Artificial intelligence and machine learning have already transformed the supply chain in countless ways, and the AI and ML tools you’re familiar with will continue to evolve. But generative AI based on large language models (LLMs) will completely reshape the industry — and the broader economy — in ways we’re only beginning to understand.

Supply chain executives who keep up with fast-moving developments in AI applications will find new opportunities to transform their operations. Those who don’t remain at risk of being outmaneuvered by more efficient and nimble competitors.    

Keeping up won’t be easy because of how quickly the technology is evolving. A brief history tells the tale: In 2013, Google introduced Word2Vec, a breakthrough technique in natural language processing using neural network models. In 2018, Google followed with BERT with 110 million parameters, and OpenAI with GPT (which stands for generative pre-trained transformer) using 117 million parameters. (A parameter is a value that defines the behavior of a machine-learning model.) In 2019, GPT-2 was released with 1.5 billion parameters; then just one year later, GPT-3 debuted with 175 billion parameters. So while it’s critical to know what generative AI technology can do for the supply chain today, keep in mind that tomorrow’s conversation will be quite different.

Generative AI applications like ChatGPT are focused on language. Yet data is king at tech-forward logistics companies, which use it to gain insights for real-time decision-making.

You’re probably familiar with AI and ML tools that are currently used to optimize the supply chain, including routing and dispatching tools based on operations research. Pricing and matching recommendation models are also widely used to connect loads to carriers according to shipper requirements. Such technologies also support the sales process in a number of ways.

Advanced forecasting tools are another example of current AI and ML applications, and they’re improving steadily as data scientists gain new insight. All of these applications will continue to add value, but it’s worth noting that today’s problem-solving models are in a different category than generative AI. Both categories will continue to evolve, but generative AI, which can create new text, images, video, audio and code, is evolving much more rapidly. 

Generative AI is already making inroads into the supply chain; expect it to accelerate further because of the dynamic and challenging nature of the freight industry. Numerous variables affect delivery, including weather and available capacity, so optimization entails the sending and receiving of massive amounts of information. To a certain extent, applications already on the market are doing that work now, such as advanced tracking. But communication or coordination that requires less human touch can be automated further to enable faster responses. 

Sales intelligence in the logistics space is also ripe for disruption via generative AI, which helps salespeople zero in on potential client needs with much greater specificity than standard applications. With better sales intelligence, they can offer highly personalized solutions that solve real business problems.  

In the not-too-distant future, we’ll likely see systems that are capable of consuming massive amounts of documentation and other information about business processes to provide a conversational interface. This “copilot” model would act as an assistant, providing recommendations on what to do, or whom to contact to resolve problems. 

Models like the copilot exist today, but they aren’t as advanced as they will be when generative AI is applied. Even then, they won’t work perfectly every time. Accuracy and reliability issues in LLMs like ChatGPT are a widely recognized problem, but because the models are complex, there isn’t a simple solution. 

For most models, data scientists conduct statistical analyses to determine their level of confidence in the models’ accuracy. Data scientists use similar techniques when working with generative AI, but those tools are relatively new, and the difficulty level in the analysis is high due to a wide distribution of potential outcomes, which can include an inaccurate answer. 

Of course, that possibility is ever-present in human intelligence, too. But risk levels make it critical for supply chain leaders to clearly define when to use generative AI to make decisions, and when a human must be in the loop. For example, anything related to safety should always involve humans. Generative AI can help, but in some cases a human must make final decisions because the risks are too great otherwise. 

In the near future, supply chain leaders will have to make decisions about incorporating generative AI into their operations, weighing tradeoffs like accuracy rates against response speed. In all likelihood, the technology will debut in the supply chain in the form of advanced tools that logistics providers, shippers and carriers use to drive efficiency internally. After that, we’ll probably see advanced customer-facing interactions that might blend generative AI capabilities in text, video and audio formats. 

That said, given the rapid advances in generative AI, expect advances to move fast. In a year’s time, the technology’s capabilities won’t just be 5% better than they are now — they’ll be 50% or even 100% better. This is an exciting time, but it requires proactive participation to realize the possibilities. If you keep abreast of AI developments and harness emerging technologies to drive efficiency, you’ll prosper in the years ahead. If you don’t, you’ll be left behind — so now is the time to get ready for what’s next. 

Alex Schwarm is head of data at Arrive Logistics.