Integration of Artificial Intelligence Applications on the Supply Chain

Artificial intelligence (AI) is the ability of a machine to perform human-like functions such as perceiving, reasoning, learning, interacting with the environment, problem-solving and exercising creativity to form plans, make decisions and achieve goals (Cheater, 2018). While the initial field of artificial intelligence research began on Dartmouth College in 1956 it wasn’t until the 21st century where machine learning was successfully applied to advanced problems; however, significant advances were made in the early 1990s with supercomputers such as Deep Blue. This supercomputer developed by IBM was the first of its kind and could process 200,000,000 moves per second becoming the first AI framework to defeat a world chess champion. By the end of the 20th century, significant progress was made in computing power, and algorithmic advances. That progress has led towards the creation of viable artificial intelligence applications which are on the rise. When properly integrated into the supply chain AI offers considerable ways to improve current processes and improve supply chain visibility.  

Futuristic AI Applications

Artificial Neural Networks

Artificial Neural Networks (ANN) would work as an elaborate network of computer memories in the same way the brain cells operate, at least in theory. These memories would allow for learning, recognition of designs and structures, procedures, and intangible statistics. ANN would be able to achieve this by using nodes of neurons much like the brain, as storage. These nodes connected in links would give their formation gross weight, some small others large. Each link offers a means of memory storage and processes information output from one neuron to another while the weight of each link is the primary response to the strength or weakness of information passed.

The use of ANN has already been proven to assist autonomous vehicles using processing techniques. ANN is helping vehicles to maneuver motor vehicles down highways however, at this moment this technology is still very limited to road conditions and certain traffic environments. Additionally, ANN has been successfully utilized in the development of supply chain planning that helped determine specific time and capacity needs, lot-size requirements, inventory and scheduling, and demand and production planning. 

Rough Set Theory

Zdzislaw Pawlak, a professor and research scientist at the Institute of Theoretical and Applied Informatics, is accredited for creating the rough set theory. Its methodology is concerned with the classification and analysis of imprecise, uncertain or incomplete information and knowledge, and of is considered one of the first non-statistical approaches in data analysis (Pawlak, 1982).  Through the development of rough set theory, Pawlak outlined several classification attributes needed for implementation mechanisms. Attributes such as set approximation, similarity classes, rough membership, and indiscernibly relations. The use of classification attributes is necessary for indiscernible objects to be found.

Supply chain management can use rough set theory to categorize decision data and develop rules relevant to suppliers. That is completed by introducing qualitative data to supply chain systems. Once the data is entered, it can then be evaluated, and modified in a manner to select the top distributors. Furthermore, the rough set theory can be used to measure supply chain performance through developing decision-making criteria.

Machine Learning

In 1959, Arthur Samuel defined machine learning as a field of study that gives computers the ability to learn without being explicitly programmed (Puget, 2016). Learning categories are broken down into three main categories: (1) supervised learning that is a system trained using past data and is able to take decision or make predictions when new data is encountered; (2) unsupervised learning which is when a system can recognize patterns similarities and anomalies, taking into consideration only the input data; and (3) reinforcement learning which are decisions made by the system on the basis of reward/punishment it received for the last action it performed (Sunil Ray, 2017). 

Machine learning techniques can be a useful tool within supply chain management regardless of learning tasks. It can be used to understand behavior between competing suppliers or on ways to better improve potential partnerships. It can also be used to enhance the understanding of critical information shared and organizational learning processes. Machine learning has been shown to reduce freight costs while improving delivery performance and reducing supplier risks. The use of machine learning can potentially lower operating costs by cutting waste and improve customer service through accurate inventory representation. One example of machine learning is IBM’s Watson. It can determine if a shipping container or product were damaged, classify it by damage time, and recommend the best corrective action to repair the asset (Columbus, 2018).

AI Agent-based systems

According to Jennings (2000), agent-based computing provides the theory and the practice of modeling, designing, and implementing real-world computer applications. The term agent especially has several proposed definitions.  An agent as defined by (Russell and Norvig, 2014) is anything that can perceive its environment through sensors and act upon that environment through actuators. As global supply chains focus more on collaborations among partners, agent-based technologies will play an invaluable role, and existing techniques within web-enabled systems will need to be used to seek superiority in the supply chains.

