The UK government’s Industrial Strategy highlighted the important sectors that will create the jobs of tomorrow and will propel the economic health of the nation forward. Among the sectors discussed in the strategy document, the artificial intelligence (AI) and data analytics sector has been identified as one that will influence a number of other sectors in the future. The Made Smarter report also presented the future opportunities with the advent of industry 4.0. In light of this and the current Covid-19 situation a common thread of discussion across the business world is in regards to the use of AI as a tool to solve supply chain challenges.
The challenges faced by grocery and pharmaceutical supply chains have been reported extensively as the Covid-19 lockdown resulted in consumer stockpiling leading to stock-outs and supply chains struggling to fill shelves. A number of articles and newspaper reports discussed this issue under the terms of supply chain risk management (SCRM) or supply chain resilience. A number of discussions asked whether AI could be used to solve or alleviate supply chains risks during Covid-19 and after. This article will provide a perspective on AI and its potential use within SCRM. It will also highlight different techniques and potential issues related to data. The discussion in this article emanates from the research conducted with my co-researchers and the papers for reference are provided below.
Supply Chain Risk Management
SCRM as an academic topic was identified as important in 2001-2002 and has since gained extensive recognition as an important topic both within academia and industry. With changing weather patterns, natural disasters, political turmoil, financial upheavals, and now a pandemic, supply chains are under constant pressure to deliver. The basic challenge in the scenarios mentioned above is the aspect of supply disruption. In general, the focus in most cases is to strengthen the upstream supply chain to maintain continuity of supply. However, as seen in this pandemic the cause of the risk upstream was the lack of human resource due to the lockdown (with operations shut) and an unknown demand profile due to stockpiling on one end coupled with reduced footfall in the retail environment on the other. SCRM in general borrows from risk management techniques from multidisciplinary sources such as engineering, finance and mathematics. There is an increasing amount of literature on SCRM but the basic process is comprised of three activities: identification of the risk, quantification of the risk, and identifying potential solutions to mitigate the impact of the risk. Simplistically, risk can be defined as an event that has the probability of creating a negative impact for the organisation or supply chain. There are a number of papers and articles that focus on the sources of risk (for example, whether the risks are external or internal to the firm, whether the risks are upstream or downstream) the different quantification techniques (for example, risk trees, FMEA), and the solutions for mitigating the impact of the risks (transferring the risk, hedging, postponing the risk, avoiding the risk).
Another aspect of SCRM is to consider both proactive management - creating systems in anticipation of the risk to mitigate; as well as reactive management - reacting quickly to minimise the impact of the risk.
Artificial Intelligence
The research on AI has had an exponential growth in the past decade. The focus initially was to apply machine learning techniques for creating intelligent systems for taking decisions or making conclusions from data. Future AI research is pushing for learning or the more advanced deep learning so that the system can learn and improve further. Research has progressed very swiftly in the past few years and the emergence of big data systems along with increased computing power has provided an incentive for pushing the boundaries in this area. Major companies such as IBM, Google, Amazon, Facebook, Tesla are investing in research to harness this technology.
There are a number of techniques used within AI systems. These techniques or methods can be broadly represented as Petri nets, Multi-Agent Systems, Automated Rule-based Reasoning and Machine Leaning. They have different uses and approaches and it is important to understand their application capability for a certain context. For example, within a supply chain environment, network-based approaches like Petri Nets could prove useful for tracing interactions and assessing the dynamic behaviour of a supply chain. Multi-Agent systems could be used to manage conflicting or coordinating interactions across supply chain entities. Rule-Based Reasoning techniques are useful when systems need to function specifically on the basis of rules. Machine Learning techniques can be applied across a variety of tasks but the effectiveness is influenced by the availability of an adequate amount of data as well as the right kind of data. Deep Learning, Neural Network Programming, Deep Neural Networks are upcoming techniques which are a subset of Machine Learning and will prove useful when AI systems will learn, take decisions, and improve their behaviour on an ongoing basis. The future of AI and its influence on SCRM will stem from the growth of research in two capabilities: Predictive analytics: the ability of the data analytics system to identify data patterns and anticipate future scenarios and Prescriptive analytics: the ability to take a decision to meet predefined objectives.
