Rate-Negotiation AI Agent

What this agent does

Counters inside limits, and stops at acceptance

Your own rate history, lane costs and published market indices, from supported TMS, rating or contract sources.

Per-tenant isolation that is architectural and auditable, rather than a filter applied to what a screen shows.

Reason

Non-public competitor rate data never pooled into a recommendation at runtime — the line the DOJ decree draws.

Cross-tenant training data, where any is used, kept at least twelve months old, non-price and non-hyperlocal.

Authority limits you set — floor, ceiling, lane, volume and term — bounding every counter the agent drafts.

Decide

Symmetric governors, with no rate floor and no control that resists a decrease more than an increase.

Auto-accept off by default, so acceptance waits on the named person who holds the authority to give it.

Out

A full transcript retained: every offer, every counter and the authority each one was made under.

Write actions only inside the approval boundaries agreed during implementation.

→ Product statement

The agent counters; a named person accepts, and under UETA §14 whatever it sends is your firm's own word.

Example workflow

One negotiation, offer to acceptance

AgentHuman

1Rate request receivedCarrier bid, shipper request, tender response or lane review

2Position assembledYour own rate history, the lane cost, published indices and the authority limits in force, each with its source

3Counter preparedOffer, limits, data sources and confidence

4Controls appliedRuntime data-source checks, authority-limit checks, symmetry checks and the confidence threshold

No human action required

Stages 1 to 4 run unaided and accept nothing — the counter is drafted, and the person's lane opens at the confidence gate.

5DecisionBranches at the confidence threshold

High confidence

Goes to the named person to accept.

Low confidence

Adds a desk read first.

A person accepts

The offer is held with its limits, its data sources and the confidence.

Accept · Counter · Send to the desk

Accepted — confirmation released▼

6Rate records updatedOnly where write access and approval policy allow it

7Outcome evaluatedLimit breaches stopped, counters drafted, acceptance rates and what each negotiation settled at

Overrides

Every override of a limit is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required8 items

Accepting a rate or returning a rate confirmation.

Using non-public competitor rate data at runtime.

Setting a rate floor or an asymmetric governor.

Turning auto-accept on, or widening what it covers.

Automation boundaryAgent acts unaided

✓Assemble the position from your own rate history and indices.

✓Draft a counter that sits inside the limits you have set.

✓Show the data each figure came from, and how old that data is.

✓Hold the offer for the named person who holds the authority.

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

Agreeing accessorial, detention or fuel terms.

Committing volume, lane or term beyond your limits.

Deciding that a limit breach may be waived this once.

Changes to the authority limits or the data rules.

Example output

One counter, annotated

Everything the agent offers is attached to the limits it was working inside.

Negotiation output · single laneIllustrative example

Lane

Counter drafted

Offer made

Data behind it

Confidence

Authority applied

Dry van, regional

Held at the mid-point of the range, and expires on the stated date

$2,140.00

Own rate history, published index

91%

Ceiling, lane and term limits

As receivedTaken from your own rate file and a published index — nothing on this side came from another broker.

What it was built from Your own lane rate history Published market indices The authority limits in force

Why it could be offeredIt sits inside the ceiling you set and a person still decides and it is logged..

ActionAcceptCounterSend to the desk

What the score decidesBelow the configured threshold the offer picks up a desk read before it is put to anyone at all.

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value landsValue 01 – 04

Every negotiationFrom the TMS or the bid queue

03Countering

Counter inside your limits

Draw on your own rate history, your lane costs and the published indices you subscribe to.

01Approved path

It can bind you — so it holds

Routine lane counters come back drafted and inside the range.

02Human review

Send the rest to a person

Limit breaches, thin rate history and low-confidence offers are marked, so the desk's read starts where risk concentrates.

04Build an evidence trail

The offer, the limits it was working inside and the person who accepted it stay on the confirmation.

Integrations

Typical integrations

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

TMS and rate recordsMcLeod · MercuryGate
Trimble · Revenova

Rating and market dataDAT · Greenscreens
Published market indices

Contracts and tendersMaster agreements · RFP tools
Rate confirmations

Agent

Rate negotiation

Reads your own data
Drafts the counter
Holds for acceptance

Finance and settlementSettlement · accounting
Audit and pay platforms

Observability & evaluationOpenTelemetry · 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 offer

Each layer contains the next. What gets past the set is listed in the map below.

L6 · Outermost — last line of defenceInward → L1 · closest to the model

L6Rollback / safe modeReduce the agent to rate lookup when evaluation or production signals degrade.Roll back

L5Version monitoringTrack model, prompt, authority-limit and data-rule configuration changes.Track

L4TraceabilityRecord every offer, the limits in force and the data each figure came from.Record

L3Acceptance heldHold the offer for the named person; it governs who accepts, not whether the rate was a good one.Gate

L2Authority limitsTest each counter against the floor, ceiling, lane and term you set; a breach returns the offer.Restrict

L1Confidence thresholdsRoute low-confidence offers to a desk read before they are put to anyone.Require review

Model coreCounter prepared — the offer, the limits, the data sources and confidence

L1 – L2Test whether an offer may stand

L3Puts acceptance in a person's hands

L4 – L5Keep the offer and the limits behind it

L6Reduces to rate lookup when signals degrade

How Nestack evaluates it

Evaluate the whole negotiation — not only the rate it settled at.

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

Surface — the offer the counterparty reads

Depth of coverage ▼

E1Final-output evaluationDid each counter sit inside the limits in force at the time?

E2Step-level evaluationDid the agent use your own data, and data of the right vintage?

E3Tool evaluationDid it read the correct lane and write the correct record?

E4Confidence calibrationDo low-confidence offers actually attract more desk reads?

E5Slice evaluationHow does performance change across specific lane types?