Although there are current techniques used for partnership within the supply chain we are in an era where web usage is flourishing, and e-commerce has intensified. For companies to maintain their dominance within this sector businesses will need to pair existing technologies and techniques with agent-based approaches. That can be accomplished by developing new frameworks founded on agent-based web services in areas where demands are high. Agent-based systems can also identify ideal solutions and recommend price adjustments and levels of future production quantity.

Expert Systems

Expert systems are computer programs that use AI methods to find answers within a specific field that would typically require human expertise. It was first experimented within 1965 by Edward Feigenbaum and Joshua Lederberg when they designed a system to analyze chemical compounds (Zwass, n.d.). There are three components of an expert system: (1) Knowledge Base which is where the information is stored within the expert system. Data is stored in the form of realities and rules (2) User Interface is where the user interacts with the system. That is where questions are asked, and results or advice is distributed. There will be justification as well in the form of how, or why. (3) Inference engine will apply the facts to rules. That will be where the systems ask questions of the user within the system interface. Problems would be laid out in a distinct order with the reasoning for each asked question. It is also considered the brain of the expert system.

The informal use, terms, and concepts which expert system operate make it a desirable alternative to solve practical supply chain problems. Expert systems are already known to have increased productivity while managing logistical aspects for the Airforce. In 1986 Allen successfully explained several inventory control problems while using more than 400 inputted rules and guidelines. Additionally, IBM developed a real-time, transaction-based system with the ability to schedule, monitor, and control the flow of logistics at one of their semiconductor facilities. Reports after implementation showed a 35 percent increase in product output and $10 million in savings. Future applications within the supply chain field are limitless. Established applications have already been implemented to air traffic control, 3-D mapping, vehicle repair, and maintenance scheduling.

Fuzzy Logic

Fuzzy logic was first advanced by Dr. Lotfi Zadeh in 1960 while attempting to form a computer-based understanding of ordinary language (Hokey, 2010). Fuzzy logic is a computer-based approach based on “degrees of truth” instead of “good or bad” or “true or false.” Unlike Boolean logic which values variables on true or false, usually annotated in 1 or 0, fuzzy logic allows for techniques to deal with uncertainty and can be a valuable approach to managing supply chains.

Using fuzzy logic, an object takes a value between 0 and 1 and provides a gradual transition from “membership” to “non-membership.” This gradual transition allows for the mathematical expression of objects with fluctuating conditions. For example, while moving goods in a box truck or railroad, I could set at a cold temperature of 0.1 and a warm temperature of 0.7; another example would be if I gave a poorly made item a fixed number of .2 and a better-made item a 0.5. Using these metrics or any set by suppliers or manufacturers, fuzzy logic can assist in developing frameworks associated with performance criteria within supply chains. Additional fuzzy logic applications in supply chain management could include order fulfillment, inventory control costs, measurement of the bullwhip effect, supplier selection and performance evaluations.

AI Applications in Supply Chain Management

Control and Planning Inventory

Having available inventory ready to distribute to consumers is required to maintain high rates of customer service but, comes with substantial costs. That is why most company’s financial success hinges on their ability to plan and control their inventories while also providing enough products to consumers as needed. Having the ability to accurately represent real-time information of consumer demands, stock on hand, and order fulfillment times would ensure these guidelines are met. Although that kind of information is typically complicated to estimate expert systems can replace inventory managers and provide sound judgment to better suit companies inventory control.

As mentioned above improved artificial intelligence techniques like expert systems provide a different approach to supply chain management. These systems offer greater flexibility to inventory control and planning whether the problem is straightforward or complicated. Dynamic issues might arise forcing inventory experts to inaccurately estimate correct levels of goods to restock causing a bullwhip effect. With the implementation of expert systems, you can incorporate it into your systems thoroughly. This allows for complete capture of inventory, material costs, and order patterns. You can even estimate future order levels and potential for inventory replenishment.

Transportation Networks

Transportation network design has been one of the most popular places for emerging AI applications. Oddly enough these applications have also come with the most significant issues, and global solutions are trying to find. Problems have included; vehicle routing, road network design, traffic assignment, freight consolidation, and intermodal connection. Due to the complex nature of these problems genetic algorithms have emerged as a popular solution to some of these issues.  Algorithms such as the colony optimization or ant colony optimization algorithm have provided techniques for solving computational problems centered around finding the most optimal way through parameter space. That has proven very successful in dealing with vehicle routing and traveling salesmen problem.

