AX Field Notes

All casesNo. 054

C.H. Robinson Hands Missed LTL Pickups to Two AI Agents

One agent calls the carrier, another picks the next step; 95% of checks are now automated

Case no.
054
Primary source
Company press release and earnings call
Source published
2026-01-26

1. Why this case

C.H. Robinson is a third-party logistics provider (3PL) that, by its own count, works with 74,000 customers and 450,000 contract carriers and manages 37 million shipments a year. In less-than-truckload (LTL) shipping, where one truck collects freight from several shippers, the company says it “moves more LTL freight than any other 3PL in North America.” On its Q4 2025 earnings call it described the LTL business as “greater than $3 billion.” Both are company statements.

This case is about one exception in that business: the pickup that did not happen. Before, people followed up on each missed pickup by hand, checking carrier websites, calling carriers, recording updates, and notifying customers. When they could not confirm the shipment’s status, they retendered it, and another truck often arrived to find the freight still not ready or already gone.

In a press release dated 2026-01-26, C.H. Robinson says two AI agents now handle this work. One calls carriers about missed pickups. The other decides what steps to take. The company reports that 95% of checks on missed LTL pickups are now automated, saving over 350 hours of manual work per day, and that unnecessary return trips for missed freight are down 42%.

The case is worth reading for three reasons. The target was chosen by measuring wasted time, not by looking for a place to use AI. The work is split between gathering facts and deciding, which maps onto where the hours and the wasted trucks came from. And the call results became a new dataset that carriers now receive every day. All figures are company-reported.


2. Problem

2.1 Why LTL pickups go wrong

The press release explains the mechanics: “With one truck carrying freight from up to 20 different shippers, LTL shipping requires complex coordination to pick it all up, take it to a terminal and recombine it on other trucks with other freight heading the same direction. Across the industry, that leads to missed pickups and costly delays that ripple through LTL networks.”

Greg West, Vice President for LTL, describes what a single miss does: “A missed pickup isn’t just a minor inconvenience,” he said. “When a truck arrives and the freight or packaging isn’t ready, or the carrier couldn’t make it because they got stuck in traffic, it forces another truck to come back the next day. That might not even be our shipper’s freight, but it creates a domino effect for other freight that was supposed to get picked up and for all the other trucks down the line.”

2.2 The manual chase

West describes the work before the agents:

“Before this transformational tech, teams of people spent over half the day chasing missed pickups – manually checking carrier websites, making calls, recording updates and notifying customers.”

The chase had four manual tasks: check the carrier’s website, call, record the update, and tell the customer. If those did not produce an answer, the fallback was costly:

“If we couldn’t nail down the shipment’s status, we’d have to retender it and carriers would send another truck, often to find the freight still wasn’t ready or had already been picked up.”

So the problem had two costs. One was labor: hours of follow-up, of which the agents now save more than 350 per day. The other was capacity: trucks dispatched on a retender without a confirmed status, which could find nothing to pick up. On the earnings call, Michael Castagnetto, President of North American Surface Transportation, called the 350 hours “outsourced manual work.”

Confirmed facts: up to 20 shippers’ freight on one LTL truck; causes named = freight or packaging not ready, carrier delayed by traffic; a miss forces another truck the next day and delays other freight; manual follow-up = checking carrier websites, calls, recording updates, notifying customers; teams spent over half the day on it; unknown status led to a retender and another truck, often finding freight not ready or already picked up; the call describes the 350+ hours as outsourced manual work.


3. Solution

3.1 Finding the waste first

Mark Albrecht, Vice President for Artificial Intelligence, frames the project as an example of what the company calls Lean AI. “We don’t just throw AI at anything and everything,” he said. “It’s not a hobby for us. We use AI agents only where they can deliver tangible business results. Our Lean AI processes helped us uncover the extent of time wasted in handling missed pickups and where artificial intelligence had the most potential to augment our automation software.”

Two points stand out. The size of the problem was measured before the agents were built. And the agents were placed to “augment” existing automation software, not to replace it.

3.2 One agent calls, one agent decides

The release describes the division of work: “a new AI agent calls carriers about missed pickups and then another new AI agent decides what steps to take.”

