Wednesday, November 4, 2026
The news that moves freight
Tech & AI Nov 4, 2026 · 7 min read

Flexport: volumes up 40%, staff up 5%

Flexport’s 40-to-5 ratio sets a new bar for forwarder productivity

Flexport says its forwarding volumes are up 40% on last year, while its headcount grew just 5%. Chief executive Ryan Petersen put the gap down to artificial intelligence. “Our volumes are up 40% annually. And headcount’s only up 5%, so you can see the AI’s working,” he told FreightWaves in an interview published on Monday 2 November.

Do the sum and it comes to about a third more freight per employee in twelve months (1.40 divided by 1.05 is roughly 1.33). That is our arithmetic, not Flexport’s.

Our view is simple. The figure is unaudited and loosely defined, so do not take it as gospel. But it is now the benchmark. Any forwarder that cannot show its own volume-per-head trend will be asked why, by customers, investors and its own board.

What Flexport actually claimed

The numbers come from one interview, written up by Thomas Wasson at FreightWaves. Flexport is a private company. It does not publish audited accounts. The article does not say whether “volumes” means containers, shipments, tonnes or revenue. Nor does it give the base numbers behind the 40% and 5%.

There is also history to keep in mind. In October 2023, Petersen told staff that Flexport would “reduce the size of our global team by approximately 20%”, according to CNBC. The company employed about 3,500 people at the time, CNBC reported, citing Pitchbook data. A 5% rise in headcount starts from a base that was cut hard.

None of that makes the claim false. It means the ratio says as much about a leaner company as it does about AI.

Petersen gave other figures in the same interview:

  • Flexport’s error rate on US customs filings in 2025 was 0.2%, which he put at about 100 times better than rivals.
  • Audits of entries filed by competing brokers found an average error rate of 20%, he said.
  • AI tools make an engineer about seven times more efficient, he estimated. Flexport is considering engineering teams of two or three people, down from eight.
  • Flexport has held two large technology releases a year for the past two years. The fall release was its fourth. It may move to three or four releases in 2027.

Flexport’s fall 2026 release page lists AI customs classification, an “AI Business Rules Engine” that turns customs procedures into code that checks every entry, and AI agents that quote and book from emailed requests. It also lists a link that lets users query Flexport from AI assistants such as Claude, ChatGPT and Copilot.

The customs claim is the one to watch

The 0.2% versus 20% comparison is the most commercially pointed thing Petersen said. It is also the one that needs the most care.

The 0.2% is Flexport’s own measure of its own work. The 20% comes from Flexport’s own audits of other brokers’ entries. The article does not say how those entries were picked. In February Flexport launched a free tool that lets any company audit its broker’s past filings, and firms that already suspect a problem are the likeliest to come forward. So the 20% should not be read as a random sample of the market. A Flexport release quoted by FreightWaves calls the 100-times estimate “based on what we know of other customs brokers.”

Even so, the direction fits what Flexport has said before. In January, Flexport’s global head of airfreight, Alexis Boutet, told Air Cargo Week that its Customs Analysis product “reviews up to 100 percent of customs entries alongside a human”. He compared that with “the industry norm of five to ten percent”. If most brokers only spot-check one entry in ten or twenty, errors in the rest go unseen.

The legal limits matter here too. Flexport’s AI suggests tariff codes, but a licensed customs broker must review them before anything is filed. FreightWaves points to US law (19 U.S.C. 1641) and a January 2026 CBP ruling that keep customs business with licensed brokers. Petersen called tariff classification “the hardest problem in customs”. The US tariff schedule has more than 18,000 ten-digit codes, according to Flexport’s release page.

For customer-specific rules, Flexport has AI write code, and a human expert signs it off before it runs. Petersen’s reason: “If it misses, the misses are worth too much.”

Our read: this is the model most forwarders will copy. AI does the first pass. A licensed human signs. The productivity gain comes from the human checking more entries, not from the human disappearing.