Purchasing and Supply Management    

Purchasing and supply management involves the procurement, storage, and monitoring of goods sold in a retail store, machinery, supplies, or other raw commodities. Inventory is a big part of purchasing and supply management. When shipments are delivered, they are usually taken to warehouses and compared to a product that was initially purchased allowing for managers to gauge the item which need to be ordered for frequently and which ones don’t. You might think this is a pretty straightforward process; however, there are a lot of “what if” scenarios that go into a make-or-buy decision.

 It is because of this type of complexity that decision aides are starting to become more evident such as expert systems. As mentioned above expert systems have already been used by our armed services in the early 80’s and saved roughly $10 million in associated costs. To handle a broader spectrum of purchasing goods AI offers agent-based systems to automate online order processing. Additionally, in 2004 Cheung, Wang, Lo, and Lee developed a hybrid agent and knowledge-based system that would evaluate on-line bids and supplier performance before fulfilling orders. Another more recent applications from Nissen and Sengupta in 2006 proposed software agents they could automate searching for carriers through a network of online catalogs. Those agents would also look at supplier evaluations, dividing them by attributes, screen suppliers, and complete order procurements. Agent-based systems are gradually assisting supply chain managers purchasing decisions while traditional techniques only handle one aspect of the purchasing decisions.

Projecting Consumer Demand

The problem with demand forecasting in today’s businesses is that they lack the equipment to leverage the capacity and assortment of data because they can’t account for real-time changes that make the difference between profit and loss. Having the ability to forecast future demand goes hand in hand with company’s capability to plan successfully, schedule, and control one’s inventory. The company must find ways to reduce uncertainty inherent in future orders.  Although this task can be difficult, AI agent-based system frameworks can be used to successfully forecast product demand for almost every business.

The most important thing with demand forecasting is the ability the accurately reflect what your demand is. Companies like Merck are putting AI to work optimizing demand forecasting. After years of dealing with demand inefficiencies, Merck was able to increase its level of service to hospitals from 97 percent to 99.9 percent according to Merak CIO Alessandro de Luca. With incorporating AI, forecasting technology companies can predict not only product demand but associated sales accurately.

Warehouse Management

Order-picking consists of selecting the correct item that has been ordered. While this seems easy enough, this can be somewhat labor-intensive and account for huge portions of warehousing and its expenditure. For that reason, AI systems are being implemented across warehouse operations and are assisting warehouse managers. AI computerization has allowed for automation sequencing to helping in filling orders. The use of intelligent agent-based systems automated machines can be assigned a zone to pick orders. The purpose of these AI techniques has been shown to better handle the complexity of warehouses leading to an increase in e-fulfillments.

Customer Service Management

Customer relationship management programs are used to ensure parts and service get to customers when needed after sales are completed by automating business processes used for sales, service, and support (Duggan, 2018). To maximize your operations both SCM and CSM, functions need to be as integrated as possible. AI plays a crucial role in this integration specifically in SRM by providing the tool necessary pattern data analysis, configuring correct course of actions, and addressing customer concerns (Dickson, 2017).  For example, Salesforce developed an AI assistant called Einstein that will continuously study data collected from sales, e-commerce, emails, and social media streams. The data once gathered will be surveyed by the AI engine can adequately provide recommendations to sales representatives based on data analysis. Another example is SAP software solutions and what they are developing is AI functionalities with their current cloud software. That, in theory, will allow for data analysis to monitor vendor accounts and can prepare performance lists from the data it collects.

E-synchronized SCM

Synchronized planning generally describes a state in which a constant flow of data from throughout the supply network enables organizations to accurately plan production to match actual demand (Gaus, Olsen, and Deloso, 2018). Supply chain partners usually share large amounts of information to assist in demand forecasting and distribution planning through web-based outlets. That massive amount of data allows for AI applications like machine learning to assist in web-mining. By powering advanced AI technologies to interconnected systems, large amounts of data can be analyzed. That will allow companies to make smarter decisions, reducing costs. Interconnected AI application can examine revenue and sales trends, as well as customer profiles stored on different websites. That will ultimately result in a more dynamic, and effective supply chain that combines traditional planning and execution. 

