A container ship seen from above, mid-ocean.

Public evidence shows that AI quoting can work in selected freight workflows. It also shows why buyers must check the mode, workflow and stage behind each claim.

C.H. Robinson reports scaled truckload and less-than-truckload quote handling. DSV has demonstrated an Air and Sea workflow as part of a product roadmap. DP World says it combines an internal quoting tool with bought rate aggregation. These are different kinds of evidence.

This guide separates what each company has disclosed from the decisions a forwarder still needs to make. Nexcade sells a bought workflow layer, so read our conclusions with that interest in mind.

Public examples have different evidence levels

C.H. Robinson reports scaled road-freight quoting

C.H. Robinson’s October 2024 release says its software reads email quote requests and retrieves the price from its Dynamic Pricing Engine. The release says the workflow started with truckload and expanded to less-than-truckload quotes.

Its June 2025 update reports more than one million quotes and a 32-second response time. These are company-reported operating measures. The update does not identify an air or ocean forwarding workflow.

A 2026 company article describes hundreds of connected agents. The company’s first-quarter presentation separately names Ocean Quoting. It gives no equivalent workflow, volume or result measures for that example.

DSV has demonstrated an Air and Sea roadmap

DSV’s May 2026 Capital Markets Day included an Air and Sea email quotation demonstration. The workflow reads a request, identifies missing information, queries DSV’s rate engine and returns a portal quote. It also records the agent’s steps for human control and transfers an accepted quote into a booking.

The official presentation labels spot quotation enhancement as a product roadmap. DSV has not published a production launch date, coverage, processed volume or measured result for this workflow.

DP World describes a mixed architecture

A cargo.one release quotes DP World saying that it is developing internal quoting tools and using cargo.one to aggregate airline rates. This is a vendor-published statement about system design. It does not provide operating results.

Supporting systems come before the agent

The public examples show that the agent depends on other systems.

  • C.H. Robinson’s agent reads the email while its Dynamic Pricing Engine supplies the price
  • DSV’s demonstrated workflow queries its rate engine and hands an accepted quote into booking
  • DP World says it buys airline rate aggregation for an internally developed quoting tool

A build therefore needs product ownership across source access, pricing rules, exceptions, approvals, integrations and monitoring. The language model is one component.

Each layer has a different ownership decision

Layer What ownership involves What to test
Rate access Carrier and marketplace connections, commercial terms and coverage Whether an established provider covers the required modes, regions and carriers
Rate management Importing, normalising, versioning and distributing contracted rates Data quality, update process and access from the quoting workflow
Pricing policy Margin rules, customer terms, approvals and exceptions Whether business owners can change policy safely and audit each decision
Workflow agents Request intake, missing details, source selection, exception handling and monitoring Performance on difficult live requests and the team’s capacity to maintain it
System of record Permissions, field mapping, writeback and recovery from failed actions Reliable behaviour in the forwarder’s TMS and a clear audit trail

The build decision can differ by row. A forwarder can own its pricing policy and buy rate access. It can also retain its TMS while building or buying the workflow that connects requests to it.

Use five tests to decide

Use these questions before choosing a build or a vendor:

  • which layer provides a real commercial advantage when you own it
  • which existing systems already provide the rates and official records
  • whether your team can maintain changing formats, credentials, workflows and exceptions
  • how you will measure quote quality, response time, margin and operator intervention
  • how the system behaves when information is missing or an integration fails

Test the hardest representative requests. A successful standard quote says little about an agent’s handling of missing rates, unusual cargo, customer-specific rules or a failed TMS writeback.

Test the workflow with real requests

Run a controlled trial with representative requests from your own inbox. Protect customer and supplier data under your normal security rules.

Include requests that test:

  • different email formats, attachments and missing shipment details
  • a missing rate and partial replies from suppliers
  • customer-specific pricing and approval rules
  • correct TMS fields, links and status after writeback
  • safe recovery from a failed login, timeout or interrupted action
  • duplicate requests, changed instructions and operator takeover

For each request, keep the source material, proposed action, operator edits, approval record and final system state. This gives the build team and any vendor the same evidence to review.

Nexcade is a bought workflow layer

Nexcade provides agents for coordination, quotations, tenders, bookings and analytics. The agents work with a forwarder’s existing inbox, rate sources and TMS. Review gates control consequential work.

Buyers should assess Nexcade against the same tests as an internal build. Check source coverage, difficult requests, approval controls, writeback, monitoring and ongoing ownership.

For the wider category comparison, see the field guide to freight quoting software.

Common questions

Has anyone deployed AI quoting at scale?

C.H. Robinson reports scaled quoting agents in truckload and less-than-truckload workflows. It has not published equivalent operating measures for its separately named Ocean Quoting example. DSV's Air and Sea quotation workflow remains a roadmap demonstration in public evidence.

How long does an in-house AI quoting build take?

Public evidence does not provide a reliable benchmark. C.H. Robinson says its first AI agents date to 2023 and rely on an existing pricing engine. DSV presents its quotation roadmap within a staged core-system transition. Treat the work as ongoing product ownership.

Do companies that build still buy components?

Yes. DP World says it is developing internal quoting tools while using cargo.one for airline rate aggregation. A company can own its workflow while buying rate access or other established components.

What should a mid-size forwarder do?

Separate the decision by layer. Keep ownership of pricing policy and approval rules. Assess bought rate access and management where suitable services exist. For the workflow layer, compare internal engineering capacity with the need to maintain source access, exceptions, controls and TMS writeback.

Sources

These public sources support the factual claims and product descriptions above. Evidence last reviewed 16 July 2026.