Global News Generative AI likely to be impactful in logistics, just not in a public-facing way

Logistics technology experts say the immediate impact of so-called large language models (LLMs) will be to accelerate software development and product iteration, while the longer-term impact will be a reshaping of how shippers manage the workflows underpinning their international freight moves.

Nearly six months after ChatGPT, the most famous LLM, brought the potential of artificial intelligence (AI) into clearer focus for non-technical people, there remains some debate over how significant large language models are in pure business-to-business use cases.

Part of the skepticism around such tools, also known as generative AI, is that the logistics industry has been burned in recent years by overhyped technologies that didn’t necessarily yield tangible progress. The most notable of those overhyped concepts was blockchain, which came to prominence in business circles in 2018, but has yet to find a transformative application in global logistics.

Some logistics technology experts believe LLMs will change the way people in the industry interact with software. But others believe the industry is far from a reality where a logistics manager merely prompts a system to autonomously execute a shipment of goods across continents, as a consumer might prompt a smart home device to adjust the temperature in his or her house.

Two clear patterns of impact around generative AI are, however, clearly emerging in logistics. First, the idea that industry-specific LLMs will be far more valuable than general purpose ones, such as ChatGPT. And second, that the immediate value of LLMs will be leveraged by software engineers and programmers that understand how to prompt such systems better than non-experts.

The effectiveness of prompts — the art of asking an LLM the right questions to get the desired answers — is at the heart of whether generative AI can reach the potential many believe it has in the logistics world.

“The most relevant use cases today that you see of AI are in customer support and customer service, with so much time spent emailing and answering phone calls,” Michael Wax, CEO of forwarder Forto, told the Journal of Commerce. “So just imagine how much time can be spent on resolving exceptions, making sure that clients are being treated in a much wider agenda and in a much more proactive fashion. I see tons and tons of potential there that I feel is still, today, very much undiscovered.” Years in the making To be clear, LLMs are no novelty. Software providers in the logistics industry have, for years, yearned to build tools that make it easier for users to extract information from ever-increasing pools of data, or to execute transportation management commands on existing legacy systems.

Some companies are even designed purely around the creation of logistics-specific chatbots, which leverage LLMs to enable machines to “converse” with humans.

“The idea is to have a bot get you a shipping quote, track a shipment, book a shipment, responding back in real time,” said Matt Motsick, CEO of Rippey.ai, which builds logistics-specific bots for forwarders based on a proprietary LLM it has built.

“With ChatGPT, you’re querying the internet, but it won’t answer business-related queries,” Motsick said. “You can’t get a shipping quote from ChatGPT. ChatGPT is great as an internal tool and as a research tool. But for a customer tool, ChatGPT alone won’t work.”

Rippey.ai is an example of LLMs interfacing directly with users that aren’t AI experts. The goal, Motsick said, is to use LLMs to make interactions between businesses more efficient. After selling widely used freight rate management software vendor Catapult (now part of forwarding software vendor Magaya), Motsick founded Rippey.ai, formerly called RPA Labs, in 2017. He said logistics users should think of LLMs as “logistics language models.”

“It’s trained on terms like FCL, LCL, CBM, so the chatbot knows exactly what the customer is looking for,” he said. “We spent four years building the logistics language model.”

If Rippey.ai’s bots are an example of how LLMs are already assisting software users, some in the industry believe the biggest gains are to be made by programmers that know what they’re doing. The theory is that LLMs will accelerate software development by essentially expanding the capacity of engineer teams and enabling them to iterate products faster, because LLMs can also include coding languages, not just human languages. Expanding development capacity An executive at a global transportation and procurement management software provider responsible for 100 software engineers said LLMs are having exactly that impact.

“It's hard to pin down, but it's probably 10 to 30% more velocity per person,” the executive, who did not want to be identified, told the Journal of Commerce. “It's like getting around 20 engineers for $1,000 per year per person.”

“I don't think there’s been a genuine paradigm shift like this in my lifetime since cloud computing or maybe the arrival of smartphones,” the executive said. “That kind of economics simply has to result in massive changes. I think the world wants more code than it can get right now, so it won't lay people off. But we'll find out.”

Expedock, a software provider catering to forwarders, earlier this year built a chatbot-like interface designed to inspect a business’s supply chain data and answer queries in responses understandable to humans.

“The idea was for users not to have to slog through TMSs to get to the shipment data they need,” Expedock CEO King Alandy Dy said. “Personally, we believe that this will be one of the ways that businesses will interact with their supply chain data in the future. There are also many other interesting use cases that these AI models can address.”

Expedock is also experimenting with using LLM in other ways: to improve supply chain forecasting and decision-making to return results in an understandable format; extracting insights from unstructured data sources such as emails, customer reviews and social media posts; and generating human-like text that can be used to automate responses to customer inquiries and improve communication with suppliers.

“My team has some pretty crazy stuff they build for fun on the weekends,” Dy said. “There’s tons to be written about the wave of these new AI models that can drastically improve existing supply chain processes, and the innovation is just beginning.” “A great awakening” John Motley, CEO of logistics software vendor LOG-NET, said generative AI has resulted in a “great awakening to years of progress in AI.”

“ChatGPT and LLMs provided the epiphany of how close massive data sets combined with adaptive algorithms has gotten to human intelligence,” he said.

AI is largely based on computational intelligence, the ability to determine outcomes based on processing larger and larger data sets through increasingly sophisticated algorithms, Motley said. “The common element of AI is that it is now generally accepted that the cognitive intelligence of machines, in some instances, is exceeding human cognitive intelligence,” he said.

Motley, similar to Motsick, cautioned that LLMs without logistics context are of limited value.

“You have to understand the business model deeply,” he said. “Training on human models may just make a creative, fast and what appears to be stupid machine. You need business, algorithmic and AI expertise, not just LLM expertise.”