Conclusion

In closing, there is an abundance of potential applications for artificial intelligence that could positively affect the supply chain and its managers. From the theory of artificial neural networks, and fuzzy logic to existing applications in use this very moment it’s clear to see that artificial intelligence in SCM is here to stay. While the use of some of the outlined technologies is only in its infancy, it has already been proven to work by numerous elite companies. As AI continues to evolve mid-level business will be able to reap the benefits within the global supply chain.

References

• Hokey, M. (2010, February). Artificial intelligence in supply chain management: Theory and applications. Retrieved on September 17, 2018, from https://www.researchgate.net/publication-/247523024_Artificial_intelligence_in_supply_chain_management_Theory_and_applications
• Ray, S. (2017, September). Essentials of Machine Learning Algorithms (with Python and R Codes). Retrieved on September 17, 2018, from https://www.analyticsvidhya.com/blog/2017/09/common-machine-learning-algorithms/
• Columbus, L. (2011, June). 10 Ways Machine Learning Is Revolutionizing Supply Chain Management. Retrieved on September 18, 2018, from https://www.forbes.com/sites/louiscolumbus-/2018/06/11/10-ways-machine-learning-is-revolutionizing-supply-chain-management/#6bb3845f3e37
• Kwon, O., Im, G., Lee, K. (2011, November). An agent-based web service approach for supply chain collaboration. Retrieved on September 18, 2018, from https://www.sciencedirect.com/science/article/pii/S1026309811002112
• Glass, S. (2018, January). AI and the Evolution of Demand Forecasting. Retrieved on September 19, 2018, fromhttps://medium.com/@Aera_Technology/ai-and-the-evolution-of-demand-forecasting-147dd4e783aa
• Dicksen, B. (2017, June). How artificial intelligence is revolutionizing customer management. Retrieved on September 21, 2018, from https://thenextweb.com/contributors/2017/06/12/artificial-intelligence-revolutionizing-customer-management/
• Miller, R. (2016). SAP adds AI and integrated analytics in latest cloud release. Retrieved on September 23, 2018, from https://techcrunch.com/2017/02/09/sap-brings-ai-and-integrated-analytics-in-latest-cloud-release/
• Duggan, T. (2018). Supply Chain Management Vs. Customer Relationship Management. Retrieved on September23, 2018, from https://smallbusiness.chron.com/supply-chain-management-vs-customer-relationship-management-10231.html
• Gaus, T., Olsen, K., Deloso, M. ( 2018, May). Synchronizing the digital supply network. Retrieved on September 24, 2018, from https://www2.deloitte.com/insights/us/en/focus/industry-4-0/artificial-intelligence-supply-chain-planning.html
• Cheater, A. (2018, March). An executive’s guide to AI in supply chain management. Retrieved on September 31, from https://blog.kinaxis.com/2018/03/an-executives-guide-to-ai-in-supply-chain-management/
• History. (2018, August). Kasparov defeats chess playing computer. Retrieved on September 31, 2018, from https://www.history.com/this-day-in-history/kasparov-defeats-chess-playing-computer
• Benjamin, J. (2018, May 11). En Route Automation Modernization (ERAM). Retrieved October 6, 2018, from https://www.faa.gov/air_traffic/technology/eram/
• Pawlak, Z. (1982, September). Rough sets. Retrieved on October 08, 2018, from https://sci2s.ugr.es/keel/dataset/includes/catImbFiles/1982%20-%20Pawlak.pdf
• Puget, F. J. (2016, May) What is Machine Learning?. Retrieved October 8, 2018, from https://www.ibm.com/developerworks/community/blogs/jfp/entry/What_Is_Machine_Learning?lang=en
• Zwass, V. (n.d.) Expert System. Retrieved on October 10, 2018, from https://www.britannica.com/technology/expert-system
• Jennings, N. (2000, March). On agent-based software engineering, Artificial Intelligence. Procedia Computer Science, 117(2), 277-296.
• Nissen, M.E., Sengupta, K. (2006, March). Incorporating software agents into supply chains: experimental investigation with a procurement task. MIS Quarterly, 30(1), 145–166.
• Cheung, C.F., Wang, W.M., Lo, V., Lee, W.B. (2004, January). An agent-oriented and knowledge-based system for strategic e-procurement. Expert Systems, 21(1), 11–21.