How is Flexport using AI to increase volume capacity without hiring proportionally?

Flexport deployed AI tools across three functional areas: customs classification (suggesting 10-digit tariff codes with supporting CBP rulings), route optimization for rail, barge, and air modes, and internal engineering workflows. The 40% volume growth with only 5% headcount expansion reflects AI handling routine analytical tasks that previously required manual effort. However, all AI outputs remain subject to human review before implementation, particularly in customs work where regulatory compliance is non-negotiable.

Why can't AI independently handle customs tariff classification?

Under 19 U.S.C. 1641, only licensed customs brokers can conduct customs business on behalf of other parties. A January 2026 CBP ruling found that an AI tool deriving tariff codes beyond the six-digit level constituted unauthorized customs business. Additionally, AI models frequently hallucinate when classifying goods because CBP leaves outdated binding rulings publicly posted alongside newer superseding rulings, creating training data ambiguity that experienced brokers resolve through professional judgment.

What is Flexport's error rate advantage in customs filings?

Flexport reported a 0.2% error rate on U.S. customs filings in 2025, compared to an audited average of 20% at competing brokers. Flexport attributes this performance difference to continuous expert-driven tariff database maintenance, AI-assisted classification with mandatory broker review, and systematic workflow integration. However, the competitive comparison is based on what Flexport knows of other brokers rather than industry-wide audits, so actual performance variation across brokers may exist.

How does Flexport prevent AI from making customer-specific business rule errors?

Rather than letting AI execute customer-defined delivery rules autonomously, Flexport has AI write code that encodes those rules (such as 10-day notice for one warehouse or three days for another). A human expert reviews and approves the code before it becomes part of the operational workflow. Once signed off, the code runs deterministically, eliminating the risk of inconsistent AI interpretation of complex 'spaghetti web' logic that varies by customer.

What market tailwinds is Flexport experiencing alongside AI productivity gains?

Flexport's ocean freight peak began in April 2025 and remains strong, with customers booking cargo unlikely to arrive before Christmas. CEO Ryan Petersen attributes sustained demand partly to tariff refunds: the Supreme Court invalidated IEEPA-based tariffs in February, and approximately $126 billion in refunds had been sent to the Treasury by October 2, 2025. Many customers are using these unexpected refunds for inventory investment, creating demand tailwinds that coincide with Flexport's AI-driven capacity improvements.

Technology & Innovation

High Impact

Flexport volumes surge 40% with only 5% headcount growth via AI

Nov 2, 2026

Get tomorrow's supply chain signal

Daily supply-chain brief. Free, unsubscribe anytime.

The signal

Flexport has achieved a significant productivity breakaway: freight volumes increased 40% year-over-year while headcount grew only 5%, demonstrating the operational leverage AI can unlock in logistics. The company attributes this gap to artificial intelligence deployed across customs classification, shipment routing, and internal engineering teams. However, the implementation reveals important constraints around AI in regulated environments.

Customs work remains highly controlled, requiring human licensed brokers to review all AI-generated tariff codes before submission to U.S. Customs and Border Protection, highlighting the reality that AI augments but cannot replace regulatory expertise. Flexport's own customs filing error rate of 0.2% reportedly compares favorably to a measured 20% error rate at competing brokers, though this advantage stems partly from continuous tariff database maintenance and human oversight.

The company's technology strategy underscores an emerging principle in supply chain automation: AI excels at pattern recognition and data processing but must integrate with human judgment where regulatory or high-consequence decisions are required.

Companies affected by this story:

Gap Inc.

Open the full daily intelligence brief · Pulse 78

Target

Open the full daily intelligence brief · Pulse 79

Wayfair

Open the full daily intelligence brief · Pulse 81

#ai-customs-automation

#flexport-volumes

#tariff-classification

#logistics-efficiency

#supply-chain-technology

#customs-broker-ai

#freight-forwarding

#operational-efficiency

#regulated-ai-deployment

Frequently Asked Questions

What This Means for Your Supply Chain

Simulation Suggestion

strategic

What if competitors achieve similar AI-driven productivity in the next 12 months?

Model the impact on Flexport's volume growth if competitors deploy comparable AI tools for customs classification and routing optimization, reducing the company's efficiency differentiation. Assume competitor error rates decline toward Flexport's 0.2% baseline and that operational labor requirements per shipment converge. Assess pricing pressure, market share volatility, and margin compression.

Run this scenario

Simulation Suggestion

this month

What if tariff refunds dry up and inventory demand normalizes?

Simulate demand normalization if the $126 billion tariff refund program exhausts and importers shift from inventory buildup to just-in-time ordering. Model the impact on Flexport's volumes, pricing environment, and freight rate volatility. Assess how quickly ocean capacity normalizes if peak demand recedes and carrier discipline tightens.

Run this scenario

Simulation Suggestion

strategic

What if CBP tightens AI automation requirements beyond tariff codes?

Model regulatory escalation: CBP issues guidance restricting AI assistance in other customs functions (entry summary filings, valuation, origin determination) or requiring pre-approval of AI tools before use. Assess operational friction, compliance costs, and time-to-market for new AI features if broker approval cycles lengthen.

Run this scenario

Go deeper:

Supply Chain Intelligence Software: The 2026 Buyer's Guide

The risks intelligence software should surface, how it compares with supply market intelligence, and how to evaluate platforms.

Technology & InnovationHigh Impact

Flexport MCP Server Lets AI Agents Book Freight Autonomously

Sep 29, 2026

Demand PlanningCritical

AI boom drives $23B data center logistics market to $35B by 2030

Oct 8, 2026

Technology & InnovationHigh Impact

Flexport AI Agents Book Freight: Who Bears Liability?