The Freight Status Gap

The measurement vacuum

This review examined government labor and transportation data, federally sponsored research, trade-association material, independent logistics reporting, company disclosures, vendor studies and product documentation, and public operator discussions. We found precise measurements for truck cost, dwell, detention, employment and market activity. We did not locate a government agency, trade association or independent study that quantified the core workload of freight check calls.

The wording matters. This is not proof that no brokerage has measured its own work, or that no private dataset exists. It means no public independent benchmark was located across the sources reviewed. A gated Bloomberg Intelligence/Truckstop study disclosed a sample of 187 freight professionals, but its detailed findings were not publicly accessible and were not used to fill the gap.¹⁵

The one hard status-volume signal

C.H. Robinson's disclosure is the strongest public indicator we found of phone-based shipment-status volume. It is unusually concrete: a count, a month and a defined call type. Independent analyst coverage at S&P Global repeated the 318,000 figure, while the original company release remains the direct source.

The missing denominator prevents broader conclusions. The annual shipment figure covers all modes, while the status count covers one call type in one month. Dividing one by the other would create a ratio that the source does not support. The evidence shows scale at one large brokerage—not a national workload rate.

What freight can measure

Detention is supporting evidence, not the subject of this report. Its value is contrast: the industry can measure the downstream consequences of delay with considerable precision while leaving the communication labor surrounding a shipment largely unmeasured.

ATRI reported average dwell of 1 hour 49 minutes per stop in 2025, up from 1 hour 38 minutes in 2024. Its detention research estimated 117–209 detained hours per driver per year, more than 135 million hours across for-hire trucking, $3.6 billion in direct expenses and $11.5 billion in lost productivity in 2023. ATRI also reported an operating cost of $106.69 per truck-hour with fuel in 2025, and that 94.5% of fleets charge detention while fewer than half of those invoices are paid.²³

These are detention and truck-cost measures. They are not the cost of a check call. Their relevance is that precise, defensible timestamps and status context have operational value—yet the work required to obtain and communicate that context remains unbenchmarked.

Conflicting detention figures are not interchangeable

ATRI reported that drivers experienced some detention at 39.3% of stops in its 2023 survey. A separate ATRI publication reported that 9.9% of stops had more than two hours of detention. Federally sponsored telematics research using more than 1.3 million stops found detention beyond two hours at approximately one in ten stops. These values use different definitions and instruments. We preserve them rather than averaging them.

Why status calls still exist

TIA, citing April 2020 FMCSA data, reported 928,647 for-hire carriers and 799,342 private carriers on file, with 97.4% operating fewer than 20 trucks. The data is old and should be labelled as such, but the structural point remains useful: the counterparty is often a small fleet without a clean system integration. TIA's visibility paper also argued that no single provider covers every mode and node.⁴

Public operator discussions add qualitative context, not prevalence. Brokers and drivers describe cadences that vary by customer procedure, disagreement over whether a short call is burdensome, tracking failures, and the need to confirm check-in, breakdowns or appointment problems. We do not convert those comments into percentages. They help explain why verification, accountability and exception context can matter even when GPS is available.

Visibility is not the same as an answer

Coordinates can place a truck near a facility without confirming that it has checked in. A geofence can show arrival without explaining why loading has not begun. A tracking link does not necessarily say whether a revised ETA threatens an appointment or customer commitment.

Vendor documentation is useful here when treated as product documentation rather than market evidence. FleetWorks' published track-and-trace fields include check-in time, loaded time, reefer temperature, waiting-at-facility status and ETA. Tai's documentation describes four workflows, a scheduled weekday run and single-stop scope. These details show what agents are being designed to collect—and where product scope still ends.¹¹¹²

The operational answer is therefore a chain: identify the correct load, establish state, capture the reason and revised expectation, update an authorised system, communicate what affected parties need to know, and give an exception to the person who owns the commercial decision. Location is one input to that chain.

The AI check-call market arrives

At TIA's April 2025 media day, Transport Topics counted about 20 companies announcing new capabilities; more than half highlighted AI functionality or virtual agents. The count describes vendors at one event, not adoption across brokerages.⁵

Capital and deployment signals followed. FreightWaves reported HappyRobot's $150 million Series C at a $1.2 billion post-money valuation. FleetWorks announced $17 million in funding. Transport Topics reported DHL Supply Chain using HappyRobot for routine phone calls, emails, driver follow-ups, warehouse coordination and appointment scheduling. Truckstop launched a voice-native carrier assistant, while Tai and MacroPoint introduced agentic status products. C.H. Robinson said it had deployed more than 30 AI agents across shipping tasks.⁶⁷⁸

These signals answer “is the category real?” They do not answer “what does it save?” Funding is not efficacy. A launch is not an independent outcome study. A broad automation rate is not a check-call rate.

The claim audit

Current vendor claims cannot be combined into an industry benchmark. They describe different customers, workflows, denominators and measurement methods—often without publishing those methods.

The Tai example is especially instructive. Its broad marketing language exceeds the scope visible in its own documentation. That difference is not evidence of fraud; it illustrates why category definitions, product boundaries and measurement standards matter.

What remains human

The evidence does not justify a universal boundary, but it consistently points to situations where judgment and accountability remain important: ambiguous identity or load matching, conflicting status information, unusual multi-stop movements, complex exceptions, appointment renegotiation, sensitive customer commitments and commercial decisions.

Automation is better suited to repeatable collection and permitted follow-through when the record and next action are clear. A person remains important when the facts conflict, policy does not determine the answer, a commitment must be renegotiated or a customer relationship is at risk. Product documentation that limits agents to defined workflows reinforces this boundary more credibly than absolute elimination claims.

What the industry still does not know

The evidence gap is not a footnote. Without common measures, buyers cannot compare products, vendors cannot make comparable claims, and the industry cannot tell whether automation removes work or merely moves it.

What would settle the question

A credible next phase needs primary measurement, not another vendor estimate. The highest-value study would combine a broker time study with status-touch counts: calls, messages, tracking checks, TMS updates, customer notifications, unsuccessful attempts and after-hours events tied to a consistent load definition.

A second study should measure tracking failure and exception context: when telemetry is unavailable, when it is technically present but operationally insufficient, and which status questions require a person. A third should compare before and after automation using the same workflow definitions, observation window and ownership rules. It should report failures and human work transferred—not only successful automated contacts.

Claire Research has prepared a methodology for this primary work. It has not yet collected or published results.

Conclusion

The freight industry has entered a period in which AI agents are being deployed faster than independent measurement standards are being created. The volume signal is real, the market activity is visible, and the downstream cost of delay is well documented. But the basic operating denominator—how much human status work exists, where it occurs and what automation independently changes—remains missing.

That missing benchmark is now a meaningful research problem. Claire Research intends to continue investigating it.