From Hype to Reality: Practical AI Applications Boosting Logistics Performance

Over the decades, analysts, marketers, and software companies have told us about “next-generation, paradigm-shifting technologies poised to disrupt conventions while delivering ROI and competitive advantage by seamlessly leveraging synergies across the value chain.” Ay, caramba!

Fact is that transport and logistics folks are a dubious bunch. Our skepticism may be an innate trait that self-selected us into our chosen profession, or perhaps it is a learned behavior. I tend to side with the latter proposition. We don’t believe 3PLs, carriers, and brokers when they make extraordinary claims around service or cost, and we certainly don’t believe self-described supply chain “experts.”

Examples of overhyped technologies that we were told would be existential threats to our businesses and careers if we didn’t take immediate action include:

• Extensible markup language (XML) as a replacement to EDI in the 1990s,
• RFID to eliminate manual scanning in the 2000s,
• Blockchain / IoT for smart contracts and visibility in the 2010s.

While one can argue that some of this technology is in use, the “paradigm-shifting” promises have not yet come close to meeting initial expectations.

Is It Different This Time?

This brings us to Artificial Intelligence, the current la mode du jour.

The skeptics in us say, “Here we go again.” However, I would argue that this time really is different. Sure, the carnival barkers at analyst firms, consultancies, and software companies overhype the technology — instilling FOMO into their clients is a critical aspect of their jobs. But unlike the previous examples, AI differs in 3 crucial ways:

• Most of us are already using AI. It has become indispensable to how we go about our day. Capabilities as mundane as grammar and spell checks in Microsoft Word and Netflix/YouTube video suggestions are common examples. More interesting and relevant advancements include robotics, self-driving vehicles, video/image recognition, and natural language processing.
• AI is not a binary technology that is either used or not. It is a broad spectrum of technologies that span an infinite and varied set of use cases.
• Advancements in the underlying AI technologies are not contingent upon supply chain use cases. These technologies will materialize and advance regardless of whether we, as supply chain professionals, adopt them or not.

Our well-earned skepticism has left many of us overly jaded and quick to dismiss AI as just the latest fad. This is a mistake, one that is only exacerbated by the purveyors of log-tech who are often woefully uninformed about the real value of AI technologies they are peddling and how they can be most effectively used by practitioners.

At JBF Consulting, we see AI and the underlying technologies not as a system and solution in and of itself but instead as a means to enhance existing processes and increase employee productivity. Examples include:

• Using Generative AI for code development. I used ChatGPT 4o to write Python code that converts an Oracle TMS JSON file to an ANSI X12 EDI 204 message. It took me less than 2 minutes. We have also begun experimenting with GenAI for test script development and other artifacts developed during our implementations.
• Using Natural Language Processing (NLP) for general queries and reporting. TMS systems are notoriously poor at customized reporting. NLP technology is now being embedded in TMS applications to eliminate the need for IT involvement in report generation. TMS providers, including Loadsmart and Pando, have built strong capabilities in these areas. NLP technology is also being deployed as Virtual Assistants for both operational and training purposes.
• Using Machine Learning to increase model/forecast accuracy. TMS and Real-Time Visibility providers started using ML to make more accurate predictive ETA calculations. Now, we are seeing progress made in new and more important areas. An area we are keen to see developed is ML being used to combine internally managed data (historical shipment information, rates, invoice information) with external data (market rates, network congestion, weather conditions) when developing models.
• Using Machine Learning to validate data. ML can quickly identify and even correct anomalies, including data outliers, missing data, and incorrect data, within large data sets.

Summary

Skepticism is a maligned trait. A healthy dose is required to be successful in life. For example, I should have been more skeptical when a Nigerian prince told me that I had great riches coming if only I sent him a few thousand dollars.

However, being overly skeptical can lead to missed opportunities.

AI is real, and it is changing how we work. Educate yourself on the various technologies that make up AI — there are great resources online. Challenge both your company and your logistics technology partners to adopt the technologies in meaningful ways that add value. And know that hype is occasionally warranted.

A version of this article was first published in Talking Logistics