High invoice volume——✓
Multi-entity audit 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 carrier contracts, rate confirmations and rules tariffs

    Contract, rules-tariff and accessorial-schedule mapping

    Week 1

  2. 02Representative paid and disputed invoices

    Audit baseline, charge normalisation and authority matching

    Week 2

  3. 03Your accessorial, fuel and demurrage schedules

    Audit-rule mapping and automation-boundary definition

    Week 1

  4. 04Access to relevant APIs, feeds or exports

    TMS, payables and EDI assessment, then integration setup

    Week 2

  5. 05Deductions you would not want taken

    Window cases and failure-mode testing

    Week 4

  6. 06What must reach an analyst before money moves

    Confidence scoring, finding routing, guardrails and approval controls

    Week 3

  7. 07Named analysts to decide on findings

    Analyst decision 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 are drawn over the weeks they actually occupy, which is why week 5 doubles up rather than padding.

Phase

W1

W2

W3

W4

W5

W6

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

Week focus

W1Freight audit workflow discovery, rule mapping and the automation boundary

W2Source integration and the audit baseline

W3Audit workflow, confidence logic and approval controls

W4Evaluation suite, window checks and failure-mode testing

W5Payables integration, pilot invoices and targeted corrections

W6One billing cycle audited under the payables team, then handover

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

Next step · Transportation AI agent

Build a freight audit agent around your payables chain.

Bring your carrier contracts, your rules tariffs and a month of paid invoices. We'll map the audit workflow, set the automation boundary and name what stays with the analyst.

Typical delivery
4–6 weeks
Deployment
Your TMS and payables
Analyst signs
Evidence-first
After launch
Agent Care available

Nestack Agents · Freight audit, pay and dispute

AGT-TR-13 · Agent Care available after launch

Build a freight audit agent

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

About you
Your invoices

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