Agent evaluation for logistics teams shipping AI
An agent assigning a route or preparing a customs filing is making a call with a real physical or legal consequence, and most teams cannot reconstruct why it made that call.
Every run scored for quality, performance and risk, and checked against the activity schema you define.
Every logistics agent run gets watched, evaluated for quality and risk, and checked against the rules you set. One record answers what any agent did, with evidence ready for review.
What's at stake in logistics
For logistics operations, compliance, and engineering teams, an agent's command moves real freight, so a bad route or a wrong customs code lands as a safety incident, a fine, or a delay before anyone reviews the run.
A route that breaks an hours-of-service limit is a safety incident, not a data error a team can quietly correct later.
Agents issue routes and declarations on shipment, driver, and inventory data that was already incomplete or out of date when they acted.
Few teams can show, for a given route or filing, exactly what data the agent used to reach that decision.
One source of truth for your logistics agents.
Instrument the agent frameworks you build on, and ingest the systems those agents touch as custom spans. Every run, score and signal lands in one place.
LangChain
Claude Agent SDKSDKCore SDK · TS + PythonOTLPOpenTelemetry · closed toolsObserve. Evaluate. Act.
Every route assignment and customs declaration is recorded as it happens, checked against the source record it claims to be based on, and held before the freight moves. Each step below closes one of the problems above.
- Observe. Checked against the driver's remaining hours-of-service window before dispatch
- Evaluate. Each command matched against the source record it claims to be based on
- Act. Written to an immutable audit trail, queryable per route and per filing
- DOT hours-of-service rules
- Customs regulations (CBP and equivalents)
- SOC 2
- GDPR (EU operations)
Built for everyone who owns logistics AI.
The same record, read the way each team needs it.
Instrument the logistics agents you already run — no gateway in the request path, no re-architecture.
For developers →Ship logistics features without regressions — every run scored before a customer ever sees it.
For product teams →One portfolio view of every logistics agent — its owner, cost and risk in a single place.
For heads of AI →Audit-ready evidence for every decision a logistics agent makes, ready when a regulator asks.
Security & governance →Two agents logistics teams are already shipping.
Hypothetical, but grounded in how logistics teams deploy agents today.
Route and dispatch agent
Low risk · 92/100A LangChain agent assigns delivery routes and drivers based on real-time traffic and driver hours data from a transportation management system, with a dispatcher approving assignments before they go out. Every assignment is scored through quality assessment → before it reaches the dispatcher.
- Each route assignment recorded as its own span, with the traffic and driver-hours data it read
- Validated against an activity schema that confirms the assignment respects the driver's remaining hours-of-service window
- Scored for quality and risk on every run
A compliance team can confirm every route assignment respected driver hours limits at the moment it was made, not reconstruct it after the fact.
Customs documentation agent
Medium risk · 87/100A Claude Agent SDK agent prepares customs declarations for cross-border shipments by pulling item classifications and values from the inventory system, drafting the filing for a customs broker to review. Each filing carries a risk score →, so an unusual classification pattern reaches the broker flagged rather than buried.
- Every classification the agent applies and every value it declares captured as a span
- Validated against an activity schema that checks the declared value matches the source inventory record
- A configurable risk profile flags a filing with an unusual classification pattern for mandatory broker review
A customs broker can verify every declared value traces back to the actual inventory record before a filing goes to a border agency.
Runs stay isolated — terminate one without touching the rest of your fleet via the kill switch → · Prefactor vs. observability tools →
The problems this solves
Agents fail quietly and the first signal is a complaint, not an alert.
How it gets caught →Stuck at POCThe pilot worked; sign-off takes months because risk has no evidence.
How it gets caught →No kill switchWhen an agent misbehaves, nothing can stop it short of stopping everything.
How it gets caught →See agent evaluation on your own logistics agents
Book a demo and we'll walk through span-level scoring and audit evidence on a fleet like yours.
Unified performance platform for agents, authentication, and risk management