specialised agents collaborate in a coordinated manner to solve complex problems. Each agent performs a specific function, whether analysis, planning, supervision or execution, and operates in synchronisation with the others to achieve a shared objective. This structure enables continuous data processing, real-time decisionmaking and a transition from a purely assistive role to the autonomous execution of complete tasks in industrial environments.
As Aizea Lojo, Head of the Artificial Intelligence and Generative AI Research Team at IKERLAN, explains,“ Agentic AI works like a team of specialists coordinated to achieve a common goal.” The platform’ s design allows this coordinated-agent logic to be applied across multiple industrial sectors, facilitating scalability and technology transfer. In practice, this may translate into automatically generating a quotation upon receiving an email, triggering product replenishment in retail environments through computer vision, or assisting in the resolution of technical incidents without manual intervention in repetitive tasks.
GALAXIA represents a conceptual shift beyond traditional conversational AI models.
Jose A. Lorenzo, Data Area Manager at Lantek, underlines the strategic focus of the project:“ Artificial intelligence is not intended to replace operators. Its role is to take over automatable and analytical tasks so that professionals can focus on supervision, decision-making and continuous improvement.”
GALAXIA is being developed in compliance with the requirements established by the European Artificial Intelligence Regulation( AI Act), which obliges companies to adapt their systems to new standards in security, transparency and governance. The multi-agent architecture designed by IKERLAN incorporates mechanisms that clearly define what each agent can do, in which context, and with what level of autonomy. The consortium also includes companies from strategic sectors such as Eroski, Mondragon Assembly, Goizper, Xabet, Ubikare, Lis Data and Fagor Arrasate, which are validating the technology through specific use cases.
For further information, please visit www. lantek. com
Issue 80 PECM 17