A version of this article was originally published in Progressive Railroading.
Trucking often wins over rail before freight is even loaded. A shipper considering rail may have to spend days or weeks figuring out who to call, waiting for a quote, working through internal railroad handoffs, and then learning new systems and processes just to transact. In many cases, the truck alternative can be priced and booked before the rail option is fully understood.
According to Oliver Wyman's Bridging the Rail-Shipper Gap research, freight customers consistently point to four areas where they want rail to improve: responsiveness, reliability, pricing accessibility, and ease of doing business.
These are not new findings. What is new: agentic AI may offer a way to close this gap. Better still, it can do so quickly, cost-effectively, and at scale. While the trucking industry has gained a competitive advantage using it, rail has yet to come on board.
The impact of agentic AI in trucking has been impressive
Agentic AI is already reshaping how shippers interact with trucking providers. J.B. Hunt, for example, has automated booking for most of its highway and intermodal shipments, saving customer experience teams thousands of hours per quarter. In 2026, it is deploying agentic AI across all of its business units to spot appointment problems, reschedule them, and provide updates.
Similarly, C.H. Robinson runs hundreds of connected AI agents across the shipment lifecycle, helping truckload shipments move through the order/appointment workflow faster and increasing on-time pickups by double digits. Uber Freight runs more than 30 agents automating procurement, tracking, payments, and analytics, and expects AI to cut customer support wait times from 5 minutes to 30 seconds. These are not pilots, but production systems handling ordinary transactions that used to require a phone call.
How agentic AI is different from prior technology
Both rail and trucking have continually advanced their use of analytics and automation technology over the past two decades: traditional automation that follows predefined rules; machine learning that predicts if a shipment is likely to be late; generative AI that can explain a delay in plain language and respond to a customer’s questions.
Agentic AI, however, offers a more holistic approach. For example, it can identify a shipment risk, pull together the relevant information, coordinate an internal response, alert the customer, and recommend or initiate next steps (before the customer needs to call, press zero, and yell “representative” into the phone) — and all without human intervention. It is this ability to understand intent and respond autonomously that makes this generation of customer-facing technology different. Instead of another digital interface for the customer to navigate, agentic AI can increasingly understand the customer’s objectives and navigate complexity for them.
Why railroads need customer-facing agentic AI
Even when rail is economically competitive, the difficulty of doing business with railroads remains one of the most consistent reasons that shippers choose to stay with truck. Agentic AI has the potential to reverse that equation, without requiring the industry to rebuild itself from the ground up.
Railroads have already been using AI for years and using it well. For example, BNSF uses AI to analyze more than 35 million wayside detector readings a day to predict maintenance needs. Norfolk Southern uses AI and high-speed imaging to find minute wheel defects and maintains a real-time digital twin of its network. Others use AI in drone surveys of track and in fuel optimization. But customer-facing agentic AI is still the exception rather than the norm.
Building a better customer experience is associated with intensive training costs for employees. Agentic AI, however, can make responsive customer experience scalable without adding costs at the same pace. This frees up customer service teams to solve more complex problems and build the kind of customer relationships that drive growth — rather than chasing information and handling routine transactions.
Railroads can unlock agentic AI with the data they already have
Railroads already have much of the data they need to power customer-facing AI agents, including shipment records, network performance data, pricing parameters, and customer history. Individual railroads can start creating value with that data now.
The key is to use such early deployment to improve and connect underlying data in ways that are reusable across use cases, rather than creating new data silos. Work will still be needed to align, normalize, and govern data across customer and operational systems, but all that can happen in parallel with targeted deployment. Oliver Wyman Forum research across transportation and industrial sectors has come to the same conclusion: data readiness grows through use, not in preparation for it.
Other industries are beginning to show what this can look like in practice. In financial services, agentic workflows have improved both client retention and productivity by automating and simplifying customer interactions. In another example, consumers can compare and purchase insurance across dozens of carriers through a single AI-enabled interaction.
Railroads can adopt agentic AI without waiting for industry standards
There is an industry-wide dimension to agentic AI that needs consideration as well. Rail is more standardized than it gets credit for: EDI messages, car location messages, location and carrier codes, and equipment registries work across more than 600 railroads in North America. The industry also has standardized handoffs between railroads.
The customer-facing layer, however, is far less consistent. The same underlying event can be described, interpreted, or presented differently from one railroad to the next. Constructive placement is one example; dwell is another. Even when underlying movement data is standardized, customers may still encounter different terminology, event definitions, reporting conventions, and processes depending on which railroad is handling a shipment.
A market even exists to address this fragmentation: shippers pay third-party platforms to merge data feeds, translate railroad-specific language into one set of standardized terms, recalculate arrival estimates, and create a more unified view of shipments moving across multiple carriers. But while these platforms can help make sense of information across rail networks, they largely translate fragmentation, rather than eliminate it. Those parts of the customer experience that most influence whether future freight moves by rail still must be addressed by the railroads themselves.
What this means is that no railroad needs an industry agreement to improve its own customer experience through the use of agentic AI — and none should wait for it. Simply putting a dedicated agent behind each account — one that knows the customer’s freight and can quote a move, manage an invoice dispute through to resolution, or flag a delay and put options in front of the customer before they ask — will help improve customer perceptions of rail.
Creating common definitions then becomes work that can progress in parallel, rather than a precondition. This has been done once already: since 2021, federal rules have required Class I railroads to make certain information on demurrage invoices available in machine-readable form, which is precisely the kind of input an AI agent needs.
How agentic AI can help railroads win back freight
Agentic AI can enable railroads to improve the customer experience without giving up the cost discipline the industry and its investors require. Near-term gains will come from making the transactions shippers find the most difficult faster, easier, and more proactive.
Rail already wins on cost, capacity, and carbon footprint. But the industry is giving away freight it should be winning on experience alone. The first railroad that can provide answers before a shipper figures out who to call will find out how much of that freight comes back.
Read the original article in Progressive Railroading.