It also describes the agents as “tracking down missed pickups and using advanced reasoning to determine how to keep freight moving.” On capacity: “Because the two AI agents working together can make 100 calls and 100 decisions simultaneously, missed pickups are resolved faster, freight gets on the road faster, and shippers and their receivers get visibility to their freight sooner without hours of manual intervention.”

The 100 calls and 100 decisions are a measure of how many the agents can run at once.

3.3 Missed-pickup data goes to carriers every day

The agents also produce data that did not exist before. The release says they are “collecting and analyzing previously unavailable data that LTL carriers are now using to improve their technology, scheduling and operations.”

West names three uses: “Sharing our new missed-pickup data with carriers every day allows them to see which electronic communications could be improved, isolate operational issues to particular terminals and optimize their scheduling.”

3.4 Rollout: Freightquote first, then all LTL customers

Albrecht describes the order of rollout: “The agents first started assisting the small and medium-size customers who use our Freightquote platform. Then in July we started scaling the agents across our LTL customers.” On the earnings call two days later, Castagnetto said the agents were “launched in 2025,” which places that July in 2025.

As of the release, “the new AI agents are resolving hundreds of shipments a day across more than 11,000 customers so far.”

The two agents join a larger set. The release says they are part of “C.H. Robinson’s fleet of 30+ AI agents” and that “AI agents also handle LTL price quotes, orders, freight classification, shipment tracking and proof of delivery.”

3.5 What people do now

On the Q4 2025 earnings call on 2026-01-28, an analyst asked how the company balances headcount reductions with the human touch that some shippers and carriers prefer. Castagnetto used the missed-pickup agent as his example. As transcribed:

“we said in the release, 350 hours a day of human work that was done just to follow up on missed pickups. Well, that’s an example of, you know, we won’t have to have humans doing that in the future, but it’s not really a high-value concept in terms of do we need somebody doing that role. The relationship is now our people will get that information more accurately, they’re gonna get it quicker, and they’re gonna be able to call their customers and talk to them about a solution to a pickup that maybe was missed, and what are we gonna do about it. How do we fix this for the next time?”

The follow-up itself moves to the agents. The customer conversation about what to do next stays with people, who now start it with the status already in hand.

Confirmed facts: Lean AI process measured the time wasted before the build; agents augment existing automation software; one agent calls carriers, a second decides next steps; capacity = 100 calls and 100 decisions simultaneously; missed-pickup data shared with carriers daily for three named uses; rollout = Freightquote small and medium-size customers first, then all LTL customers from July (2025 per the call); hundreds of shipments a day across 11,000+ customers so far (as of 2026-01-26); part of a 30+ agent fleet; LTL agents also handle quotes, orders, freight classification, tracking, proof of delivery; per the call, humans will no longer do the follow-up and staff will use the information to talk to customers about fixes.


4. Impact

Figures from the 2026-01-26 press release, repeated on the 2026-01-28 earnings call. No baseline period or measurement window is given.

MetricFigureBasis
Checks on missed LTL pickups automated95%share of checks, not of pickups resolved
Manual work savedover 350 hours per day“outsourced manual work” per the earnings call
Freight speedmoves up to a day fastermaximum, no average given
Unnecessary return trips for missed freightreduced by 42%baseline not stated
Volume handledhundreds of shipments a day across 11,000+ customers“so far,” as of 2026-01-26
Agent capacity100 calls and 100 decisions simultaneouslyconcurrency, not daily volume

The release names who gains. For carriers, fewer return trips; for shippers and receivers, earlier visibility and faster freight. West ties the return-trip saving to other shippers as well: “Now that all that time and capacity aren’t being wasted, it keeps other shippers’ freight from getting delayed.”


5. Insight

The target was an exception queue, picked by measuring waste. Missed pickups are not a product or a customer-facing feature. They are a follow-up task that consumed more than 350 hours a day. Albrecht says the Lean AI process found “the extent of time wasted” before the agents were built. The selection rule is narrow on purpose: agents go “only where they can deliver tangible business results.”