E6Business outcomeHow many offers were overridden, and how many should have been?

Floor — the confirmation your firm is bound by

Failure modes

Where each failure originates in the agent

Seven ways an offer goes wrong, placed at its stage.

Agent lifecycleDirection of processing →

01 · Retrieval1 mode

AO-03

Stale rate basis

The counter is built on an index reading since moved.

Stage gathersRate history, lane cost, indices and the limits

02 · Reasoning2 modes

AO-04

Ceiling read as a target

A bound is treated as the number to aim at.

AO-06

Asymmetric movement

The agent concedes upward more readily than downward.

Stage proposesThe offer, the limits applied and confidence

03 · Tool / write2 modes

AO-02

Offer sent without a person

A counter goes out before anyone has cleared it.

AO-05

Duplicate counter

The same lane is countered twice on one thread.

Stage writesOnly where write access and approval policy allow it

04 · Output1 mode

AO-01

Cross-tenant rate leak

Another tenant's non-public rate reaches a recommendation.

Stage returnsThe offer the counterparty reads and may accept

05 · Change / Version1 mode

AO-07

Silent limit regression

A model or config change widens what the agent will offer.

Stage tracksModel, prompt, authority limits and data rules

Sev-1 · accepted outside limits Sev-2 · outside data reaches an offer Sev-3 · basis degrades, offer routes to review

See the complete Transportation failure-mode catalogue →

Affected slices

The aggregate is not the lane you are losing on

An aggregate override rate can sit comfortably inside tolerance while two lane types absorb almost all of it. Nestack reports the override rate by lane type as well as in total, so the slice stays visible instead of being averaged away.

Slice performance — reported separately, not only in aggregateIllustrative example

Slice	Failure rate	Lift	Status
Spot lanes with thin rate history	7.2%	3.8×	Review
Contract renewals at term	5.1%	2.7×	Review
Accessorial-heavy moves	2.7%	1.4×	Watch
Committed contract lanes	1.5%	0.8×	Normal

Bar: override-rate lift vs. committed-lane baseline · scale 0–4.0× · tick marks the 2.0× review threshold 2 of 4 slices over threshold

Evidence-linked improvement

The artefact is a test, not a note

Nothing closes because it was understood. It closes when the next release has to pass a case, and that suite is what the next offer made is measured against.

Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn

01Detect

Override rate rises in a lane type.

02Diagnose

Offers go back on the desk beside the authority in force at the moment each was made.

03Improve

Changes ship against a version with the negotiations that caused them.

04Verify

A failing case blocks release until it clears.

05Learn

The case is permanent, and the authority limits move with it.

Learn → DetectThe return edge. The next detection runs 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, negotiation workflow, evaluation, integration, then production validation and handover.

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

01Negotiation workflow discovery and boundary definition.

02TMS and rate-source assessment.

03Authority-limit, data-isolation and governor mapping.

04Rate-history and index ingestion.

05Counter logic and limit binding.

06Confidence scoring and desk routing.

07Acceptance workflow.

08TMS and rating integration.

09Formation and authority cases.

10Guardrails and acceptance controls.

11Offer-trail instrumentation.

12Deployment, documentation and Agent Care handover.

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 tier PilotOne lane set, one entity ProductionProduction TMS and rating AdvancedMultiple entities / regions

Introduced at Pilot

Countering on your own data✓✓✓

Acceptance held for a person✓✓✓

Offer-quality baseline✓✓✓

Introduced at Production

Reporting by lane and partner—✓✓

Acceptance workflow in your systems—✓✓

Approved write-back—✓✓

TMS and rating integration—✓✓

Introduced at Advanced

Multi-entity authority rules——✓

Multi-stage desk approvals——✓

High negotiation volume——✓

Multi-entity authority controls——✓

Build price From $5,000 From $8,000 Custom quote

Final build priceConfirmed after discovery based on integrations, workflow complexity, transaction volume, approval controls and deployment requirements.

Separate from buildBuild 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.

You bringWe build with it

01Your authority limits and rate policy → Rate-history and index ingestion and limit mappingWeek 1

02Representative past negotiations → Counter baseline, limit binding and data-isolation setupWeek 2

03Your data sources and their vintage → Authority-limit, data-isolation and governor mappingWeek 1

04Access to relevant APIs, feeds or exports → TMS and rate-source assessment, then integration setupWeek 2

05Offers you would not want accepted → Authority cases and failure-mode testingWeek 4

06What the agent is never given authority to accept → Confidence scoring, desk routing, guardrails and acceptance controlsWeek 3

07Named people who hold the authority to accept → Acceptance workflow, then pilot and production validationWeeks 5–6

Nothing else is required Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

Each band covers the weeks it genuinely occupies, so the fifth week holds two kinds of work.

Phase W1W2W3W4W5W6

Discovery W1

Build W2 – W3

Evaluate W4 – W5

Pilot & Launch W5 – W6

Week focus W1Negotiation workflow discovery, limit mapping and the boundary W2Rate-source integration and the counter baseline W3Negotiation workflow, confidence logic and acceptance controls W4Evaluation suite, authority cases and failure-mode testing W5TMS integration, a pilot lane set and targeted corrections W6One negotiation cycle run under the desk, then Agent Care handover

Reading the bandEach band sits on the weeks it truly needs, so the fifth holds evaluation and launch at the same time.

At the end of W6Live negotiations close the validation and monitoring moves to Agent Care.

DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Transportation AI agent

Build a rate-negotiation agent that stops short of acceptance.

Show us your lanes, your rate sources and the limits you would set. Being wrong here is expensive — the counter your agent sent is the rate your firm is held to, and UETA forms that contract precisely because nobody read it.

Discuss this agent →

Nestack Agents · Rate negotiationAGT-TR-07 · Agent Care available after launch