Marisol runs an eleven-truck refrigerated fleet based in northern Illinois. Regional lanes across the Upper Midwest, primarily food distribution and pharmaceutical cold chain. Fourteen-year history. She started with two trucks that her late husband had operated as owner-operators, and she has built the fleet up one unit at a time since 2018. Twenty-two people on the payroll counting drivers, dispatch, maintenance, and the accounting side. Revenue in 2025 was $6.4 million.
In late 2025 a fleet software vendor came in with an AI voice agent package. The pitch was aggressive. The AI voice agent would answer inbound dispatch calls from brokers, negotiate rates in real time using historical pricing data, book loads, and update the dispatch board without human involvement. Forty-two thousand dollars a year. The vendor's sales deck showed carriers replacing entire dispatch teams with this tool. Marisol's dispatcher had been with the company for six years and was one of her most trusted employees. She was uncomfortable with the pitch but the numbers looked compelling.
She came to the diagnostic before the sales guy could close her.
What The Diagnostic Found
Return to Owner ran the biomarker set for the transportation subsegment. Marisol's labor productivity was strong. Her Working Capital Days was tight but healthy for the segment. Her Fixed Cost Capacity coverage was excellent. Her fleet utilization was where the flag went up.
Utilization on Marisol's eleven trucks was 71 percent against a segment-standard target of 82 percent for regional refrigerated. Eleven percent of the fleet's earning capacity was sitting empty. Deadhead miles ran 22 percent of total miles against a segment standard of 15 percent. Every empty mile was a direct cost against no revenue. Every underutilized truck was fixed cost being consumed against no revenue.
Convert the numbers. Eleven percent underutilization on an eleven-truck fleet is roughly 1.2 truck-equivalents of empty capacity. At an average revenue per truck of $580,000 a year in her regional refrigerated segment, that is $696,000 a year of revenue Marisol was not booking because her dispatcher and her routing decisions were leaving miles empty and units idle.
The AI voice agent the vendor was selling would have done nothing about that. Voice agents answer phones. Marisol's phone was already being answered by her dispatcher, competently, at industry-standard cost. Her dispatcher was not the bottleneck. Her load matching and route optimization was the bottleneck, and the AI voice agent had no ability to touch either.
The AI That Actually Belonged Here
Regional refrigerated fleets have a documented AI use case that produces measurable utilization gains. Load matching against carrier characteristics, driver hours-of-service constraints, and lane profitability. Route optimization against fuel cost, traffic, weather-sensitive routing for refrigerated loads, and detention risk at specific receivers. Real-time reassignment when a load cancels, a driver runs into a delay, or a better lane opens up. All of it running as a decision-support layer for the dispatcher, not a replacement for the dispatcher.
Two-thirds of fleet managers reported active AI adoption plans in 2026, and the early adopters were showing 20 to 30 percent efficiency gains on dispatch and routing. For an eleven-truck fleet the size of Marisol's, the utility software costs roughly $19,000 to $28,000 a year plus integration cost with her existing TMS. Less than the AI voice agent, focused on the actual bottleneck, and paired with a dispatcher who kept her job and got better at it.
The doctrine framing came out sharply. AI in trucking is fast pattern matching on load, rate, and sensor data. It is genuinely useful. It is not autonomous judgment. The routing math is a pattern matching problem. AI wins that one. The Friday afternoon call from a difficult broker who is trying to renegotiate a rate mid-transit is a human judgment problem. Dispatcher wins that one. Buy the tool that solves the routing math. Keep the dispatcher for the human judgment.
What Happened After The Purchase
Marisol bought the dispatch and routing AI package in November 2025 for $24,000 a year plus a one-time $6,200 integration fee against her existing TMS. Deployment took six weeks. By the end of Q1 2026, the fleet utilization number had moved from 71 percent to 76 percent. By the end of Q2, it was 79 percent. By August 2026, it was 82 percent, hitting the segment-standard target for the first time in the fleet's history.
Deadhead miles fell from 22 percent of total miles to 17 percent over the same period. Cost per mile dropped 14 percent. Driver retention improved because drivers were making more revenue per shift and spending fewer hours running empty. Marisol's dispatcher, who had been nervous about the AI purchase, became the most enthusiastic advocate for the tool inside the company. Her job shifted from constantly rearranging the load board to managing exceptions and handling the broker conversations the AI could not touch. She said it was the first time in six years she felt like she had time to think.
Revenue on the same eleven trucks moved from $6.4 million in 2025 to a run rate of $7.9 million by mid-2026. Same fleet. Same customer base. Same dispatcher. Different AI purchase.
The Villains The Diagnostic Named
The software vendor. His job was to sell his highest-margin product, which was the AI voice agent package. The dispatch and routing AI was in his catalog too, but at a lower margin, so it did not lead the pitch. Marisol had to specifically ask for the second package after the diagnostic conversation. If she had signed the first proposal in the first meeting, she would have paid $18,000 more for a product that did not fit her business.
The trucking trade press. Every trade publication in late 2025 ran features on AI replacing dispatchers. The narrative was clean and provocative and it drove clicks. It did not accurately describe what AI actually did well in trucking. The real story, which was decision-support software making existing dispatchers meaningfully more productive, did not have the same headline energy. Marisol read the headlines and almost bought the wrong tool because the trade press had already framed the debate.
The peer group at the state trucking association. Two carriers in her regional peer group had bought AI voice agents in the prior six months. Neither had run the biomarker analysis before signing. Both were now discovering that voice agents did not fit their businesses and were quietly looking for exit clauses in their contracts. Neither had told the peer group they were regretting the purchase, because nobody likes to admit a bad decision in front of competitors. Silence in a peer group is not evidence a decision worked.
The Pushback Marisol Almost Made
"Every other carrier is investing in AI voice agents." Some are. Some regret it already and are not talking about it. The peer group data is contaminated by survivorship silence. Do not price a purchase decision off what your competitors are willing to admit they bought. Price it off the biomarker.
"But what if AI voice agents get so good that I lose my competitive edge if I do not buy one?" When the utility of voice agents surpasses the utility of a competent human dispatcher on real broker calls, the tool will be worth buying. Right now it is not, and paying $42,000 a year to be an early adopter on a tool that does not fit your bottleneck is an expensive way to signal innovation. Wait for the utility. Buy the routing AI now, because that utility is already proven.
"My dispatcher will feel threatened." Not by the routing AI. She will feel threatened by the voice agent, correctly, because the voice agent was designed to replace her. The routing AI makes her better. Frame it as her tool, not the company's replacement plan. She becomes the operator of the AI, not the target of it.
The Lesson
Every AI vendor pitching a small business is optimizing for the vendor's product margin, not the business's actual bottleneck. The tool that fits the bottleneck is often not the tool the vendor leads with. It is usually the second product in the catalog, the one with the lower price and the higher fit. You will not find that tool unless you know your own bottleneck before the sales meeting starts.
Marisol's fleet is now running at segment-standard utilization for the first time in fourteen years. Revenue is up 23 percent on the same eleven trucks. Her dispatcher is still there, doing better work than before, using the AI as her own tool. The AI voice agent that was pitched as the future of trucking would have replaced the person who most needed to be kept and left the bottleneck untouched. Cost of running the diagnostic conversation before signing the vendor contract: two hours of Marisol's time. Value of running the diagnostic conversation: roughly $1.5 million in additional annual revenue and a dispatcher who did not have to be recruited.
Note on the field note format. Names and identifying details are composited from actual Aldebert diagnostic engagements. The mechanism, the biomarker sequence, and the outcome pattern are faithful to the underlying case.