Why accuracy is worth money right now

The timing is not an accident. As of 2 October, US Customs and Border Protection had sent about $126 billion in refunds to the Treasury Department for payout, FreightWaves reports. The refunds come after the Supreme Court in February struck down tariffs imposed under IEEPA, the International Emergency Economic Powers Act, the emergency law used for many of the 2025 tariffs.

Petersen called the refunds a “massive stimulus for the industry”. He also had a warning for shippers getting cheques. “Most lottery winners go bankrupt,” he said he tells customers. “Don’t go crazy buying inventory.”

In our view, clean entry records matter for refunds. An importer whose broker filed wrong codes or values may wait longer, get less, or face questions. That gives a broker with a provable low error rate a strong pitch. We expect Flexport to keep pushing its audit tool at rival brokers’ customers while refunds flow.

The rest of the market tells the same story

Flexport is not the only firm claiming more freight per person. C.H. Robinson’s third-quarter results, released on 28 October, showed average employee headcount down 10.4% across the company. In its global forwarding unit, average headcount fell 15.8% year on year. Over the same period, ocean shipments rose 2.0% and air tonnage fell 4.0%.

Chief executive Dave Bozeman said that, across its North American Surface Transportation and Global Forwarding units, “both businesses delivered a higher level of year-over-year productivity improvements”. He said “the team continues to turn manual processes into agentic workflows” (agentic means AI software that carries out a task end to end, such as booking a load, rather than just suggesting).

The two firms are getting there in different ways. C.H. Robinson is holding volume roughly flat with far fewer people. Flexport says it is growing volume fast with roughly the same number. Both end up with more freight per head.

C.H. Robinson’s customs unit is a warning sign. Its customs adjusted gross profit fell 12.9% in the quarter, driven by a 13.0% drop in profit per transaction. Our read: customs is getting cheaper to deliver, and that saving is going to customers.

What it means for forwarders

Four things follow, in our view.

First, cost per shipment is now the comparison. If a rival moves a third more freight per head, it can quote lower and still make money. Mid-sized forwarders without the budget for in-house AI will feel this first.

Second, customs quality is becoming something you have to prove with a number. Saying “we’re careful” will not win against a competitor that quotes an error rate. Forwarders with brokerage arms should start measuring their own entry error rate now, before a rival’s audit does it for them.

Third, the human role shifts, it does not vanish. US law keeps a licensed broker in the loop. The job becomes checking far more AI-prepared entries, not typing them.

Fourth, demand is still there to fight over. Petersen said Flexport’s ocean peak began around April: “Basically since April, we’ve been just ripping.” Freightos put Asia–US West Coast rates at $8,126 per 40-foot container and East Coast rates at $9,700 as of 20 October, FreightWaves reported. In air, Petersen estimated data-centre cargo at about 8% of global air capacity, projected to triple next year. He said lanes from Taiwan and Southeast Asia to the US will be “really constrained”. An air freight expert at Flexport’s conference put the number of full freighters flying from Taiwan to the US just for data centres at 30 a day.

On the AI labs themselves, Petersen was clear. “The frontier companies are not going to become a logistics provider,” he said. In our view, the threat to a forwarder is another forwarder using AI better, not OpenAI or Anthropic moving freight.

Our prediction

By 31 March 2027, at least one of DSV, Kuehne+Nagel, Expeditors or C.H. Robinson will publish a customs-accuracy figure or a shipments-per-employee figure in its full-year results or investor material. Customers and investors will want a number to set against Flexport’s 0.2% and its 40-to-5 ratio, and the big listed firms will not want a private rival setting that bar unanswered.

We also expect Flexport’s next release, due in early 2027 if it keeps its twice-a-year schedule, to lead with customs and refund tools rather than new freight products.

Sources

Read us on Google? Add Forwarder News as a preferred source to see more of our stories in Top Stories.