Flexport is moving significantly more freight without adding staff at the same pace, a sign of how artificial intelligence is changing the logistics company’s operating model. Its headcount has increased only 5% year over year, while its freight volumes have increased 40%, founder and CEO Ryan Petersen told FreightWaves in an interview.

“The productivity gap is because we are integrating more and more AI into the business, in areas like customs operations, transport routeing and software engineering,” says Petersen. The technology is enabling employees to automate complex processes, improve operational consistency and develop new services without commensurate increases in staffing.

But the company’s approach additionally emphasises the limitations of automation in highly regulated environments, particularly customs clearance, where supervision by humans remains a must.

The comments are coming after Flexport's fall launch of tech including an AI-based classification tool in its customs software and the expansion of its free Flexport Atlas mapping platform to include rail, barge and air freight routeing.

On October 15, Time magazine named the company's Customs Technology Suite one of the Best Inventions of 2026.

Customs classification continues to be supervised by humans

One of the most important uses of AI at Flexport relates to tariff classification, which Petersen calls one of the most difficult problems in customs operations.

The company’s new tool can suggest a 10-digit Harmonised Tariff Schedule classification and link to the underlying U.S. Customs and Border Protection rulings that support its suggestion.

But the AI is not legally able to perform customs operations for another party by producing a classification.

The distinction was noted in a January 16 ruling by CBP, H350722. In that instance, the agency found that a foreign online platform was engaging in customs business when its AI system classified tariff subheadings below the six-digit level for goods destined for entry into the United States.

Section 19 U.S.C. 1641 Customs brokers A licensed customs broker may conduct customs business for another.

Therefore, Flexport has added a human verification step to its process. If a customer wants to use an AI generated classification, the proposed code is sent to a licensed customs broker within the company for manual review before the filing is submitted.

Petersen said the arrangement is an example of how AI meets the practical needs of the real world.

There are technical problems too. AI models can generate incorrect results because CBP does not automatically delete old binding rulings when new decisions render them obsolete.

Petersen explained that both the old and the new rulings are available. So an AI model may depend on an outdated decision, but a seasoned customs broker would know that this decision should not be applied anymore.

Flexport identifies significant customs filing errors among competitors

Flexport is also using technology to scrutinize the accuracy of customs declarations submitted by other brokers.

In February, the company introduced a free auditing tool allowing businesses to examine historical customs entries prepared by their existing brokers. Petersen described the initiative as probably the most aggressive product Flexport has developed.

According to audits conducted by the company, customs entries filed by competing brokers showed an average error rate of 20%. Petersen attributed much of this inaccuracy to the considerable tariff changes and regulatory uncertainty experienced throughout 2025.

By comparison, Flexport reported an error rate of just 0.2% on its own U.S. customs filings last year. Petersen said this represented approximately 100 times better performance than competing brokers.

The comparison builds on comments he made when the auditing tool was introduced in February. At that time, he estimated that Flexport's 0.2% error rate represented a 100-fold improvement over other customs brokers, while the company's announcement clarified that the assessment was based on its knowledge of the wider brokerage market.

Maintaining such accuracy requires continuous updates to tariff databases, Petersen explained. Specialists must monitor regulatory developments and incorporate new requirements into the information used by customs professionals.

Flexport makes its tariff simulator available to the public free of charge. The same underlying data is also integrated into the company's internal brokerage workflows to support customs clearance.

Petersen believes that importers failing to use reliable tariff technology are approaching a level of negligence, given the financial and compliance risks associated with incorrect classifications.

AI-generated software helps to fulfilll customer-specific logistics needs

In addition to customs, Flexport is using AI to tackle another operational headache: managing the increasingly complex delivery instructions that individual customers set up.

Logistics businesses often have different requirements depending on destination, warehouse or shipment. For example, one facility may require 10 days’ notice prior to a delivery, while another may only accept deliveries with three days’ notice.

Petersen describes these requirements as a “complex web of business logic.”

Compliance becomes a necessity due to such factors as customer relations and commercial agreements and freight forwarders have to adjust. But it’s a risk to place all the faith in AI to comprehend and act on those instructions.

Mistakes can be especially costly when automated systems don't correctly apply a customer's requirements, Petersen said.

Flexport's solution is to use AI to generate software code to convert individual customer rules into repeatable operational instructions.

Once an expert has reviewed and approved the code it can run the requirements consistently every time and reduce the chances of unpredictable decisions being taken by artificial intelligence (AI) systems.

Tariff refunds may be driving up ocean freight demand

Petersen also spoke about the surprisingly robust ocean freight demand in 2026.

“The company’s peak shipping period began much earlier than we expected,” says Flexport CEO. Since April activity has been particularly strong.

Initially, Flexport attributed the spike to the typical early start of the peak season. But customers kept booking shipments that wouldn’t get to their destinations before Christmas. Petersen took the activity as genuine market growth, not just a seasonal swing.

He has also indicated the boom in demand may be associated with tariff rebates.

