Agriculture AI agent · Freight & logistics
Ag-Freight & Logistics AI Agent
Plan loads against harvest and delivery windows, book from an approved carrier and rate list, track in transit and reconcile the invoice — with welfare, food-safety and hours rules held by a person.
- Typical delivery
- 4–6 weeks
- Deployment
- Your stack
- Tender and book
- Agent, stoppable
- After launch
- Agent Care
What this agent does
Plans and books, never approves a carrier
In
- 01
Take the loads to move — commodity, quantity, origin and destination, and the window each one has to hit.
- 02
Read the approved carrier list, the contracted rate table, the equipment available and the appointment calendar.
Reason
- 03
Match equipment to the commodity — hopper, reefer, live-haul, flatbed — and to what the loading site allows.
- 04
Plan the run against hours rules, the agricultural exemption's radius and season, permits, weights and route restrictions.
- 05
Apply the commodity's temperature spec, pre-cooling and record path, or the journey plan a live-haul needs.
Decide
- 06
Hold any load whose plan leans on an exemption, permit or welfare limit that does not clearly apply on that leg.
- 07
Hold any carrier that is not on the approved list, or whose authority, insurance or identity does not verify.
Out
- 08
Tender to an approved carrier at a contracted rate, book the appointment and track the load against it.
- 09
Reconcile the freight invoice to the rate and the accessorials, and put every difference in front of a person.
Product statement
The agent plans, books, tracks and reconciles. Carrier approval, any exception to a welfare or safety rule, and every payment stay with a person.
Example workflow
One load, end to end
- 1Load requestedThe commodity, the quantity, the origin and destination, and the window it has to hit
- 2Plan builtEquipment matched to the commodity, then hours and exemption radius, permits, weights and route restrictions
- 3Requirements attachedThe temperature spec, pre-cooling and the record path, or the live-haul journey plan, density and rest
- 4Carriers checkedApproved list, contracted rate, and authority and insurance verified at source rather than from the paperwork sent
No human action required
Stages 1 to 4 run before anyone is asked to look — planning, load requirements and carrier checks finish first. A carrier that will not verify, or a plan leaning on an exemption, ends the stretch there and then.
Tenders at the contracted rate.
Goes to the logistics lead first.
Logistics lead
Reads the plan, the carrier's authority and insurance, the exemption or permit in question and the rate, then approves the carrier or the exception.
Approve carrier · Amend the plan · Reject
Approved — handed back
- 6Tendered, booked and trackedTendered at the contracted rate, the appointment booked and the load tracked against the plan
- 7Outcome evaluatedAppointment adherence, temperature-record completeness, rule conflicts caught and invoice differences
Exceptions
Every carrier approval and every rule exception is recorded against the load.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required
8 items
- Approving a carrier or adding one to the list.
- Deciding an hours exemption applies to a leg.
- Signing off a permit, route or weight exception.
- Accepting a livestock journey outside the welfare limits.
Automation boundary
Agent acts unaided
- ✓Plan loads against harvest, delivery and appointment windows.
- ✓Match equipment to the commodity and check hours, weights and the route.
- ✓Tender to carriers already approved, at rates already contracted.
- ✓Track the load, collect the records and reconcile the invoice.
Write actions stay inside the boundaries agreed in implementation: approved carriers, contracted rates, and a tender the logistics lead can pull before pickup.
- Judging a temperature excursion acceptable.
- Releasing payment on a freight invoice or accessorial.
- Agreeing a detention, demurrage or damage claim.
- Changing rate tables, carrier lists or booking rules.
Example output
One load plan, annotated
Everything the agent proposes is attached to the load, the rate line and the checks it came from.
Freight output · single load
Illustrative example- Load
- Chilled produce, 20 pallets
- Lane
- Packhouse to a retail DC
- Window
- Tue 06:00 slot
- Tendered to
- Approved carrier — verified
- Confidence
- 92%
- Temperature
- Spec set, records due
As received
The load, the lane and the window as the order and the appointment calendar hold them.
Evidence used
- Authority and insurance
- Contracted rate line
- Pre-cool and spec
Why this carrier
It is on your approved list, and its authority and insurance were read at source.
What the score decides
Confidence decides how much of the plan a person re-reads, not whether a rule applies.
Action
Approve carrierAmend the planRejectValue
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 load to move
From the order, the harvest plan and the schedule
Apply the rules that govern the run
Use the approved carrier list, the contracted rates, the hours, permit and weight rules, the temperature spec and the welfare limits for that journey.
Settle the plan while it can still change
Equipment, hours, permits and the appointment are worked out before the truck is loaded, not while it is standing at a gate with a crop on it.
