Using AI To Optimise Supply Chain Management

AI is central to many of the most efficient supply chains in the world, with AI-powered systems becoming more and more powerful as larger datasets can be compiled and stored as more advanced technology develops. For example, Shopify, which announced in 2019 that it would be investing $1 billion USD in the creation of a new fulfillment network to rival Amazon, has recently confirmed that AI, machine learning, data and algorithms are at the heart of its new fulfillment network.

 

This article will look at common supply chain problems and how these can be addressed through the use of AI, machine learning and data to produce increasingly efficient supply chain management and strategies.

 

 

 

The supply chain is how a manufacturer moves their products or services to the customer. Supply chain can also refer to how a manufacturer sources raw materials necessary to create products and services, before their supply to the end consumer. The supply chain will typically involve a very complex array of people, activity, data and organisations and these all interact to ensure the products and services are created and moved as efficiently as possible. Most large retailers will have a supply chain strategy in place to ensure their supply chains are functioning as efficiently as possible. Supply chain strategy will typically encompass several elements which all need to interact and work together simultaneously to ensure the overall functioning of the supply chain. Although all supply chains are different, supply chains will typically encompass elements including purchasing, inventory management, sourcing, warehouse management, order fulfillment and production planning. Undue delay in the supply chain can cause losses to a supplier for all sorts of reasons, including orders not being fulfilled, reputational damage to a brand or increased storage or logistical costs.

 

Supply chains are important, as when these are not administered efficiently, costs that manufacturers must pay to manufacture and distribute their products and services can increase. The operation of any supply chain is costly and it is a major consideration in the overall running of a successful business. In terms of the operation of a successful supply chain, both time and cost are very important considerations. The more time it takes for goods or services to move from supplier to end consumer, the more costs are likely to be incurred. Successful supply chains tightly control the overall time and cost in their supply chains. Most large retailers will have a supply chain strategy in place to ensure their supply chains are functioning as efficiently as possible. Supply chain strategy will typically encompass several elements which all need to interact and work together simultaneously to ensure the overall functioning of the supply chain. Although all supply chains are different, supply chains will typically encompass purchasing, inventory management, sourcing, warehouse management, order fulfillment and production planning. AI has been shown to be instrumental in reducing problems associated with the supply chain. Moreover, many retailers are turning to AI to assist their operations, for example in the reduction of undue delay.

 

 

 

Demand forecasting can be extremely beneficial in saving time and money wasted through sub-optimal functioning of a supply chain. Take for example, decisions made about purchasing certain products. Real-time data about actual sales can be compared with purchasing plans and purchasing strategy can be adjusted as a result, in real time. This can, for example, avert a situation where a supplier buys more products than it can actually sell.  

 

 

 

Before they reach where they need to be, products, services and raw materials that are moved within the overall supply chain will typically need to be moved more than once. So, for example a raw material may need to be purchased, stored, moved to a processing plant and then delivered to the customer after it has been used to create an end product. All stages in this process can be costly and variable. Unexpected events can upset the supply chain and lead to additional costs. Brexit is a prime example. Goods moving in and out of the EU may be subject to increased or additional tariffs or taxes and this increases the overall costs of the supply chain. A simpler example is an increase in the cost of fuel, which is needed to supply a fleet of vehicles used to transport goods to customers, or to where customers will buy them. Even the slightest fluctuation in the cost of fuel can add up to massive costs in the overall operation of a supply chain.

 

Movement within supply chains can be greatly assisted by AI when data relating to transportation fleets is analysed systematically. This allows a supplier to make predictions about its fleet of delivery trucks for example, which trucks are more likely to break down and which are most reliable at a given point in time. When a supplier is able to plan and manage a fleet in this way, downtime can be significantly reduced and this has valuable consequential impacts for suppliers such as fewer instances where demand isn’t met due to transportation problems like breakdowns. Like any supply chain efficiency, there are reverberations that “travel” throughout the supply chain, in many cases. In the case of the supply chain that makes accurate predictions about their transportation fleet and reduces fleet downtime as a result, a benefit will also be gained in terms of fewer returns made because of delayed delivery. This all adds up to more profit for suppliers.

 

 

 

Another very typical aspect of a supply chain will be storage. As goods and raw materials move between different stages of a supply chain, they need to be stored and kept safe. The storage stage can be very problematic though, as items can be vulnerable to theft, spoiling or damage merely as a result of remaining in one location for any length of time. While being stored raw materials and products will have to be inspected and included as part of an itinerary. Every single day of storage will result in increased costs in the overall supply chain, so supply chain experts recommend that supply chains move as quickly as possible. This reduces the risk that the items moving in the overall supply chain will be lost, damaged or stolen as well as reducing the actual cost of physical storage, for example in a warehouse. Warehouse storage will also attract cost in terms of utilities required to keep items stored at a given temperature and many require trained maintenance staff to transport and store raw materials or products. Items stored will also need to be insured and this adds to the overall cost of the supply chain. If items enter and reenter the supply chain, for example if a customer buys something which is later returned, this adds costs into the process as the item will need to be checked before it re-enters the supply chain.

 

 

 

Order fulfillment, or being able to deliver a product or service to the end customer is one of the most critical aspects of the supply chain. If this takes too long, a customer may decide to place their order with an alternative supplier. If delivery of an item takes too long and this is not what was initially agreed between the supplier and the customer when an order was placed, the supplier may face negative reviews from their customers, or may see an increase in items returned. Furthermore, unexpected events can interrupt this aspect of the supply chain, and these may have a “domino” effect on the later stages of the supply chain. A simple example might be if a delivery truck breaks down, items may not be delivered to customers on time, or items might be damaged in transit.

