Artificial Intelligence in Corticeira Amorim: reinventing the future

At Amorim Cork's factories, where around 25 million cork stoppers are produced every day, Artificial Intelligence is no longer a distant promise, nor even a prototype: it's a concrete reality. The first step was taken over 15 years ago, as part of an ambitious technological innovation strategy that is having a profound impact and is continually being updated. Perhaps this is the biggest secret of technological transformation: it's a continuous process, constantly improving, that is never fully complete. That is ultimately the very nature of innovation. Robotics was only the first step. Since 2020 Artificial Intelligence has entered the scene. Initially, machine learning was used to improve the classification of cork stoppers, primarily based on their visual appearance. Using photographs and other data (such as X-ray images), the systems began to learn from large volumes of information, and increased the accuracy, efficiency, and homogeneity of the cork batches, thereby reducing the unpredictability of the process. Prior to the introduction of AI, stoppers were classified using algorithms with thousands of lines of code: "We had codes with more than 10,000 lines. And then only those who wrote them could understand what was there," explains Tiago Pinho, Amorim Cork's industrial director. The machines analysed 2D and 3D images, but the decision depended on 70 adjustable parameters. However this complexity didn't enhance the consistency of the analytical process and it soon became clear that this type of tool, with limited computing power, needed to evolve. The goal was to reduce human inter vention and make the system even more reliable and predictable. "We wanted a project where people only had to change two or three parameters. We didn't want to be dependent on highly skilled labor, because that's a highly scarce resource," he emphasises.

Neural networks offered a solution since they are capable of learning directly from specific examples rather than from pre-programmed rules. To develop this system, Amorim Cork partnered with a start-up company - Sentinel, from Guimarães - that is specialising in Artificial Intelligence using a Microsoft programme that is forging a bond between industry and this type of technology partner. Sentinel is developing a new algorithm model, continuing the work that has already been developed in this area with the company 3Control. Commu nication was initially difficult. It was necessary to reverse roles: the computer engineers came to the factory floor, and Tiago Pinho and his team went to study AI. But this led to change and the training of the algorithm - a continuous, never--ending process - began to deliver results. We are now in the field of deep learning, a branch of machine learning that uses more sophisticated algorithms, capable of replicating, in a simplified manner, the functioning of the human brain. With one difference: unlike humans, the neural network does not forget and is continuously learning. "Our brain cannot add indefi nitely, without overwriting past memories. But the machine can. It increases the available data and reduces error," explains Tiago Pinho. Human evaluation is, by nature, imperfect, and can never achieve absolute consistency, and the repeata bility rate (i.e. the ability to apply the same parameters in the same way) is never higher than 75%. The machine, by contrast, makes the same decision about the same image in 90% of cases. Deviations are drastically reduced, extremes disappear, which makes it possible to achieve the desired level of homogeneity and uniformity.

As Tiago Pinho summarises, "The secret is not in the algorithm. It's in the customer knowing precisely what they want and clearly defining the precise output they want from each machine, A, B and C. The quality and quantity of data is what enables the output to provide a reliable decision." In practice, the introduction of deep learning constituted a genuine paradigm shift, moving from fixed rules, manually entered parameters, and fragile code to a neural network that is capable of producing consistent decisions, in a process of continuous improvement. Amorim Cork has chosen to internalise skills, training its teams to continuously train, evaluate, and improve the models. Today, the system has undergone several versions of incremental improvement, which have drastically reduced human intervention, thereby making the system fairly autonomous. At Amorim Cork's factories, each cork stopper is currently checked by the algorithm four times, for a total of 12 decisions per unit, based on analysis of its visual appearance. The use of AI increases the efficiency and homogeneity of the classification process, reducing the variability associated with human evaluation and freeing up human resources for higher value-added functions. The system is based on continuous learning, that has made it possible for the repeatability of decisions to progress from 60% to 80%. Amorim Cork has already opened this transformative "Pandora's box" but the other business units are now following the same transition process. For example, AI is advancing in Amorim Cork Solutions, with plenty of room to grow, and the impact is already verifiable, especially in specific areas such as sealing, insulation, and the explore sectors. In this case AI is used in a different way, linked to knowledge management and sharing. Sector Knowledge is a strategic initiative by Amorim Cork Solutions that focuses on creating, maintaining, and sharing knowledge within teams. Coordinated by Sofia Paixão, head of Sector Knowledge, the project has implemented two AI-based initiatives within the company. The goal is for each team - in the fields of sales, innovation, or marketing - to have access to the greatest possible knowledge and expertise about the materials, applications, customers, and competition of their sector, and thereby exponentially leverage the core competencies of each BU.

The first initiative consisted of imple menting a chatbot, powered by a specific database for each sector, which allows employees to ask questions about the portfolio, technical data sheets, strategic reviews, and market information and deliver answers in a matter of seconds. The initiative works with knowledge that already exists within the organisation and offers an innovative, easier, and more accessible way to maintain and share internal knowledge across the team. By accelerating and simplifying the flow of communication, this initiative benefits people who are already in the company and facilitates onboarding - ensuring that new employees can learn faster. This tool, available on a permanent basis, makes it possible to find answers to many of the questions that arise on a daily basis about the sector. This initiative is already being implemented in the sealing and insulation area and will soon be extended to other sectors. The second initiative, already available in the sealing, marine and energy, and energy areas, and that has been extended to the remaining areas during the first half of 2026, is more focused on obtaining and maintaining new knowledge, scanning the above all, to optimise the relationship between time and level of knowledge - these tools effectively provide access to documentation and information that would otherwise take us hours to find." web for relevant information about the industry, and bringing it into the company. The goal is to ensure that teams consis tently remain up to date on global news that may affect the business. For this purpose, an Artificial Intelligence agent has been developed which, on a weekly basis, scans a set of websites that have been pre-iden tified by the team and then sends short summaries via Teams, of all the news published that week, as well as a PDF report. This tool delivers considerable time saving, freeing up teams for strategic thinking that is essential to the company. As Sofia Paixão, who is also technical and product manager for the aerospace and energy areas, explains: "Knowledge management is increasingly important in our organisation, in particular to ensure that teams have as much knowledge as possible about the business area in which they operate, and that this knowledge is both quickly acquired (...) and also maintained over the years. In addition to maintaining knowledge, the Sector Knowledge team is responsible for generating new knowledge and aims to ensure that all members of each team systematically receive up-to-date and relevant market information. Both initiatives have used AI and aim,