What is the AI?
Artificial intelligence (AI) is a technology that simulates, through computers, the problem-solving and decision-making abilities peculiar of the human mind and can be used for a variety of purposes. So far, the areas of application are many and it is indeed hard to find an area that has not already been affected.
In the context of Industry 4.0, and more specifically in S&OP and MOM applications, AI plays a major role since, by enabling analysis of data that would be difficult to process manually due to its numerosity and complexity, it provides quantitative elements to support decision-making processes. This enhances both the ability to plan already established and robust S&OP and MOM processes and to improve their responsiveness to adverse events.
In both cases, it is worth emphasizing the importance of the term “support” for decision-making processes. AI can also be used to automatize certain tasks: however, this is often seen as a synonym for human replacement, with associated ethical issues whose discussion is beyond the scope of this article.
What’s the difference between artificial intelligence and machine learning?
Artificial Intelligence and Machine Learning are not the same thing; these two technologies are related but different. AI creates the architecture, while machine learning enhances it.
Artificial intelligence, in fact, is the science that for years has been aiming to develop machines capable of making decisions in perfect autonomy: it can be defined as the science that develops the architecture necessary for machines to function like the human brain and its neural networks. Machine learning, on the other hand, is the algorithm that allows intelligent machines to improve over time, just as happens with the human mind. Indeed, without advanced learning, it would not be possible to put artificial intelligence “in motion.”
According to the strategy defined by the sedApta group, the focus is on the inclusion of AI-based capabilities in the suite’s products that enable, through properly trained Machine Learning (ML) algorithms, to help individuals make better decisions based on what has already occurred under similar conditions in the past (human-in-the-loop).
This process has resulted in a virtual assistant integrated within the sedApta Suite that boosts users’ ability to identify patterns among data, enhancing the information assets that companies commonly exploit only minimally.
Artificial intelligence for sedApta
The sedApta Group is in a privileged position in implementing AI capabilities due to the cross-sectionality and heterogeneity of the data processed.
One of the key features of the sedApta Suite is its end-to-end approach to S&OP and MOM processes-from demand forecasting processes to long and medium-term planning, inventory management, scheduling and management of logistics services.
In addition, considering that the Suite can be integrated with major ERP systems, it becomes clear how significant and distinctive such a database can be.
Within the sedApta group there are two organizations, AImesys and I2Lab, which collaborate closely with the R&D function in identifying, developing, and integrating innovative AI features into sedApta products in both Process and Discrete domains.
Use cases realized by sedApta in the AI field
Successful inventory planning is an increasingly complex process due to the high variability of factors that must be considered: first and foremost, demand, the behavior of which is more difficult to predict due to increasingly complex variables (more transactions, greater customization required by customers, and more players involved among the main causes) and supply lead times that tend to be less stable and reliable.
According to Gartner, supply chain managers are under constant pressure to contribute positively to cash flow through inventory reduction. Key steps to take to increase cash flow while maintaining high levels of service, profitability and growth include better management of reorder points based on lead times, demand patterns and target service levels established based on safety buffer analysis [Gartner’s Top Actions for Supply Chain Inventory Reduction, 2018].
For this reason, among the main S&OP processes for which sedApta has trained ML algorithms, underlying the operation of the virtual assistant, are Demand Planning and Inventory Management processes.
Today, sedApta provides its customers with functionality that enables them to more accurately forecast demand and generate a more robust inventory replenishment plan. This makes it possible to optimize inventories and increase the ability to react to unforeseen events brought about by the dynamism of the market and the state of the logistics infrastructure as a whole.
Demand Planning
- Use of exogenous variables to increase the accuracy of demand forecasting. Variables used may include macroeconomic variables (e.g., GDP), energy price trends, sector-specific variables, and variables related to particular situations (e.g., movement flows during pandemic to cleanse time series of the ‘covid effect’). Depending on the type of exogenous variable, each appropriately weighted, one can forecast demand in the medium to long term (e.g., GDP exogenous variable) or refine it in the short to very short term (e.g., weather forecast exogenous variable).
- Demand forecasting optimization. Traditional algorithms, selected through best-fit logics, are applied alongside multivariate ML algorithms. The latter, by leveraging the endogenous variables in the time series and any available exogenous variables, make it possible to identify conditions that are inherently complex to predict in which, in the past, traditional algorithms and/or user-made corrections within a collaborative environment have proven inaccurate. ML algorithms signal the likelihood that such conditions will reoccur and propose a corrective factor for the prediction to the user.
- New product rollout. One of the most critical aspects of forecasting demand is the calculation of the forecast for new products, for which there is no historical basis available to use. This problem tends to be overcome through manual selection by the user of an existing product that he or she believes is similar to the new product. In this area, sedApta has trained supervised and unsupervised algorithms to produce a quantitative ranking of the products most similar to the new product. This ranking is consulted by the user to analytically validate the choice they would have made without the support of the virtual assistant and/or identify similar products they had not thought of.
Inventory Management
- Supplier Ranking. Supplier performance (e.g., punctuality, quality, sustainability, etc.), typically calculated in ERP systems, is contextualized with respect to the specific conditions under which they were processed in order to provide an even more timely ranking. Given a specific situation, the ML algorithms implemented by sedApta, make it possible to estimate which supplier is most likely to deliver the required performance in line with the objectives of one or more analysis dimensions (e.g., punctuality-quality, quality-sustainability, etc.).