Artificial intelligence (AI) is a hot topic. From computers taking over many jobs to the potential extinction of humanity, it cannot be said that predictions lack scale and imagination. Those leading the technical development of AI have called for government regulation to mitigate risk. Notwithstanding the need for frameworks to emerge that balance the potential benefits of AI with the risks, what are the opportunities AI might present for the challenging activity of managing supply chains?
Like many technological developments, AI is generating much hype and a lot of impenetrable jargon. So, what is AI? It is perhaps unfortunate that the term ‘Artificial Intelligence’ was coined in the first place; this conveys a sense of consciousness to machines and suggests some form of sentient being. Maybe a better term might be ‘Rapid Applied Statistics’ (Accelerated Applied Statistics) given the foundational aspects of techniques such as Machine Learning that use statistical algorithms to detect patterns and make predictions, be that related to numbers, words, images. It is the development of computational power that now brings unprecedented volumes of data within the grasp of such software programs. More data, more tested correlations, more assessed probabilities, and the results get better. But, as ever, data must be qualified to avoid error and probability does not necessarily equate with reality.
Imagine a supply chain with seamless connections; as goods are purchased in stores, demand requirements are assessed at inventory holding locations across the network and activities are initiated as and when required without any human intervention. Based on the parameters that are used to manage service levels and capacities across the supply network, production is planned and executed, stock is moved, all at a cadence that optimises the efficiency of factories, warehouses, transport modes. Such a ‘hands-off’ supply chain would be unheard of today. Despite technological advances, many nodes in current supply networks operate in isolation, second guessing what customers and suppliers will do and require; often with the added benefit of human interactions that seek to add opinion with limited factual grounding. Current supply chain management is very ‘hands-on’ with significant human resource assigned to monitoring, assessing, and planning activities.
With techniques such as Machine Learning, Machine Vision, Neural Networks, Natural Language Processing being utilized in rapidly expanding settings, how should we understand and qualify the opportunities to transform ways of working in supply chains? Although a somewhat traditional hierarchy, perhaps it is helpful to consider activities at strategic, tactical, operational levels, for example:
AI applications have already been identified in demand forecasting, production and logistics optimisation, supply risk assessment, and quality control amongst others. Positioning these developments in a context such as the hierarchy above will support an assessment of the extent of change and scale of impact for a business. Equally important is to understand how a development in one activity is likely to impact related activities and the broader supply chain network. Thinking of the supply chain as a networked system will be essential to understanding opportunities offered and how they might transform ways of working.
Ultimately, creating a more ‘hands-off’ mode of operation requires co-ordination of activity across the processes running in the network. Techniques such as value stream mapping and SIPOC can help define activities and their interrelationship. However, with potential AI developments, it will be critical to define and map what data is generated and captured through the operational network, including any implications as to whether this data might be shared and transferred. Applications such as ChatGPT access large pools of publicly available data whilst much data in the supply chain is proprietary.
Marketing campaigns seem fixated on ‘disruption’, ‘optimisation’, ‘transformation’ as they seek to drive sales of a myriad of applications for a ‘digital world’. There is, however, a long track record of system and technological developments promising much and struggling to deliver. What could get in the way of reaping the benefits of AI? Some potential hinderances include:
Data is the lifeblood of AI applications. From a supply chain perspective, whilst internal processes in a business can offer significant improvement opportunities, often it is collaboration and co-ordination with both customers and suppliers that can yield greater overall benefit. Businesses will need to consider how, and to what extent, they are prepared to share proprietary information to supply AI applications with the comprehensive datasets that will maximise their capabilities.
If there is ever a business process that benefits from cross-functional working, it is supply chain management. The alleged benefit of ‘creative tension’ between the objectives of production, finance, sales, for example, typically undermine supply chain efficiency. Similarly, either pursuing AI approaches to drive purely functional goals or seeking to constrain AI applications to isolated operational activities risks limiting the potential for improvement.
There is a natural caution and scepticism around software that generates outcomes when we struggle to understand the methods used and, in some cases, dislike the results. This can be healthy, and we will need to maintain oversight to the recommendations and actions of AI applications. That said, there will need to be an adjustment in thinking about AI with a greater focus on the inputs and logic that these applications apply rather than solely on outputs. The quality, validity, and scope of the data inputs will govern our ability to apply these technologies with confidence. Recognising that comprehending the vast array of data that AI uses is beyond human cognition does not mean that people no longer have a role to play. It is the nature of activities and tasks that will need to change in an AI enhanced environment.
AI will offer the opportunity to re-think and re-design much of how supply chain networks currently operate. Key dimensions that will drive benefits are likely to be:
A consequence of the second guessing that pre-dominates much of supply chain management today is that time is used as a buffer for many activities; there are many instances where elapsed time is far greater than active time in a whole range of operational settings. Decision making often waits for some form of hierarchical approval that either affects minimal change or causes re-loops based on limited grounding in fact. There are likely to be significant opportunities to remove non-value adding time and increase the clock speed of decisions, thereby improving service, lowering costs, and reducing inventories across the supply chain.
The large datasets that support AI applications, when suitably developed, lend themselves to greater predictive capabilities in a range of contexts. From better demand forecasts to more reliable rhythm wheel settings for production, there is significant scope for aligning supply to demand. The skill will be in configuring the AI applications to practicalities on the ground. Adjustments to plans and schedules have a threshold beyond which they tip into destructive disruption to operational efficiency; a production quantity or a replenishment volume must be fixed at some point for things to get done.
The ’Internet of Things’ is not a new concept. With sensors, radio-frequency identification, and cloud computing, we are in an increasingly connected world. Machine to machine communication offers increased visibility across the end-to-end supply chain. The potential role of AI will be to determine patterns in this mass of data and signal impending variations or highlight issues. It is likely also to be able to offer adjustments and solutions that have a demonstrable record of successful results.
For businesses, the path to AI adoption needs to be planned as far as practicable. Any transformation programme will need to adapt and adjust in light of experience and the emergence of new applications. That said, such an agile approach will benefit from a fundamental understanding of the key dimensions that will drive most benefit and being able to qualify the nature and scale of the change implied by AI adoption. Whilst our focus may naturally be drawn to the technology, greater attention will need to be paid to the changing processes and the new organisational structures and job roles that will be required. It will be how well the transition for the people in the business is led and managed that will determine success.