Loads do not fail cleanly.
They fail through late pings, ignored emails, missed FCFS windows, bad appointment data, lumper surprises, and carrier replies buried in inbox threads.
Humans
The Escalation Layer
Human-owned work
- High-value shipper updates, recovery plans, and service failure messaging
- Carrier negotiation, fall-off recovery, accessorial disputes, and detention approvals
- Priority calls when hot freight, premium customers, and margin collide
- Customer-specific judgment around scorecards, routing guides, and service commitments
xGold Design Rule
Reps should handle judgment, negotiation, and trust moments. They should not spend the day refreshing FourKites, checking DAT, copying emails into McLeod, or chasing routine check-call updates.
AI Agents
The Freight Interpretation Layer
Agent-owned work
- Interpreting carrier emails, SMS replies, ELD visibility gaps, and customer portal notes
- Summarizing load history across TMS notes, tracking events, CRM activity, and email threads
- Drafting shipper updates with ETA, cause, recovery action, and confidence level
- Recommending whether to re-cover, escalate, wait, or notify the customer
Common tools
McLeod, MercuryGate, Aljex, Rose Rocket, Tai TMS, Turvo, project44, FourKites, MacroPoint, DAT, Truckstop, Highway, MyCarrierPackets, RMIS, Salesforce, Outlook, Teams.
Scripts
The Freight Execution Layer
Script-owned work
- Stale tracking thresholds, geofence misses, tender age, and appointment-risk triggers
- Load status writebacks, task creation, queue assignment, and customer-update cadence
- BOL, POD, rate confirmation, lumper, detention, and accessorial document validation
- Carrier onboarding checks, insurance expiration flags, and lane-specific rule enforcement
xGold Design Rule
If the trigger and next action are known, use scripts. AI should not be paid to do threshold checks, task creation, or status writebacks a deterministic workflow can run all day.
Data & Systems
The Load Record Layer
System-owned truth
- Load status, tender status, pickup/delivery appointment, carrier, lane, and shipper records
- Visibility events, check-call history, EDI updates, and customer portal commitments
- CRM notes, routing-guide promises, escalation history, and account-specific rules
- Billing, POD, accessorial, detention, lumper, and claims evidence
xGold Design Rule
The load record wins. Automation must update the TMS and CRM trail instead of creating another offline tracker that operations stops trusting by Friday.
Exception Intake & Triage
- HOps defines exception classes: late pickup, missed delivery, no tracking, detention, rejected tender, OS&D, rate dispute
- AAgent reads carrier email, EDI 214 alerts, visibility events, customer portal comments, and TMS notes
- SScript classifies severity by SLA, appointment window, customer tier, lane, margin, and owner
- DTMS, CRM, visibility, and customer-facing notes update with status and next action
Carrier Follow-Up Loop
- STrigger fires when tracking is stale, geofence confidence drops, or appointment risk increases
- AAgent drafts carrier-specific follow-up using load number, lane, facility, hours, and prior replies
- DResponses are parsed from email, SMS, portal notes, ELD/visibility tools, or dispatcher updates
- HCoordinator handles disputes, re-cover decisions, customer-sensitive exceptions, or unresolved risk
Customer Status Updates
- DLoad status, ETA, facility appointment, dwell, and exception context are pulled from source systems
- AAgent summarizes what changed, operational impact, recovery path, and confidence level
- SUpdate is posted to CRM, email, customer portal, Teams channel, or shared exception queue
- HAccount owner approves updates for strategic accounts, service failures, or scorecard-sensitive freight
Rate & Document Cleanup
- AAgent extracts shipment details from rate confirmations, BOLs, PODs, lumper receipts, and detention forms
- SScript validates shipper, consignee, PO, weight, accessorial, and appointment fields against TMS records
- DMissing data, accessorials, OS&D notes, and exceptions are written back for billing or claims review
- HBilling, claims, or ops resolves mismatches above customer or dollar threshold
| Freight Task | Best Handled By | Why | Typical Integration |
|---|---|---|---|
| Stale tracking, geofence miss, or no-ping alert | Script | Threshold logic should run continuously across every active load | project44/FourKites/MacroPoint → TMS task |
| Carrier response interpretation | AI Agent | Dispatcher replies are short, incomplete, and context-dependent | Email/SMS/portal → TMS note |
| Strategic shipper escalation | Human | Scorecard, margin, and relationship risk require judgment and tone | CRM alert → account owner |
| BOL/POD/accessorial extraction | AI Agent + Script | Documents need extraction, then deterministic validation against the load | Document inbox → TMS/billing queue |
| Tender fallout and re-cover routing | AI Agent + Human | Automation assembles lane and carrier context; humans decide margin and service tradeoff | TMS/load board → carrier sales queue |
| Load status writeback | Data & Systems | The load record must remain the shared operating truth | TMS, CRM, portal, EDI |
The TMS is the operating spine.
McLeod, MercuryGate, Aljex, Rose Rocket, Tai, or Turvo must stay authoritative. Automation should enrich the load record, not bypass it.
Visibility events are dispatch triggers.
project44, FourKites, and MacroPoint signals should create routed work when tracking goes stale, appointments move, or ETA risk changes.
Humans handle relationship and margin risk.
Escalate shipper-sensitive, high-margin, disputed, hot, or scorecard-sensitive loads. Automate the routine chase work around them.
Every exception needs owner, SLA, and next action.
A detected issue without a queue, service clock, and action path is just another alert competing with the inbox.
Measure touches per load.
The fastest ROI proof is fewer check calls, portal lookups, copied notes, EDI corrections, and repeated customer updates.
Documents close revenue leakage.
BOLs, PODs, rate confirmations, detention, lumper, and accessorial evidence should land in the record before billing, claims, or dispute review.
Start here
Book a working conversation
Bring one exception queue and one week of load history.
We will map the trigger, system owner, manual touchpoints, and first automation target.