In February, the U.S. Supreme Court invalidated tariffs imposed under the International Emergency Economic Powers Act (IEEPA). As of Oct. 2, CBP had sent refunds of about $126 billion to the Treasury Department for distribution, Supply Chain Dive reported, citing a court filing.

“It’s a big economic stimulus for the logistics industry,” Petersen says. “In particular, a lot of companies didn’t expect to get this money.

Some customers have told him they intend to use the refunds to purchase more inventory, he says. This could be especially important for importers who have been struggling with persistent capital constraints.

But Petersen has warned customers against getting over-committed to buying new inventory, likening unexpected financial windfalls to lottery winnings that can be gone in a flash with poor financial decisions.

Ocean carrier capacity still driving freight rates

Demand is not the only factor underpinning the ocean freight market. Petersen also noted changes in the structure of the shipping industry.

He estimates there are only 10 or 11 major ocean carriers left and operators are getting better at managing the capacity that is available. By cancelling sailings and slowing vessels, carriers can control supply and maintain freight rates.

Trans-Pacific prices were $8,126 per 40-foot container for shipments from Asia to the U.S. West Coast in Freightos’ weekly market update published October 20. Rates on the East Coast jumped to $9,700 a container.

The report characterised prices as stable but still very high, although it was expected that shipping volumes would ease at some point.

Still, those rates are below the exceptional levels seen during the pandemic-era shipping boom of 2021, as FreightWaves pointed out in September.

Data center expansion is creating new pressure on air cargo capacity

While Petersen remains reluctant to make predictions about the ocean freight market, he is closely monitoring the growing influence of data center investments on air cargo demand.

He estimates that shipments associated with data center construction and development, including semiconductor chips, server racks and building materials, already account for approximately 8% of global air freight capacity.

That share is projected to triple next year, potentially creating substantial capacity constraints on routes connecting Taiwan and Southeast Asia with the United States.

During a Flexport conference, an air freight specialist estimated that approximately 30 fully loaded freighters travel daily from Taiwan to the United States specifically to transport data center-related cargo, Petersen said.

The financial value of these shipments helps explain why companies increasingly favor air transportation over ocean freight.

According to Petersen, a single freighter carrying exclusively graphics processing units (GPUs) and server racks can transport cargo worth as much as $500 million.

Waiting approximately 30 days for an ocean shipment would generate additional financing costs on such valuable goods while also delaying the revenue-generating operation of the completed data center.

For these reasons, Petersen believes that a significant proportion of this specialized cargo will continue moving by air for the foreseeable future, potentially indefinitely.

Despite the growing volumes, the effect on air freight rates has been less pronounced than he initially expected.

One explanation is the arrival of additional capacity. Airlines in the Middle East are bringing aircraft back into service and have also acquired more dedicated cargo planes, helping accommodate part of the increased demand.

Flexport considers smaller engineering teams as AI accelerates development

Artificial intelligence is also changing how Flexport structures its technology workforce.

Petersen revealed that the company is considering reducing the traditional size of its engineering teams from approximately eight people to groups of just two or three engineers.

The approach reflects the productivity gains made possible by AI-assisted development. Petersen cited Flexport Atlas as an example, explaining that the platform originally started as a project handled by a single engineer.

He also challenged companies concerned about the cost of AI tokens, arguing that businesses should consider increasing their investment rather than restricting usage.

His reasoning is based on an estimate that AI tools can make an engineer approximately seven times more productive.

The company's product development schedule is evolving accordingly. Flexport has organized two major technology releases annually over the past two years, with its autumn announcement marking the fourth release under this approach.

However, Petersen indicated that the company could increase that frequency to three or four major releases in 2027 as engineering teams accelerate their development cycles.

He also dismissed concerns that leading AI developers could eventually become direct competitors in freight forwarding.

In his view, the major AI model providers are unlikely to transform themselves into logistics operators. Instead, Flexport intends to use their technologies aggressively to improve its operations, expand its product portfolio and extend the range of services it offers.

AI is becoming a competitive advantage across freight forwarding

Flexport's latest performance figures illustrate how artificial intelligence is beginning to influence productivity, workforce organization and service development across the freight forwarding sector.

The company increased its freight by 40% with only a 5% increase in headcount, showing the potential operational gains from automation. It can be used to classify customs, monitor tariffs, define customer-specific delivery requirements, and for software engineering.

But the company’s customs operations highlight an important difference: Technology automation can’t replace legally required professional oversight.

Flexport claims a 0.2% average error rate versus a reported 20% among competing brokers—but the comparison is based on audits conducted by Flexport itself, and it’s a point the company emphasises in its emphasis on accurate data and regulatory expertise.

At the same time, tariff refunds, volatility in ocean carrier capacity and the explosive growth of data centres are changing freight demand in both ocean and air transport.

Flexport's next challenge will be to continue to maintain this productivity advantage as it grows its offerings and evolves to a more technology-driven logistics market.

SpaceX Launches Starlink Satellite Constellation

Related coverage

Runnings deploys Relex AI to unify forecasting and inventory replenishment

TrashLab secures $12 million to expand AI automation across waste management operations

Flexport boosts freight volumes by 40% while limiting workforce growth to 5% through AI