Logistics Costs Fell to $2.4 Trillion. The Rekeying Did Not
U.S. business logistics costs came in at $2.4 trillion in 2025, or 7.8% of GDP, according to the Council of Supply Chain Management Professionals' 37th annual State of Logistics Report, produced by Kearney and presented by Penske Logistics. That is down from $2.6 trillion and 8.7% a year earlier. The same report treats volatility as a lasting feature of the operating environment rather than a temporary disruption, and lists accelerating digital and automation ROI among its strategic implications. A lower national bill has not removed the manual work inside individual operations. A freight quote still arrives by email. A carrier update still lands in a shared inbox. Someone still types the same order into the ERP, the TMS and the WMS.
In its guide to logistics automation, integration platform company Jitterbit makes the case that the cost of that work rarely appears as a line item. It appears as a container held at port over an error on a commercial invoice, an invoice that does not match the inventory count, or a delivery date given to a customer that nobody can confirm. Nearly nine in ten supply chain leaders reported major disruptions in McKinsey's 2024 supply chain risk survey. When disruption is routine, the handoffs that depend on a person retyping data become the slowest part of the response.
The Work Happens Between Systems
Manual effort in logistics tends to collect in the same four places. Freight quoting and scheduling still run on spreadsheets, phone calls and email chains, which makes it hard to compare options quickly once shipment volume grows. Vendor communication across suppliers, carriers and brokers moves through long threads and disconnected portals, where updates get buried and stakeholders end up working from different information. ERP, TMS and WMS records are updated by hand, often with the same data entered more than once. Shipping labels, customs forms and commercial invoices are prepared manually, and a single mistake on one of them can hold a container for days and add storage charges on top of the delay.
Each of these is a repetitive, rules-based task. Each fails the same way. A detail is missed or mistyped in one system, and the next system trusts it. The error then moves outward to finance, operations and customer service, each of which has to find it and fix it separately.
Accuracy Comes Before Speed
The first return from automating these steps is accuracy. Software that handles rules-based work does not mistype an order number or skip a field on a bill of lading, which means fewer compliance problems and more confidence in the data that planners act on. The second is visibility. When ERP, TMS, WMS and CRM platforms exchange data in real time, inventory levels, order status and carrier performance come from one source rather than several partial ones.
Cost and capacity follow from there. Hours spent on duplicate entry and problem resolution come off the operating budget. Customers get reliable delivery dates and proactive updates instead of corrections. During seasonal peaks, an automated operation can absorb higher order volume without a matching increase in headcount. In practice, this looks like replenishment orders that trigger automatically when stock falls below a threshold, shipping documents generated in seconds for complex import and export moves, and sales orders that pass from confirmed to picked, packed and shipped without a person moving them between screens.
AI Is Already Reading the Paperwork
Artificial intelligence is the latest technology to arrive in logistics with large promises attached, following robotics, blockchain and digital twins. Jitterbit's assessment is that it is both overhyped and already useful. AI is not about to replace the workforce or drive trucks across the country on its own. It is, however, extracting data from bills of lading and invoices, populating systems and flagging anomalies in seconds. Image models trained on large volumes of cargo photos can identify damage, read handling labels and count stacked boxes, which reduces rejected shipments and delays at ports. Routing models weigh weather, fuel costs and carrier capacity to cut empty miles, and sensor data fed into predictive models can flag vehicles and warehouse equipment before they fail.
The distinction that matters is between automation and intelligent automation. Automation handles fixed, rules-based steps. AI adds a layer that learns, adapts and recommends. Some applications remain further out: autonomous long-haul trucking, autonomously operated heavy equipment at ports and yards, and warehouses run entirely by robots all depend on advances in safety standards, regulation and capital investment. What the near-term and long-term applications share is a dependency on connected systems. AI cannot act on data it cannot reach, and robotics cannot coordinate with an ERP that is updated a day late.
Start With One Handoff
Treating automation in logistics as a single overhaul tends to stall. Jitterbit recommends a narrower sequence. Start by identifying where delays and errors pile up, whether that is manual data entry, an ERP and TMS that do not share records, or shipping paperwork that requires repeated back and forth with carriers. Pick one of those processes, such as generating shipping documents or syncing orders between systems, measure the result and expand from there.
The third step is the one most often skipped. Teams need to understand how the automated workflow supports their work and who manages it when an exception appears. Training, joint ownership between IT and operations, and visible early wins build the trust that a second and third project depend on. The aim is not fewer people on the dock or in the planning office. It is fewer hours spent retyping data that already exists somewhere else.