AI and SCRM
Supply chains are complex systems comprising of various organisations, operations, processes, assets, people, and strategic stakeholders. The complexity is also manifested through the information flowing across the chain which ensures that appropriate decisions are taken for the supply chain to work both efficiently and effectively. These decisions are taken by humans in the chain. As supply chains have gone far and wide there is even more complexity due to unpredictable weather patterns and external sources of risks. However, software systems have grown in capability and with big data systems there is a lot of data available for both pattern analysis and decision making. To add to the big data available from supply chain software systems, there is a plethora of dynamic fast data from social media streams that provides valuable information about ground reality of locations, assets, sentiments, and weather which has the potential to be tapped for better decision making.
The use of AI and specifically machine learning techniques could be used to harness both big data and fast data from the supply chain systems to predict supply chain risks and prescribe mitigating strategies. The future may be a situation where the system takes over the mitigation process after the decision is made, without human intervention. The research is moving in that direction, but from a SCRM perspective the ability of the humans to control and mitigate the risks will be more important.
AI systems will be able to identify the source of risk (ideally predict it from big data analytics), quantify the risk based on past impact data, and suggest the best possible mitigation strategies for that scenario. As discussed above, this method is a regular process, however AI, and data analytics will help provide a faster analysis process, a faster quantification process, and perhaps better suggestions for mitigation based on a wide ranging analysis of past scenarios. For example, if we consider a scenario of sourcing grain from farms, the system could analyse weather patterns, predict which farms could be affected in the future (or near future) and do a portfolio analysis to redesign upstream sourcing to minimise supply chain disruption by identifying the appropriate suppliers from the portfolio. This could also work on the basis of analysing supplier quality data. In another scenario, if a supplier location is affected by a local or regional upheaval, the system could analyse big data and fast data (from social media feeds) to decide the severity of the impact, and then decide whether a supply chain redesign is required for business continuity. In the future, sustainability non-conformance will be a major supply chain risk and AI systems will need to be taught about sustainability parameters and its importance within supply chains.
It is clear from the discussion that the capability of the system will depend upon the AI algorithms and the available data. Data availability and appropriate data is critical for the system to manage supply chain risks effectively. There is debate within the community about whether the AI systems should be like humans, or should be designed to naturally remove all bias as ingrained in humans. Since the discussion will continue until the capability is fully formed, it is important that the AI (Machine Learning) system provides the following characteristics at the present:
Explainability - the ability to explain how the AI system is working.
Interpretability - the extent to which one can predict what will happen if there is a change in the input to the system or parameters in the algorithm, the ability to interpret how the system is working.
Auditability - the ability to audit the system, the data and how this data is used
Transparency - the characteristics that provides full access to how the system works and provides trust to all stakeholders
About the author
Professor Samir Dani is Professor of Operations Management and Head of Marketing, Management, and Organisation at Keele Business School, Keele University. Professor Dani engages widely with industry and has research interests in supply chain risks, sustainability in supply chains, and the use of technology and business models. He is currently conducting research in the use of AI techniques and autonomous systems within supply chain decision making and logistics. He works closely with the food sector and has published widely and presented his research to both academic and practitioner audiences.
References
Baryannis, G., Dani, S. & Antoniou, G. (2019) Predicting supply chain risks using machine learning: The trade-off between performance and interpretability, Future Generation Computer Systems, 101, p. 993-1004 12 p, 29 Jul
Baryannis, George; Validi, Sahar; Dani, Samir; Antoniou, Grigoris (2019), Supply Chain Risk Management and Artificial Intelligence: State of the Art and Future Research Directions, International Journal of Production Research, 57 (7), 2179-2202
Baryannis, G., Dani, S., Validi, S., Antoniou, G (2019) Decision Support Systems and Artificial Intelligence in Supply Chain Risk Management, Revisiting Supply Chain Risk, ed: George Zsidisin, Michael Henke, Springer Series in Supply Chain Management
Abhijeet Ghadge, Samir Dani, Michael Chester, Roy Kalawsky (2013) "A systems approach for modelling supply chain risks", Supply Chain Management: An International Journal, Vol. 18, Iss. 5, pp. 523 - 538.
Chaudhuri, A., Ghadge, A., Gaudenzi, B. S Dani, S. (2020) A conceptual framework for improving effectiveness of risk management in supply networks, International Journal of Logistics Management, ISSN: 0957-4093
