Multi-agent AI architecture for industrial digital transformation

Lantek, a multinational specializing in digital solutions for the metal industry, and IKERLAN, a Basque technology centre focused on applied research, are collaborating on GALAXIA, an R&D initiative aimed at creating an industrial agentic artificial intelligence architecture. The project focuses on multi-agent systems capable not only of analysing data but also executing coordinated actions within real production environments, advancing the integration of intelligent automation into industrial operations.

Context of the collaboration
Both organizations are active members of the Basque Artificial Intelligence ecosystem and part of BAIC (Basque Artificial Intelligence Center). The project is funded through the Basque Government’s HAZITEK programme and seeks to strengthen regional technological sovereignty by supporting locally developed advanced AI capabilities. By developing AI technologies within a controlled ecosystem, the initiative aims to reduce reliance on external platforms while maintaining governance over critical industrial data and processes.

The consortium also includes companies from strategic sectors—Eroski, Mondragon Assembly, Goizper, Xabet, Ubikare, Lis Data, and Fagor Arrasate—which contribute industry-specific use cases to validate the technology under real operational conditions.

Technical solution and responsibilities
GALAXIA is designed as a multi-agent AI architecture aligned with the European Artificial Intelligence Regulation (AI Act), incorporating requirements related to security, transparency, and governance. The architecture defines clear operational boundaries for each AI agent, specifying its permitted actions, contextual limitations, and level of autonomy.

Lantek serves as project coordinator, contributing expertise in industrial software, advanced data utilization, and intelligent assistants for production environments. IKERLAN acts as technological lead, applying its multi-disciplinary capabilities in intelligent systems and industrial research to develop the core agentic architecture.

From conversational AI to action-oriented systems
The project represents a shift from traditional conversational AI toward agentic AI capable of performing coordinated actions. Instead of responding only to queries, specialised agents collaborate as structured teams, performing analysis, planning, supervision, and execution tasks simultaneously.

This coordinated approach enables continuous data processing, real-time decision-making, and automated execution of complex workflows. Agentic AI can support industrial processes by automating tasks such as generating quotations from incoming emails, triggering product replenishment using computer vision in retail environments, or assisting in the resolution of technical incidents without manual intervention.

Implementation and expected impact
The modular design allows deployment across multiple industrial sectors, supporting scalability and technology transfer. By automating analytical and repetitive processes, the platform aims to enhance operational efficiency while allowing human operators to focus on supervision, decision-making, and continuous process improvement.

The project demonstrates how collaborative development between industrial software providers, research centres, and end users can accelerate the adoption of advanced AI within industrial automation frameworks while aligning with emerging regulatory standards.

www.lantek.com