As BNSF Railway trains move toward their destinations, a type of artificial intelligence (AI) is also at work—helping identify any potential safety issues with the cars.
“We are analyzing about 3.5 million images a day,” said Asim Ghanchi, assistant vice president of technology services at BNSF Railway. “We have 45-plus models that look for almost 30-plus defects as trains are rolling across our network.”
At C.H. Robinson, North America’s largest freight brokerage firm, its proprietary AI technology significantly shortens the time it takes for customers to receive price quotes.
“Inbound emails go through an email classifier,” said Cody Griggs, vice president of digital brokerage at C.H. Robinson. “The email classifier sends that quote request out, hits our pricing algorithms that we’ve developed over the years, and generates a quote request and response to the customer within 60 seconds.”
Ghanchi and Griggs shared their companies’ applications of AI as part of the 27th Annual CTS Freight and Logistics Symposium: AI-Powered Freight—Revolutionizing Transportation and Logistics, held in December 2024.
The goal of the annual Freight and Logistics Symposium is to bring together decision makers, practitioners, and researchers from different sectors of the freight industry and get them to talk about how Minnesota—and the country as a whole—should advance the industry.
A big question: AI means what?
The symposium offered a framework for thinking about AI from Michael Watson, associate professor at Northwestern University and co-founder of the AI firm Opex Analytics, acquired by Coupa in 2020.
When people talk about AI in different ways, it’s not always clear what they mean, Watson said. He described four AI frameworks:
- Artificial general intelligence (AGI), which involves getting machines to think like people. AGI also covers the concept of singularity—the hypothetical future point at which computers become smarter than people and those computers make the next generation of computers even smarter.
- Deep learning to solve hard problems, such as self-driving vehicles and robotics. These applications entail high-cost investments in the technology.
- Generative AI and large language models such as ChatGPT. Generative AI compiles existing visual and written knowledge to create models.
- Practical AI, which covers a “big tent” of AI, including machine learning. While ChatGPT makes algorithms more accessible, algorithms in turn can teach machines how to learn from data, identify patterns, make predictions, and perform tasks.
“If I think about the freight and logistics field, some of these practical AI algorithms are the ones that are going to have the most impact,” Watson said. “And if your organization is just focused on ChatGPT and not these algorithms, you’re going to be missing out on a lot of technology that can help.”
AI and freight: Making it happen
In addition to safety applications, BNSF Railway uses AI to enhance operations. At several of its large facilities, flying drones capture pictures and videos of parking lots for containers that need to be loaded on railcars. Analysis of that data allows BNSF to match the container with its location. Another AI application helps minimize the distance of a container for loading to the railcar. The jobs of BNSF employees on the ground remain the same, but the AI tools help them do those jobs more efficiently, Ghanchi said.
C.H. Robinson also uses AI to complete the process of creating customer orders and to more quickly generate carrier matches. “Dispatchers typically sit in an office and they’re hitting refresh on multiple brokers’ websites looking for freight,” Griggs said. “We’re helping them find freight about four times faster than what they would be doing on their own.”
As a result, C.H. Robinson has seen gains in productivity, improved accuracy in orders, and growth opportunities from quicker customer responses.
AI implementation: Moving forward with care and intention
When ChatGPT made headlines, companies had many questions about if and how to use the technology. C.H. Robinson rapidly decided to centralize its use of generative AI and carefully look at building potential applications as well as forming a team with technical and business expertise.
Bringing together diversity in such a team is critical to success, Watson said. He recommended that team members include algorithm specialists, data engineers, software engineers, product managers, designers, and leaders. He also suggested the team add a translator who understands the business and the technology in a broad sense to help bridge gaps between the technical and business sides.
BNSF also takes a centralized approach to AI, with technology services heading most of the development and multiple teams working on different projects and problems. Executives, including the CEO, review the top AI initiative list each month. With much interest in AI at the executive level, BNSF has assembled an AI ethics committee, with leaders from law, human resources, and business who examine large use and high-value cases through an ethical lens.
AI and policy: Where does this lead?
Minnesota Representative Steve Elkins spent five years working to pass data privacy legislation, and in 2023, Minnesota became the 19th state to approve a state-level consumer-level privacy act. While other states are struggling to define AI for legislation, Elkins said the data privacy bill provides a framework to protect consumers from profiling—using technology to make consequential decisions about insurance, employment, and housing, for example.
Elkins understands the difficulty of regulating fast-moving technology. “The technology changes too quickly, and you can never keep up, so focus on the use cases.”
Minnesota Senator Jordan Rasmusson noted that other related issues, such as cybersecurity, may impact AI policy. “I think AI provides some tools to address [cybersecurity], but it’s also going to make those hackers and criminals smarter than they’ve ever been.”
Industries such as freight that cross state lines also face hurdles when states have differing policies and regulations. To help achieve some consistency, freight industry representatives and legislators should work through national organizations to help develop frameworks for states as well as identify best practices.
And it may pay to develop a regulatory framework for innovative pilots, Elkins said. “A lot of the companies doing the research and development basically won’t come to your state unless there is a well-defined regulatory environment around it.”
By its nature, policy legislation takes time, and AI capabilities and use cases continue to expand, Rasmusson said. New AI tools may not fit well into the often process-driven regulatory framework for industries.
“I think the industries that are going to avoid unnecessary regulatory burden are going to be the ones talking to legislators proactively ahead of time, making sure that the regulatory framework evolves with the business case,” he said.
The Freight and Logistics Symposium was sponsored by CTS in cooperation with the Minnesota Department of Transportation, Minnesota Freight Advisory Committee, Council of Supply Chain Management Professionals Twin Cities Roundtable, and the Metropolitan Council.
—Darlene Gorrill, contributing writer