Put a person on the loads that carry risk
A carrier that will not verify, an exemption that does not fit the leg, a journey that breaks a welfare limit — those reach the logistics lead instead of a carrier.
Build an evidence trail
Retain the plan, the carrier check, the permit and exemption reasoning, the temperature record, the appointment and the invoice match — on both paths.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Freight systems
- Transport management
- load boards
- Rate tables
- tendering APIs
Carrier data
- Authority records
- insurance certificates
- Vetting services
- approved lists
Farm & grain systems
- Operations Center
- FMS
- Scale and grain tickets
- packhouse orders
Agent
Ag-freight & logistics
- Plans the load
- Books and tracks
- Reconciles the invoice
Tracking & temperature
- Telematics
- ELD feeds
- Reefer records
- appointment portals
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 truck
Each control wraps the one inside it. A load clears every layer before it is tendered, and none of the six can waive a welfare or food-safety duty.
L6 · Outermost — last line of defence
Inward → L1 · closest to the model
- L1 – L2
- Decide whether a load may be tendered at all
- L3
- Decides who may carry it, and at what rate
- L4 – L5
- Keep approval with a person and the record intact
- L6
- Pulls automation back when signals degrade
How Nestack evaluates it
Evaluate the whole movement — not only the load that arrived.
Coverage runs the whole depth of the workflow, and every layer is cut by slice.
Surface — the plan the freight desk reads
- E1Final-output evaluationWas the plan workable, and did the load move the way it said?
- E2Step-level evaluationDid it match equipment, hours, permits and the temperature spec?
- E3Tool evaluationDid it read the right rate line, carrier record and appointment?
- E4Carrier-verification coverageWas authority, insurance and identity checked before every tender?
- E5Slice evaluationHow does performance change by lane, commodity and season?
- E6Business outcomeHow many loads missed a window, a record or a rate?
Floor — whether the load arrived fit and on its window
Failure modes
Where each failure originates in the agent
Seven failure modes plotted against the five stages of the agent lifecycle.
Agent lifecycle
Direction of processing →
01 · Planning / intake
2 modes
- AF-01
Live-haul limits read stale
Journey time, density or rest from an old record.
- AF-02
Appointment moved late
The delivery slot changed after the plan.
Stage gathers
Loads, windows, rates, carriers and the rules in force
02 · Plan build
2 modes
- AF-03
Exemption misapplied
Hours planned on an exemption that does not fit.
- AF-04
Permit or weight missed
An axle, seasonal or route limit is not in the plan.
Stage matches
Equipment, route, hours, permits and requirements
03 · Tender / book
1 mode
- AF-05
Carrier not verified
A tender reaches a fraudulent carrier.
Stage books
Approved carriers only, at rates already contracted
04 · Track / settle
1 mode
- AF-06
Temperature record missing
A reefer load's condition cannot be proved.
Stage returns
Status, records and the invoice put up for payment
05 · Change / Version
1 mode
- AF-07
Silent rate-table change
A rate or accessorial change moves the match.
Stage tracks
Model, prompt, rate-table and carrier-list changes
- Sev-1 · a safety or welfare rule is breached
- Sev-2 · the load or the payment goes wrong
- Sev-3 · planning degrades, more falls to people
Affected slices
Most of the failures sit in a few weeks of the year
A load-exception rate that reads tolerably across a season sits mostly in harvest weeks, in temperature-controlled freight and on carriers nobody has moved with before. Nestack reports performance by slice, not only in total.
Slice performance — reported separately, not only in aggregate
Illustrative example| Slice | Failure rate | Lift | Lift vs. threshold | Status |
|---|---|---|---|---|
| Harvest-peak spot loads | 7.0% | 3.8× | Review | |
| New and one-off carriers | 5.1% | 2.8× | Review | |
| Reefer and temperature-controlled | 3.5% | 1.9× | Watch | |
| Contracted lanes, repeat carrier | 1.5% | 0.8× | Normal |
Bar: load-exception rate lift vs. contracted-lane baseline · scale 0–4.0× · tick marks the 2.0× threshold
2 of 4 slices over threshold
Evidence-linked improvement
A load already moving is fixed on the road, not in the loop
A truck under load cannot wait for a root cause. It is re-iced, re-routed or re-booked first — the loop only decides what next week's planning is built on.
Improvement cycle · five stages
Switchback — the path turns at Improve and returns at Learn
- 01
Detect
A lane is the unit: exceptions, missed windows and invoice differences counted against it.
- 02
Diagnose
The rate table, the carrier check, the rule set and the tracking feed are read apart until one owns it.
