Logistics Agentic AI · Freight Forwarding

Freight quotes, from inbox to booking. On autopilot.

The Freight Quotation Agent is a Stoqr module that reads a client’s email, breaks down the shipment into its service legs, sources rates from your partners, and returns a ranked, margin-applied quote — with your team in control of every commitment.

Start a pilot See how it works// A Stoqr module · pilot stage
One loop · email to booking
Air · Ocean · RoadRFQ Quote BookingHuman-in-the-loopSG · Asia
The cost of the inbox

A freight quote still means a dozen emails and a day of waiting.

Forwarders coordinate international shipments over free-form email — reading requests, breaking them into legs, chasing partner rates, and re-keying replies. It is slow, labour-intensive, and inconsistent, and a late quote often costs the business.

ISSUE 01

Latency

Breaking down a shipment, contacting several partners per leg, and chasing replies takes hours to days. The forwarder who answers fastest with a competitive price usually wins the business.

ISSUE 02

Manual effort

Skilled staff spend a large share of the day reading, copying, and re-emailing rates — not on the judgement and relationship work that actually adds value.

ISSUE 03

Inconsistency

Which legs get priced, which partners get asked, and what margin gets applied all depend on who is at the desk. Quotes get mis-transcribed; options get missed.

ISSUE 04

No memory

Rates, partner performance, and negotiation history sit scattered across mailboxes, so the business cannot easily learn from its own data.

The quotation loop

One inbound email, run end to end.

A client email arrives and the Agent takes it through the whole loop — intake and breakdown, sourcing and evaluation, then a ranked quote, booking, and bounded negotiation. Every path ends in a logged, closed state.

Stage 01 · Intake

Read & structure

The Agent watches the inbox, classifies each message's intent, and turns free text into a structured shipment — then breaks it into its service legs.

  • Classify intent · RFQ, selection, negotiation
  • Extract shipment · origin, mode, commodity
  • Breakdown into ordered service legs
Stage 02 · Sourcing

Source the rates

For every leg, the Agent selects suitable partners from your registry, dispatches RFQs, and normalises the heterogeneous replies into comparable end-to-end options.

  • Match partners · capability · lane · mode
  • Dispatch RFQs · track outstanding
  • Evaluate & consolidate · normalise quotes
Stage 03 · Quote & book

Quote & commit

Margin is applied, options are ranked, and a clear quote goes to the client behind an approval gate. On selection it books; on a counter-offer it negotiates within guardrails.

  • Apply margin · rank options
  • Send quotation · approval gate
  • Confirm booking · or bounded negotiation
The multi-agent team

An orchestrator and nine specialists.

Rather than one monolithic model, responsibility is split across a coordinator and nine worker agents, each with a narrow, testable remit — so behaviour stays predictable and every step is auditable.

ORCHESTRATOR

Coordinator

Maintains workflow state as an explicit state machine, routes each message to the right specialist, enforces approval checkpoints, and guarantees every shipment moves through a well-defined lifecycle. Agents share a versioned workflow state — not free-form chat — so every transition is predictable and auditable.

AGENT · 01

Intake & Intent

Classifies the intent of each incoming email and extracts a structured shipment request from free text.

IntentExtraction
AGENT · 02

Shipment breakdown

Breaks a shipment into its ordered service components using a service-component ontology plus LLM reasoning over its attributes.

OntologyLegs
AGENT · 03

Partner Matching

Selects candidate partners for each component from the registry by capability, lane, mode, and service level — aided by retrieval.

RegistryRetrieval
AGENT · 04

RFQ Dispatch

Composes and sends a request for quotation to each selected partner per component, and tracks which requests are still outstanding.

RFQTracking
AGENT · 05

Quote Evaluation

Parses partner replies; extracts and normalises price, currency, lead time, validity, and conditions; aggregates component quotes into options.

ParseNormalise
AGENT · 06

Pricing & Margin

Applies configurable markup rules and ranks the consolidated options by price and lead time.

MarginRanking
AGENT · 07

Client Communication

Drafts the client-facing quotation and any follow-up correspondence in a clear, consistent format.

Quotation
AGENT · 08

Negotiation

Conducts bounded negotiation on a client's counter-offer, within configured floors and round limits.

Guardrails
AGENT · 09

Booking & Confirmation

On selection, confirms the booking with each chosen partner and records the confirmations received.

BookingConfirm
Breakdown

Every shipment, broken into priceable legs.

Breakdown is the heart of the Agent. A shipment becomes an ordered chain of components — each separately sourceable and separately priced — grounded in an explicit ontology, so the output is consistent and easy to validate against expert-labelled ground truth.

Special handling reshapes the chain. Flag a reefer, GDP-pharma, or dangerous-goods requirement and the Agent inserts the right tasks, restricts partners to certified providers, and reflects the premium in costing.

Run it on your lane
Human-in-the-loop

Agents do the legwork. People own the commitments.

Because the loop ends in commercially binding actions, oversight is a first-class design concern — configurable from fully assisted, where a human approves the quotation and every booking, to autonomous for low-value, high-confidence cases.

  • Approval gates. Nothing commits before review at the two points that matter — the client quotation, and every partner booking.
  • Confidence-scored. Every extraction and breakdown carries a confidence score; anything below your threshold is routed to a review queue.
  • Bounded negotiation. A floor margin, a maximum discount, and a round limit — beyond which the Agent escalates to a person.
  • Full audit trail. Every agent action, tool call, and human decision is written to an immutable log.
  • Assisted or autonomous. Review every quote, or let low-value, high-confidence cases flow on their own. You set the dial.
Start a pilot

See the Agent quote
your lane.

Point it at a mailbox and a handful of partners, and watch a real RFQ go from inbox to ranked quote. A Stoqr module, at pilot stage.