The logistical challenges faced by international automakers are myriad, thanks to the globalized nature of the supply chain. Parts, tools and materials come from all over the world and must be constantly tracked, resulting in a slow, labour-intensive process that takes its toll on the bottom line.
Today, as automakers seek to leverage artificial intelligence for their self-driving vehicles, so too do they look for ways to use AI in their logistics processes in an effort to streamline operations and reduce costs.
Case in point: MLaaS is currently being used by a leading automaker to track and document the tools used in its manufacturing process. Previously, the client employed an eight-step procurement and documentation process wherein purchasers, suppliers and procurement staff engaged in a time-consuming exchange that had changed little in decades. The result was a system that while accurate, was very slow and ripe for technological disruption.
MLaaS was approached by the client to explore the possibility of introducing AI into this process. After some initial analysis, we identified three steps in which AI could make significant improvement:
The first step is for the supplier to email images of the tool and label to the client prior to shipping the tool. The client must confirm that a label is present in the image.
MLaaS built a model from a dataset of images of tools and labels so that the system can employ optical recognition and confirm the presence of a label in seconds via a mobile app. If the label is not visible, the tool does not ship.
Upon the tool’s arrival, the client verifies that the details in the label are accurate and that it is on the correct tool, and that the label conforms to the client’s data standards.
This is a laborious process that requires cross-referencing the characters on the label to a database, and visually confirming that the label is on the correct tool. This must be done manually, and takes a significant amount of time to perform.
Using deep learning, MLaaS performs optical character recognition to verify the label information, and employs 3D imaging to verify that the tool in question is the correct one.
This is done in seconds, saving hours of labour and catching any errors or anomalies early.
The final challenge involves classifying the tool within the client’s database. The client clusters its tools into three levels, with levels 1 and 2 used as guidelines while level 3 describes the exact tool as part of the database. The was done manually and takes considerable time to complete.
MLaaS performs tool classification quickly and accurately, alerting the client that they’ve got the correct tool, or that a mistake has been made somewhere along the way.
To date, the introduction of MLaaS into the logistics process of the client has resulted in considerably shorter times from start to end. Errors are identified much more quickly, and far less time is spent manually checking databases or poring over images.
Instead of an in-house effort in which the client hired AI experts to build a custom solution from the ground up, MLaaS provided the technology and talent required to achieve the client’s results rapidly, accurately and with considerable cost savings.