Can I get that yesterday?
As digitization and automation reshape how cargo is tracked and moved throughout the supply chain, shipper demands remain firmly rooted in speed, visibility and cost. It comes as no surprise, then, that many of the machine-learning applications in operation or development today focus on delivering these basic shipper requests.
Freight forwarders have already shown an interest in machine learning technologies, as they are easily drawn to the appeal of reduced shipment times, and intelligent consolidations. Although some forwarders are experimenting on their own, most are reliant on third-party developers to provide solutions.
Panalpina, for example, recently began using U.S.-based predictive logistics company ClearMetal’s Data Intelligence Platform to optimize booking allocation, and predict customer behavior across its ocean freight business. While it is still too early to evaluate the implementation of that program, a spokesman from ClearMetal hinted that predictive analytics for airfreight forwarding were in the pipeline.
“As a company focused on enabling the artificial intelligence supply chain and end-to-end predictive visibility, it’s absolutely where we’re going.”
Returning to Elementum, the company’s core offering — what it refers to as a “product graph” — is, in essence, a logistics platform that integrates data from all links in the supply chain in a holistic manner. Rather than viewing data flowing between different links in the chain as transactions between two parties (supplier-to-carrier, carrier-to-retailer), it visualizes the supply chain as being interconnected. At present, multiple thematic applications sit on top of the graph, and each app contains unique code and architecture with machine learning components that focus on specific problems.
In practice, this means that, even though the company is not yet working directly with airfreight carriers, it has long been capturing data on airfreight movements from existing users of the graph. If a manufacturer regularly moves goods along a certain trade route — the graph is fed with a mix of data about forwarders managing the shipments and the carriers transporting the goods. Elementum plans to use machine learning to leverage this information to create a “calculated ETA” feature.
Calculated ETA uses machine learning to predict estimated arrivals based on aggregate historical information collected from past shipments moving along the targeted trade lane. Rather than simply taking into consideration a published flight schedule, the algorithm takes into consideration “the entire journey, from origin to destination, giving carriers the ability to provide granular solutions,” said Brody. While consignments are in transit, the calculated ETA will shift in the event of any irregularity.
Machine-learning technologies also support a broader shift in how customer service teams respond to exceptions and shipment irregularities. Rather than fielding calls from disgruntled shippers trying to locate their consignments, carriers will see a greater set of tools at their disposal. For Elementum, this means synchronizing the gadgets and capabilities of the tools we use in our personal lives, with those we use in our professional lives – analysis and decision making from mobile devices.
Echoing that concept is U.S.-based Unisys. “Why not replace limited-function handheld scanners with iPhones?” asked Dheeraj Kohli, vice president and global lead of travel and transportation. In doing so, Unisys hopes to link customer service agents directly to its freight management software, called “Digistics,” which is increasingly driven by machine learning technologies.
“What is happening in business analytics is that if predictability, speed and agility are the desired outcome, more and more machine learning will be introduced,” said Kohli. Digistics derives much of its data directly from the airfreight carriers it works with, and is developing route optimization tools that can be layered on top of its core freight management platform.
In May, Unisys launched an online Artificial Intelligence Center of Excellence, which helps users develop capabilities in advanced data analytics. Digistics’ machine learning relies on a “neural network approach” – that is, algorithms which function in a manner similar to the way neurons act in the brain.
By analyzing available route options for cargo, historical flight arrival and departure times, forecasted traffic and weather, and other data, the modeling tool can detect relationships and patterns between variables. As the algorithm collects data, its predictions become increasing-ly accurate. Eventually, when a forwarder goes to book a shipment, the tool will suggest an optimal route based on the probability of delay or exception.
Although it may seem as if artificial intelligence seeks to remove the human element, in the near future it will assist and empower operations teams in making better decisions. In case of an exception, like a pallet missing from a flight, the Digistics tool will alert the carrier of its options. Then, as soon as a decision is made, instructions are transmitted to ground handlers, and an option to inform the shipper is provided.