AI in Action: Manufacturing & Distribution

This is the fourth in a series of articles highlighting the many applications of Artificial Intelligence.

Manufacturing and Distribution, the supply chain, does not always get as much publicity as the latest technology trend or the newest financial vehicle threatening to bring the global economy to its knees. Nonetheless, it remains the backbone of the economy. Some of the sexiest firms out there are in this sector — Tesla is a manufacturer, Amazon is a distributor. Manufacturing and distribution is also a hotbed of AI development. There are innumerable points along the supply chain where Artificial Intelligence can add value. Here are a few examples:

Predictive Maintenance

Twenty years ago I was working with a predictive modeling firm that had a proprietary neural network solution. One of our clients was a global semiconductor manufacturer. They told us that every hour of downtime in their chip fabrication process cost them $250,000. This was an expensive, highly complex manufacturing system. On the opposite end of the spectrum, I met an engineer with a fastener manufacturer at a trade show. He told me that they sold their fasteners for only a tenth of a cent, but that they produced them “faster than a machine gun can fire”. It wasn’t $250,000, but over the course of an hour, they ended up producing a an amount of inventory of significant value.

Regardless of the type of manufacturing you’re engaged in, downtime is the bane of your existence. For decades, manufacturing systems have been loaded with sensors to monitor the state of the equipment. With the coming of the “internet of things”, the sophistication of these sensors, and their ability to aid in diagnostics through pulling data from the web, has increased exponentially.

AI-driven predictive modeling in process control is moving beyond “signal health” into the more complex “system health”. System health is about pattern recognition — there may be group of sensor readings that serve as a precursor to an event that historically has led to a downtime incident. If a neural network can recognize this pattern, maintenance can be scheduled which will either minimize or eliminate that downtime.

System health is about pattern recognition — there may be a group of sensor readings that serve as a precursor to an event that historically has led to a downtime incident.

Machine learning solutions like neural networks are well suited to this type of predictive modeling. With the significant costs associated with manufacturing downtime, AI can have a massive impact on efficiency, and ultimately profitability.

Sales & Operations Planning (S&OP)

Sales & Operations Planning is an arcane admixture of art and science, a field for prognosticators charged with the thankless job of matching sales projections with manufacturing output. Being able to understand the ebbs and flows of your supply and demand is central to a sustainable manufacturing operation and effective supply chain management.

S&OP is fairly complex to engage in if you have one product. For most businesses, there are many skus, in some cases thousands. You don’t have to be a brain surgeon or a rocket scientist to run these multivariate predictive models, but it doesn’t hurt.

AI is here to help. Neural networks can be very effective when it comes to these complex systems. A well trained neural network has the potential to deliver better results than the traditional statistical techniques like multiple regression. Fine tuning supply to meet demand can lead to reducing the carrying cost of inventory. This type of improved planning and forecasting can lead to millions of dollars of savings through the whole supply chain.

Repetitive Tasks

A strong supply chain relies on a group of manufacturers, partners, distributors, and customers working closely together. There is a great deal of communication necessary to keep products flowing in a timely fashion to the right places. The type of communication that’s necessary often revolves around queries and workflows. Queries like:

• “What’s the status of my order?”
• “Please send me a copy of my sales receipt”
• “Please issue me an RMA number so I can return this order”
• “I want to cancel my order”

These types of repetitive questions clog up call centers, often taking up over 50% of the call center’s available time. Luckily, they also lend themselves to natural language processing — chat technology. A chat agent can be deployed on an extranet which allows your supply chain members to submit their queries. The query can trigger a workflow within the chat agent. The agent will ask questions of the supply chain member in order to collect the information it needs to resolve the inquiry. Without any intervention, order status can be updated, an RMA number can be issued, a receipt can be emailed — queries can be resolved.

Moving these repetitive tasks and workflows away from your contact center to a self-service conversational agent frees your human agents to spend more time on the more challenging issues your supply chain has to deal with every day. It also gives your members availability to 24/7 service. This new found efficiency and improved customer experience will strengthen your supply chain community and ultimately improve profitability.

These are just a few examples of how an investment in AI can pay off handsomely for manufacturing & distribution companies. Wherever you may reside in the supply chain, you should be investigating the value of this new AI technology.

AI in Action: Healthcare

AI in Action: Financial Services

AI in Action: Entertainment and Media

Ken Tucker is a business consultant specializing in AI and Analytics.