 

AI including machine learning and the use of data analysis can have considerable applications within order fulfillment. REI, a popular retailer selling outdoor and camping equipment is prime example of how AI can be used to address supply chain efficiency. With 3 distribution centres and 155 outlets REI has a complex supply chain. In 2017 REI began using IBM’s order optimiser tool. This tool uses data and analytics to make existing order management systems more efficient, by reducing the total cost of the supply chain. The system also analyses demand and helps supply chains work more smoothly by analysing factors like the number of items in any one order, item size and weight, shipping options, processing costs, availability and stock levels. By looking at this data, the tool can ensure that resources are diverted where necessary, to save costs, for example if one area of the delivery system is under a lot of pressure, some tasks can be re-routed through alternative channels, relieving the pressures and reducing the scope for error and problems with order fulfillment. The system can be used to ensure that delivery trucks are fully loaded more frequently, which is more efficient than a series of deliveries with trucks containing half loads.

 

Since moving to this system, REI has reported that their out of stock status is virtually NIL, because it now has more insight into demand and it can predict when demand rises and falls, and stock outlets accordingly. Whereas before the tool was introduced to REI, individual sales on at least 800,000 occasions were lost because REI was forced to display an “out of stock” notification.

 

 

 

Supply chains that don’t function optimally can also be costly where they lead to an “out of stock” status of products or services. If an item is in demand, but the supplier can’t meet the demand because of supply chain problems, then this, too is a cost that suppliers have to deal with.

 

Demand can be managed very effectively with automated replenishing systems. These systems work by scanning for and detecting when stocks are running low and the physical performance of the restocking process. In the USA these systems of automated restocking are in use very widely. Walmart, for example uses robots to scan their shelves and detect when restocking needs to take place. There are additional benefits, as when the restocking process is automated, there are fewer errors, and there is less incidence of theft. Robots never need to rest like human employees do, so the supplier can accomplish restocking more flexibly, for example overnight, or at faster rates, to cope with demand and other variables that might otherwise interrupt operations, and ultimately reduce profit.

 

Amazon Go stores have taken the concept of automation to another level completely. Their stores are cutting edge as they automate the whole shopping experience at so-called “grab and go” stores. Shoppers simply scan a code to show they have entered an Amazon Go premises. They do their shopping and check out using cashless technology where their banks are automatically charged for their purchases. Their exit is also monitored electronically by the store, and then the store is replenished using cameras that monitor stock levels. This system is almost completely powered by AI and robots, with little human intervention. There are many positive aspects to such an approach, for example the store does not need to pay human sales assistants and no manual labour is required to restock shelves. On the other hand though, the Amazon Go concept has been dogged by the fact that having a “cashless” status can discriminate against some people in society. Take the requirement to have a phone that is capable of scanning entry and exit from a cashless store. Disadvantaged people, with no fixed addresses or with poor credit ratings can find it hard to obtain contracts and credit for facilities like smartphones. This has the effect of excluding large groups of society from the Amazon Go shopping experience. It has also attracted controversy for the Amazon brand as a whole, with senators in some cases calling on Amazon to remedy the problem that certain groups of people, like disadvantage people, cant access these shops in the same way as wealthier people with a more settled way of life can.

 

 

 

Many experts have argued that machine learning is at the heart of data driven supply chain management. Machine learning can be defined as any machine that uses its inputs to learn from and continuously improve its outputs. As such, machines that learn are capable of improving their own functions through experience. A common application is within demand forecasting. The supply chain functions better, overall if various stages of the supply chain can be accurately predicted. This gives the supplier more time to plan the supply chain operations. So, machine learning will be used to take historical data about shipments, orders and demand, and will make predictions about future demand. Equally, machine learning can be used to make accurate predictions about more complex elements of the supply chain, like returns. Armed with a better understanding of their operations, a supplier will then be able to operate their supply chain more efficiently. A simple example is a fleet of delivery trucks that need to be made available to cope with orders at Christmas. When a supplier is working on guesswork, it may be sensible to have a greater number of trucks available for this high demand period, in case they are needed. However, any trucks that are not needed represent waste in the overall supply chain. When a supplier is able to make more accurate predictions, this makes the operation of the supply chain more efficient, with less waste and less error.  

 

Predictions made based on data available in real-time are also very valuable to a supplier in terms of their supply chain. Point of sale data can be used, for example to identify popular products. When suppliers know that a given product is selling very well, they can ensure that they don’t run out of stock because they are better able to cope with the demand they are experiencing. Equally, point of sale data can indicate when a product that was expected to sell better, isn’t performing in the way that was expected. This enables the supplier to take important decisions like clearing shop floor space, or warehouse space or deciding how many sales agents are needed in a given store. Other advantages of forecasting through applications of machine learning and AI related to advertising spend. Advertising spend on poorer performing products can be adjusted, and resources diverted elsewhere.

 

 

 

AI is also being used by suppliers to attract customers at strategically advantageous times. AI systems can be used to assess footfall, and days and times when footfall isn’t as high. Promotions and offers can then be integrated with this data to improve sales during quieter periods.

 

 

 

AI is making a considerable impact on supply chain management as a whole, mainly because an AI and data driven supply chain strategy can enable suppliers to get ahead of their competitors, cut costs and reduce overheads.

 

On the other hand, suppliers need to be wary of excluding the input of humans completely. A complete reliance on AI has been shown to be problematic in some cases, as the instance of cashless stores that disadvantage people highlights. As such, retailers need to plan their supply chain management strategies carefully, and ensure that they are tailored to the needs and nuances of individual businesses and supply chains.