How AI in Transportation and Logistics Is Creating Real-World Value

How has the AI conversation with customers changed over the past few years?

A few years ago, AI was somewhat of a novelty. Customers were interested and liked knowing that we were using it, but it was still a peripheral concept. Over time, their questions became more practical, and they started asking how it impacts them. Today, business outcomes are front and center in every conversation.

When people talk about AI in transportation and logistics, what do they usually mean?

AI has been around much longer than ChatGPT, but AI entered the public consciousness when ChatGPT brought generative AI directly to consumers. As a result, when people think of AI, they think of a conversational assistant or the automation of routine tasks with generative AI.

In supply chain, AI has a broader meaning and a wider range of applications. Much of the work delivering value today is based on machine learning, which predates ChatGPT.

Generative AI has a place, especially in agents and automation, but machine learning has enabled organizations to gain insights that were previously difficult to obtain. For example, we can better understand how our drivers are performing, whether they may be likely to leave voluntarily, and ways we can improve retention.

Where are you seeing AI deliver some of the clearest operational benefits today?

Transportation planning and routing is one of the top use cases. Traditional routing technology generally took a deterministic approach. If there were five stops, it assumed the travel time between those stops and the service time at each location would be relatively consistent.

We know transportation does not work that way. There is variability every day. AI gives us a better way to account for that variability and use it to improve decisions. Instead of treating unpredictability as a problem, AI can incorporate it into decision-making, helping us operate more efficiently.

Another growing area is defect prediction in both warehousing and transportation. That could mean identifying a warehouse order with characteristics that make a picking error more likely, or recognizing a transportation event that is more likely to create a service issue and intervening before it occurs.

With those insights, you can take steps to prevent the issue. You can ensure the right picker is assigned to the order or the right driver to the load. That can reduce the labor required while improving the customer experience by ensuring the right product reaches the right customer.

Trailer utilization is another important area. Different people can look at the same trailer and reach different conclusions about how full it is. If AI can provide a more consistent assessment, we can make systemwide decisions that improve trailer utilization. At our scale, missed opportunities to fill trucks mean we are effectively shipping air without generating revenue.

Does that also create sustainability benefits?

Absolutely. If we can cut unnecessary miles or use equipment already in the network more effectively, we consume less fuel and reduce emissions. That matters to us and to many of our customers as well.

Why are data quality and governance so important?

Trust is at the core of AI. A trusted data layer makes AI trustworthy. If you give a person bad information, they are likely to make a bad decision, and AI is no different. You need a data governance framework that supports the accuracy, integrity and consistency of the underlying data along with the right business context. You also need to centralize and unify your data, with mechanisms in place to manage it, monitor it and address issues when something breaks.

AI differs from many technologies companies have worked with in the past because it is probabilistic. You cannot always assume the same input will produce the same output. That makes guardrails, testing and human oversight particularly important. People need to know how to question AI-generated answers and recognize when a human needs to step in.

If people lose trust in an AI system because it produces incorrect answers, rebuilding that trust can be very difficult.

How is Penske Logistics using generative AI or AI agents today?

Track-and-trace is a good example. Many shipment updates can be handled through electronic data interchange (EDI), but exceptions still occur when the information isn't available electronically. At the scale of a large logistics operation, those exceptions add up.

Traditionally, people would call dispatch offices to get shipment updates and then enter that information into the system. Today, AI can handle that work. It can call dispatch offices, gather the information and update our systems automatically.

Interestingly, some dispatch operations are also beginning to use AI to answer those calls. AI agents are communicating with other AI agents, giving us a glimpse of where technology is headed.