We have a challenge in Europe. As highlighted in Mario Draghi’s report to the EU commission, we have fallen significantly behind in productivity compared to the U.S. and China, and we should heavily invest to regain technological sovereignty and industry competitiveness. Self-development of AI based systems and their integration into the production lines of our factories is a necessity. The question is no longer whether we should use AI or not, but how to implement it in a smart, safe, and efficient way, with real business value and both fast and long-term benefits. We should start now in case we didn’t do it already yesterday.
AI technology for manufacturing can start with some isolated software working as an agent doing an individual task, maybe using machine vision and other sources of information. But we could aim instead for a complete level of understanding of what is going on within the company walls and be innovative to create a larger number of agents. The hypothesis is: if we achieve a unified understanding of everything happening within the company, from management offices to the shopfloors, it would be possible to create a higher number of agents serving workers on different roles, from management to shopfloor lines.
Factories are huge sources of data, but we still have not managed to utilize the data to its full potential. We have documentation in different forms, archive images and training videos, internal logistics and human resources needs, workers with tacit knowledge and strong experience but also those still learning new tasks, schedules and production requirements, machines running processes, and so on so on. Additionally, we could have cameras to track the progress and actions of workers and internal logistics, but also acoustic, vibration and other sensors to monitor the functioning of all the machines. Even legacy machinery creates data, and that can be also extracted with digitalization techniques based on machine vision.
Developing a system able to connect all those data flows using machine learning techniques would be the way to turn data into information, and then information into the cost-saving knowledge and actions that we look for. The path is to move away from isolated solutions to interconnected real-time systems that could simultaneously help all types of workers to be more efficient and make better decisions. The number of agents we could create after that is large. At shopfloor level, solutions could be cooperative robotic systems, automatic orchestration of machinery, and digital assistance of work tasks. For the managers running the floor, tools could be quality reports based on multilevel sensor analysis, workflow planning and bottleneck prediction, and automatic or at least faster reporting and final archiving of information. At middle and senior management, agents could support strategic planning based on historical production data, resource availability, and monitoring of energy consumption, waste generation, and logistics.
It is up to us whether we start now and aim to take the lead in that global context, or we stay traditional — maybe a bit scared of investing and waiting to see how others do. There are issues related to ethics and personal information that will be solved and many people are afraid that this transformation could take their jobs. Some roles will undoubtedly disappear. Repetitive tasks on the shop floors are easy to automatize, but this applies to office work as well. In most cases the transformation should lead to improvements in working conditions. This aligns with the principles of Industry 5.0, which emphasizes human-centric, resilient, and sustainable industrial development where technology augments but not replaces human workers. The idea is to reach both physical and cognitive augmentation; improve ergonomics, safety, the speed to perform actions, and everything related to decision-taking. Yes, job roles will change, but that occurred too in all previous industrial revolutions, and we know that each one was for the good. In the end, it is not about reducing the number of employees, but about how they can produce more and better. It is also possible that companies that engage early with AI would be better positioned to attract and retain skilled workers in the years to come.
What is the price to introduce AI at the level mentioned above for manufacturing companies? There is a cost, but what is the cost of not doing it? A company can always do their own RDI work and/or purchase from a service provider. But in some cases, especially when aiming for solutions that are not yet market-ready, joining a RDI project run by research institutions can be a good and safe option. The entering price is normally quite low, as those actions are supported by Business Finland, regional development funds, or the EU Commission. We researchers always need companies to be involved, as we do applied research and need use cases. For instance, our Industrial AI solutions group from Turku University of Applied Sciences, starts this autumn two new projects with a total of 12 participating companies. In both we focus exactly on this technology of overall factory understanding and modular agent creation. We are preparing more, as we continuously come up with new ideas as we talk with new partners and other national and European universities. Participation for SMEs and larger companies is simple. The first step is just a conversation. Reach out and tell us how your company works and what your needs are. Let’s innovate together and get it done.
David Oliva
Chief Advisor
Industrial AI Solutions, Futuristic Interactive Technologies research group
Turku University of Applied Sciences (Turun AMK)