RFQ + Freight AI
At a glance
- Live at: Aramex (freight quotation), cross-border freight customers
- Channels: email (primary), API
- Languages: EN
- Owner: Freight pod
- Demo: freight-rates.app.shipsy.ai
The problem it solves
Freight quotation is one of the most labor-intensive processes in logistics. An RFQ arrives by email with shipment details (origin, destination, commodity, weight, dimensions, Incoterm). A human analyst must:
- Parse the request (often in varied, unstructured formats)
- Look up applicable rates across carriers and routes
- Apply customer-specific and route-specific margins
- Assemble the cost components based on the Incoterm
- Generate and send a formatted quote — all within SLA
At scale (thousands of RFQs/day), this becomes a bottleneck. Slow quotes lose deals; wrong quotes erode margins.
How it works
Step by step
- Intake — RFQ email arrives. Agent extracts structured data from unstructured text.
- Data extraction — Identifies: origin, destination, commodity type, weight, dimensions, required Incoterm, any special requirements.
- Rate lookup — Queries rate databases for applicable tariffs across carriers and routes.
- Margin assembly — Applies the right margin structure based on customer tier, route, and mode of transport.
- Cost component structuring — Different Incoterms include different cost components. EXW = minimal (ex-works price only). DDP = maximal (all costs through delivery including duties and taxes).
- Quote generation — Produces a formatted quotation document.
- Confidence check — High-confidence quotes auto-send. Low-confidence quotes route to a human for review.
Incoterm coverage
The agent handles all 11 Incoterms, each with a different cost component structure:
| Group | Incoterms | Seller’s responsibility |
|---|---|---|
| E | EXW | Minimal — buyer arranges everything |
| F | FCA, FAS, FOB | Seller delivers to carrier/port |
| C | CFR, CIF, CPT, CIP | Seller pays freight to destination |
| D | DAP, DPU, DDP | Seller delivers to final destination |
Tech stack
| Layer | What’s used |
|---|---|
| Models | LLM for email parsing + data extraction. Classification for Incoterm identification. |
| Memory | Long-term: rate history, customer pricing agreements, margin rules. |
| Tools (MCP) | Rate database lookup, carrier API, customer CRM (pricing tier), email send. |
| Guardrails | All quotes above a configurable threshold require human approval. Margin floors enforced. |
| Evals | Scenario-based across all 11 Incoterms; adversarial inputs (incomplete data, contradictory requirements). |
Why CS folks should study this
- Email-native agents. Most agent discussions focus on voice/chat. RFQ shows that email is a high-value channel too — especially in B2B freight.
- The Incoterm complexity. Understanding the 11 Incoterms and how they change cost structures is useful knowledge for any freight-facing CS person.
- Margin protection. The agent doesn’t just automate — it enforces pricing discipline. This is a selling point for finance teams.
Sources
- Aramex Architecture (includes RFQ flow)
- Aramex Freight Quotation Scenarios — all 11 Incoterms (21 slides)
- Aramex Freight AI Quotation Demo
- Demo: freight-rates.app.shipsy.ai
- See Aramex case study for the full engagement context
Changelog
- 26 May 2026: Full content from Aramex freight quotation decks and demo links.