We are talking about a sequential process (figure 2) in which, together with the business, production or digital transformation managers, we understand the internal aspects of their business and analyse the available data while complementing it with data from other contrasted sources and based on our experience.

This part of the process leads us to define a ‘GO – NO GO’ depending on the data provided and the quality of the data. In cases where the data does not allow us to work with AI, clients obtain a clear vision and a roadmap that leads them to sensorise their processes and start obtaining quality data, which is already a step forward and a positive impact in terms of their production strategies and an increase in their digital maturity that is critical to their future.

In a significant majority of cases, and given that large manufacturing companies tend to have increasingly high-tech and highly sensorised production systems with extensive IoT ecosystems, we have access to a good collection of quality and sufficiently representative historical data on their manufacturing processes, which allows us to build datasets on which we can deploy the next step, in which we begin to apply Machine Learning and Deep Learning techniques that allow us to detect the variables with a critical impact on each sub-process and the most relevant: modify those on which we can intervene in a feasible way according to the prescription of our Operational Research (OR) systems embedded in our solution, thus optimising the integral process.

With all the functional experience we have accumulated, exploring the limits of the application of these analytical techniques in the optimisation objectives of processes as complex as those of discrete or continuous manufacturing, we can undoubtedly affirm that even AI itself, with all its power of evident transformation in so many scenarios, has its limitations in these complex approaches and that only with the hybridisation of Artifitial Intelligence (AI) and Operational Research (OR) can we unlock this transforming value of the business and achieve this new paradigm.

Thus, with the implementation of this solution, in which algorithms and Artificial Intelligence and Operational Research models coexist and feed back, we define the operating models that ensure the maximisation or minimisation of the ratios that in each case the manufacturer considers most critical for the profitability and efficiency of the specific manufacturing process on which we are impacting.