With defence spending increasing and civil aviation passenger numbers predicted to rise, could this be the year that they put their ‘factory of the future’ strategies to the test?
Leo Whyte is Head of Digital Platform and Tom Clements is Data Insights Lead at Vendigital. They recently shared their insights with Consultancy UK.
At a time of significant cost volatility and an increased frequency of supply-side shocks, businesses across industry sectors are grappling to find ways to increase demand certainty. Forecasting has become more challenging as one month’s – or even one year’s – trading figures could be vastly different from another. Could now be the time to onboard AI and machine learning capability to start predicting the way to enhanced profits?
Global factors such as conflicts and materials shortages have created an environment of volatility, uncertainty, complexity and ambiguity (VUCA), with business and political leaders alike uncertain as to what will happen next. For businesses trying to work within these ever-shifting parameters, cost transformation projects have become an essential part of staying viable. While many businesses understand the value that AI can bring to cost transformations, such as accurate demand forecasting, few are ready to fully embrace the technology.
Accurate demand forecasts are essential for every business seeking to maximise revenue and mitigate the risk of under- or over-supply. If a business orders too much stock, this may have to go into storage, incurring fees until sufficient orders come in. On the other hand, a business without sufficient stock could lead to unsatisfied customers choosing to make purchases elsewhere as well as reputational damage due to late or unfulfilled orders. Accuracy in demand forecasts is therefore critical. AI algorithms trained on historical demand data can mitigate the risk of over- or under-supply and free up working capital.
The more data that the AI model is trained on the more accurate its outputs will become, although human evaluation is still required. Like any other supporting tool, AI systems can support human workers; helping them to do their jobs faster and more efficiently. A degree of caution is always needed with any generative AI, as depending on the quality or quantity of data used to train the system, it could develop some form of bias.
Selecting the right AI model for a specific task is not as simple as choosing from a list of ‘plug and play’ technologies. Businesses could choose to build their own and there are several methodologies and approaches available, including both supervised and unsupervised learning. Businesses should aim to use their own data as far as possible, as this will help to ensure accuracy. If businesses lack the expertise to build their own AI model, they could opt for a pre-built one instead, pre-trained on external data, although there are inherent risks associated with data privacy and security.
Whether choosing a pre-built AI system or building a bespoke model, it is important to develop a strong use case and understand the best approach to take. Businesses should therefore seek to understand the quality of the data that they have across their business and whether they have a “single source of truth” before deciding how to go about harnessing the predictive power of AI and machine learning.
For example, supervised learning will be most appropriate if the system is going to be used to predict prices, classify fraud or determine risk. However, unsupervised learning allows businesses to create groups based on key behaviours. Considering the risk of data leakage is also vital to ensure the system’s integrity. Where necessary businesses could draw on external support to boost their internal capability, however, internal teams hold the knowledge and should be responsible for validating the system.
Running models in parallel to prove value and test or validate outputs is typically the best approach. This will help to strengthen confidence in using the AI model, which means the business will soon start to realise benefits due to reduced waste, optimised sales and more efficient inventory management.
Over time, AI models could be developed further and data shared with supply partners; bringing the benefits of increased visibility to the end-to-end supply chain. For example, an AI model could be used to support a dynamic pricing strategy, enabling the business to foresee when price increases are likely and take the right decisions to optimise profits in real time. AI models could even be utilised to improve supply chain efficiencies – maximising profit while simultaneously reducing carbon emissions.
To create a successful strategy, businesses should undertake extensive value-mapping to understand where AI models could bring the greatest opportunity at the lowest risk. By taking a holistic view and focusing on deliverables that can be achieved most easily and bring the most benefit, the business will gain vital skills and knowledge which could become invaluable in the future. Crucially, project managers will also be able to demonstrate proof of value more quickly.
While there is a point at which AI systems will be able to self-manage, for the time being they should be viewed as a management tool with human oversight of outputs required. For many businesses however, the potential rewards far outweigh the risks, so securing board-level buy-in to start in a small way and grow the company’s AI and machine learning capabilities could lead to a successful and sustainable business transformation.
Share this insight
Share this insight
We share industry insights across aspects of EV Supply Chain, Battery Manufacturing and Circular economy including what businesses can do to create certainty around business growth.
Product and pricing decisions are key to survival. But how can a complex business make these decisions amidst so much uncertainty? The answer lies in data understanding.