Artificial Intelligence
When I was already writing this, a bit of shipping trouble in the Suez Canal caused a re-adjustment. The original introduction was going to point out that shipping logistics has a wider impact than most people consider. Now people are well aware of the supply chain issues that only a few days of delay can cause. Yet, there’s still more that is involved in shipping. Sanctions are a tool used by multiple nations. Multinational shipping organizations and their customers can be put in financial and legal risk due to the political environment. This is another area where artificial intelligence (AI) can be part of a solution to mitigate business risk.
We’ll get to sanctions in just a minute, but let us start with a simple, US focused shipping challenge. The Jones Act is a regulation requiring items shipped from one US port to another to be transported on ships built, owned and operated in the US. This drives up costs in Hawaii, Alaska, and Puerto Rico, as the supply of ships that meet the requirements is very small. Companies shipping products need to be aware the risks of confiscation from using an incorrectly registered vessel.
Sanctions make things even more complex, as each nation has different sets of sanctions. A multinational company has to be aware of the many ways in which it must be compliant to continue to ship to markets. Making things even more complicated, the way ships are owned and operated means it is often hard to track which vessels are safe to use. “The first ship database – the Lloyd’s Register of Ships – dates back to 1764, and since then the approach and coverage has changed at an incredibly slow pace,” said Ami Daniel, CEO & Co-founder, Windward. “Today’s shipping environment is much more complex, so organizations in the maritime and trade ecosystem need to rely on solutions that bring clarity by using AI and decision support.”
Recent changes to shipping regulations have changed some sanctions laws from using a blacklist to using a whitelist. Rather than rule out a shipping company, as ship ownership is very complex, the rules involve ships applying to be on whitelists – on a ship-by-ship basis. While this is more accurate, it adds another level of complexity in shipping. The start to any system must to accurately identify each ship in use.
There are more than 45,000 freight forwarders and millions of shippers in the global market. While major oil & gas companies ship enormous tonnage, and are heavily at risk, everyone involved in shipping needs to better track shipments.
Deceptive shipping practices are one area of fraud that the industry needs to fight. One key type is ship-to-ship transfers. There are perfectly legitimate reasons to do that, especially in the oil & gas industry and in other bulk commodity sectors. This can be for a simple reason such as moving product from a ship too large to enter a port to a smaller ship than can land the product. However, such transfers can also be used to circumvent sanctions.
Visual imagery, radio frequency (RF) communications, shipping ownership information, whitelists, and other variables can be used to analyze ship-to-ship transfers, but many are happening around the world and a lot of features must be rapidly analyzed. AI can help with the task. Windward is using deep learning (DL) modules to model vessels behaviors, pattern changes and surface recommendations for interested parties in order to check for compliance to laws including sanctions.”
Other risks, such as grounding, bad weather, breakdowns, and more, can be analyzed to help understand other logistical challenges impacting global shipping.
To wrap up with the usual soapbox about AI being a tool, deep learning is only part of the system. Much of the analysis is done using well known rules, and those rules can be used to help explain the results of DL analysis. “Deep learning is only part of the solution,” said Ami. “We provide a ‘glass box’ approach in our decision support system, not a ‘black box.’ The key difference is about explainability of the models – an issue that has long challenged machine learning and deep learning models. Customers need to better understand how our recommendations are made in order to trust the model, take action, and communicate their decision to counterparties."
Shipping has grown more complex over the last century, and not just in logistics. National and international regulations, including fluid rules such as sanctions, must be combined with opaque ownership chains and many other challenges. Shippers and their customers must be able to know where product are and when they will get to where they should go. They must also be able to show to governments that the ships and cargo are in compliance with law. Artificial intelligence is being used to address those challenges.
