Four ways how AI technology can disrupt the industrial sector

12 March 2021 7 min. read

The rise of artificial intelligence is set to disrupt all facets of the industrials industry – from business models and operations through to innovation. Aimé Lachapelle, a partner at international consulting firm Emerton, shares use cases of how AI can improve performance across four key areas of the value chain.


While production chains have been the subject of many successive process improvements over the years, artificial intelligence makes it possible, by capturing all the dimensions and complexity of processes, to carry out new and high added value optimizations.

One of the most widespread use cases, often appearing to be the most natural due to easy conceptualization, is that of predictive maintenance. The goal is then to predict, using AI algorithms, machine incidents and breakdowns, and thus optimize maintenance. The reality that has emerged from the field is that this use case often turns out to be difficult to achieve, due to the scarcity of observations of truly impactful failures. 

Four ways how AI technology can disrupt the industrial sector

While the large volume of data captured in real time may seem attractive for prediction, the order of magnitude that matters for AI algorithms to learn is that of the number of occurrences of failures at machine level observed over a year, which can be counted on one or both hands. In this case, even the best AI algorithms cannot learn the mechanisms underlying these failures. This very concrete example highlights the criticality of the correct framing of an AI use case, which must necessarily be done in contact with data and strongly involve business and scientific experts, in order to maximize the chances of success. 

The optimization of production processes being the engine of the industry represents a significant potential for gain, particularly in the metallurgical, chemical and pharmaceutical industries. It is cited by 60% of industrial leaders surveyed as appearing at the top of the list of priority applications to be implemented on their production chains. Whether for the optimization of production parameters or better production planning, the proposed AI solutions allow a significant improvement in productivity, quality and energy consumption. 

A first step in these optimizations is often to focus on a target process at the heart of production. The IT director of a group specializing in glass forming indicates: “after several failures, we have concluded that one of the two major prerequisites for the success of an AI use case is to sufficiently circumscribe the considered scope.” For example, some players in the metallurgy or chemical industries optimize the quantities of integrated reagents in order to improve product quality, while reducing the input of raw materials and energy. 

The quality of a product is difficult to optimize because it is very difficult to predict. This is true in the process industries, where quality is often only known after a long and complex process. This is also true in the assembly industries, where the quality issues of many parts are also difficult to master. 

Quality can therefore be impacted by a large number of events and is therefore very often unpredictable. This is why the applications of quality prediction and cost of quality, for example via the performance of its control, are a vast playing field for artificial intelligence, particularly thanks to computer vision, often applauded by manufacturers. Quality defects are then detected flawlessly and at negligible cost by AI algorithms. 

Finally, one of the prerequisites for integrating AI into the heart of industrial processes is to develop digital factory twins, mimicking production flows as a whole, in order to optimize them. The manufacturers who have succeeded in carrying out production optimization AI and data projects are those who have been able to put AI experts as close as possible to their industrial experts, allowing the industrialization of AI use cases. 

Supply chain

The supply chain, by its complex nature at the interface of many processes, offers a fruitful playing field for artificial intelligence algorithms that allow the simulation and optimization of a large number of scenarios. Several AI players have specialized in supply chain-related topics given the high value potential. 

A wide variety of algorithms are used for supply chain use cases. For example, we can cite machine learning models which are used to forecast customer demand and transport or delivery times. Thus, the entire supply chain can be adapted accordingly. For example, two supply chain experts in the assembly industry told us that they “use real-time data to predict transit time between depots.” 

The centralization of all data in the chain combined with the use of AI solutions enables not only stronger predictions for decision-makers in the supply chain, but also better decision evaluation on a global scale. For example, helping to orchestrate production and internal transfers is a critical application for firms with several factories. In some cases, manufacturers go further and allow the algorithm to make decisions in real time. An example of the use of the latter technique is the optimization of transport costs, a key element of operational excellence in the supply chain.

Innovation: research & development

Beyond the central applications of production and the supply chain, use cases of artificial intelligence are beginning, although to a lesser extent, to be implemented in R&D functions. Thus, AI is becoming a tool of choice for analyzing technical trials. For example, this is the case in the aviation industry where AI solutions are used to analyze the results of vibration tests on certain critical parts.

In the food, chemical and pharmaceutical industries, AI also makes it possible to limit laboratory experiments by selecting the most relevant experiments or by directly predicting their results, and above all by correlating them to customer expectations. “The real contribution of AI to us has been being able to unlock the product portfolio, to create products that meet expectations while limiting R&D investment and testing,” the CIO of a global food giant tells us. 

These tools are also used earlier in the design of products. We can cite the example of the automotive industry where so-called generative design tools use AI algorithms to automatically generate plans for certain parts according to pre-established constraints. Although these methods are still in an early phase of their development, they could accelerate prototyping if they were coupled with 3D printing. The perfume industry illustrates another product design use case where AI tools can predict whether a prototype perfume will be chemically stable. 

Despite the rise of use cases in R&D, the main obstacle to a greater democratization of artificial intelligence tools in this function remains the lack of explainability of the behavior of AI algorithms, which are often seen by engineers and designers as black boxes. 

Commercial: sales & marketing

In the B2B segment, the sales and marketing functions of manufacturers have a more limited volume of data than in B2C, and a priori represent less fertile ground for Artificial Intelligence applications. Among the actors interviewed, these themes are handled with lower priority.

Some industrial companies such as more B2C oriented consumer goods, are starting to deploy AI tools to improve their sales. In particular, this is the case in the cosmetics and perfumery sectors which use algorithms for hyper-personalization of products or the analysis of online reviews, thanks to NLP algorithms (Natural Language Processing).

More information? Download Emerton’s ‘AI for Industry’ white paper produced in collaboration with and Startup Inside.