- 03
Improve
Nothing ships until the logistics lead signs the corrected rule, list or rate line, version stamped.
- 04
Verify
Recorded loads from that lane run through again, the ones that failed among them.
- 05
Learn
The load joins the regression set, and the booking runbook is rewritten before the next season.
Learn → Detect
The return edge. A welfare limit, a temperature spec or a carrier list is not the agent's to change — a person signs it before a load runs on it.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, rates and rules, planning and tendering, evaluation, integration, then production validation and handover.
Workstream
Week 1Week 2Week 3Week 4Week 5Week 6
- 01Workflow discovery and automation-boundary definition.Week 1
- 02Freight-system and carrier-list assessment.Week 1, Week 2
- 03Hours, permit and welfare rule mapping.Week 1, Week 2
- 04Commodity-to-equipment matching rules.Week 2
- 05Order, load and appointment ingestion.Week 2, Week 3
- 06Load planning and delivery-window logic.Week 3
- 07Carrier verification and tendering workflow.Week 3, Week 4
- 08Temperature spec and live-haul rules.Week 4
- 09Rule-conflict and carrier-verification evaluation.Week 4, Week 5
- 10Invoice and accessorial reconciliation.Week 5
- 11Tracking, appointment and record integration.Week 5, Week 6
- 12Observability, deployment 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 tier | PilotOne lane, one commodity | ProductionProduction freight integration | AdvancedMulti-site / multi-commodity |
|---|---|---|---|
| Introduced at Pilot | |||
| Load planning against windows | ✓ | ✓ | ✓ |
| Tendering to approved carriers only | ✓ | ✓ | ✓ |
| Carrier approval left with a person | ✓ | ✓ | ✓ |
| Baseline evaluation | ✓ | ✓ | ✓ |
| Introduced at Production | |||
| Carrier authority and insurance checks | — | ✓ | ✓ |
| Rule-conflict, permit and weight checks | — | ✓ | ✓ |
| In-transit tracking and appointments | — | ✓ | ✓ |
| Invoice and accessorial reconciliation | — | ✓ | ✓ |
| Observability and evaluation | — | ✓ | ✓ |
| Temperature and live-haul requirements | — | ✓ | ✓ |
| Introduced at Advanced | |||
| Multi-site and enterprise controls | — | — | ✓ |
| Build price | From $5,000 | From $8,000 | Custom quote |
- Final build price
- Confirmed after discovery based on lanes and commodities, freight and carrier-data integrations, load volume, rate and accessorial complexity, 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.
You bring
We build with it
01Your approved carrier list and contracted rate tables
→Freight-system, rate-table and carrier-list assessment
Week 102The rules you work to — hours, permits, weights and welfare
→Hours, permit, weight and welfare rule mapping
Week 103What each commodity needs from equipment and a site
→Commodity-to-equipment and site-constraint rules
Week 204Access to order, appointment and tracking feeds
→Order, load and appointment ingestion, then tracking
Week 205Loads that went wrong, including the near misses
→Evaluation suite, regression cases and failure-mode testing
Week 406A year of freight invoices with the disputes attached
→Invoice, accessorial and detention reconciliation
Week 507A named logistics lead to approve carriers and payments
→Carrier-approval workflow, then pilot loads and 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. Week 5 carries both the invoice reconciliation and the first supervised loads.
Phase
W1
W2
W3
W4
W5
W6
- DiscoveryW1
- BuildW2 – W3
- EvaluateW4 – W5
- Pilot & LaunchW5 – W6
Week focus
W1Workflow discovery, the rule set and the automation boundary
W2Rate tables, carrier lists, and order and appointment feeds
W3Load planning, window logic and the tendering workflow
W4Rule-conflict, permit and carrier-verification testing
W5Tracking and invoice integration, then supervised loads
W6Your desk books from live plans, then Agent Care starts
- Reading the band
- Carrier verification and rule-conflict checks are tested in week 4, before a load is tendered on them in week 5.
- At the end of W6
- Plans have been booked alongside your existing desk and checked against your own invoices, then Agent Care takes over monitoring.
- Duration
- Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Agriculture AI agent
Build a freight agent around your lanes and your carrier list.
Show us a season of loads, your rate tables and how a carrier gets approved today. We'll replay those loads and show you which plans broke a rule, which carriers would not verify and which invoices you paid twice.
- Typical delivery
- 4–6 weeks
- Deployment
- Your freight and farm stack
- Tender and book
- Approved carriers, lead can pull
- After launch
- Agent Care available
Nestack Agents · Ag-freight & logistics
AGT-AGR-10 · Agent Care available after launch