Facts first, then the decision. The old process had a gap between not knowing and acting. When people could not confirm a shipment’s status, they retendered it and a second truck went out, often for nothing. The new process puts a call agent in front of the decision agent, so the decision is made after the carrier has been asked. The release does not break down where the 42% reduction in unnecessary return trips came from, but West’s description of the old retender loop shows where that waste sat.

The calls became a carrier dataset. The agents collect data the company calls “previously unavailable.” Shared with carriers every day, it lets them see which electronic communications could be improved, isolate problems at particular terminals, and adjust scheduling. The by-product reaches parties outside C.H. Robinson.

The follow-up job goes away; the customer conversation stays. Castagnetto says humans will not be needed for the follow-up in the future, and that staff will instead get the information faster and more accurately and call customers about a solution. He called the saved hours “outsourced manual work,” which suggests the follow-up had been contracted work rather than C.H. Robinson staff time.

Limits. All figures are company-reported, with no baseline, period, or error rate. A day faster is the upper end, not an average. The release does not describe how the calls are made, what the decision agent may do on its own, how mistakes are caught, or what happens to the 5% of checks that are not automated. The “outsourced” wording comes from one third-party transcript of the call.

One-line takeaway: C.H. Robinson took a follow-up job that cost hundreds of hours a day, gave the phone calls to one agent and the next-step choice to another, and left people to talk to customers once the status is known.


6. Implementation hypotheses (not in the source)

6.1 A trigger, an electronic check, then a call

Basis: The release says the agents are “tracking down missed pickups,” that the old work was “manually checking carrier websites, making calls, recording updates,” and that 95% of “checks on missed LTL pickups” are automated.

The likely flow starts in the shipment record. When a scheduled pickup passes its window without a pickup confirmation from the carrier, the shipment is flagged as a possible miss. The automation software would first try the cheap checks, such as the carrier’s tracking data or website, because a pickup that did happen but was never confirmed electronically can be closed without a call. Only when that check returns nothing useful would the call agent phone the carrier’s dispatch or terminal. The call would aim to fill a small set of fields: was the pickup attempted, why it was missed, and when the carrier will return. Those fields would be written back to the shipment record, which replaces the manual “recording updates” step. Running 100 calls at once suggests the calls go through a telephony service that can place many outbound calls in parallel, with each call tied to one shipment ID.

6.2 A bounded set of next steps for the decision agent

Basis: “another new AI agent decides what steps to take”; the old fallback was to “retender it” when status was unknown; the agents “augment our automation software.”

The decision agent most likely chooses from a short list of actions that the existing software already supports, rather than acting freely. Plausible options are: accept the carrier’s new pickup date and update the estimated delivery; retender to another carrier when the first cannot return in time; ask the shipper to confirm the freight is ready before a truck is sent; and notify the customer and receiver. The order matters. Retendering would be allowed only after the call has confirmed the freight is ready and not already picked up, which is the condition the old process could not check. Cases where the carrier cannot be reached or the answers conflict would go to a person. That queue is the likeliest home for the 5% of checks that are not automated, and the call summary would be what a staff member reads before calling the customer.

6.3 Reason codes rolled up into a daily carrier report

Basis: West says the daily missed-pickup data lets carriers “see which electronic communications could be improved, isolate operational issues to particular terminals and optimize their scheduling.”

For that to work, each call result would need to be stored as structured data: carrier, terminal, scheduled pickup time, reason for the miss, and whether the carrier’s electronic status matched what the call found. A daily job would group these by carrier and terminal. The comparison between electronic status and call result is the part that points to “electronic communications”: a terminal whose pickups were made but not reported shows up as a data problem, not a service failure. Reasons such as freight not ready would point back to shippers, while repeated misses at one terminal on certain days would point to scheduling. Each carrier would likely see only its own data.


7. Sources

  1. C.H. Robinson — C.H. Robinson launches AI agents to combat industrywide problem of missed LTL pickups (press release)
    https://www.chrobinson.com/en-us/about-us/newsroom/press-releases/2026/chrobinson-lean-ai-agents-for-ltl-pickup-efficiency/
    2026-01-26
  2. C.H. Robinson — Q4 2025 earnings call, transcript via AllMind
    https://allmind.ai/earnings/event/2857283
    2026-01-28