Multiagent systems for modern industrial processes and value chains

About the project

Multi-agent systems can multiply knowledge and capability of an organisation, deliver added value and drive cost reduction and improve operational outcomes, with 24/7 availability. 

At the same time, research shows that companies still find it difficult to profit from AI, as trust remains a major hurdle; businesses require confidence in AI systems before granting them any level of autonomy. The specific needs of Swedish industry have set the direction for this collaborative project.

Challenges

Trust-calibration: systems must be designed so that users neither under-rely on agents (leading to low adoption and missed value) nor over-rely on them (leading to uncritical acceptance of flawed outputs)

Quality: how both individual agents and system-level behaviour can be evaluated and monitored over time as systems evolve.

Security handling: 

• Attacks

• Trust escalation across agent boundaries

• Inter-agent permissions and delegation

• False consensus and inter-agent influence

Project purpose

Multi-agent systems (MAS) have great potential to transform work processes across all organizations, including those in industry. To fully realize this value, we must both develop the individual components and learn how to integrate them into a complete system. While each individual capability provides value, the true impact is achieved only when we understand how to effectively implement these systems within the organization.

By actively participating in the development and evaluation of core capabilities, companies gain the necessary skills, the confidence to benefit from MAS, and the ability to manage risks effectively. The companies participating in the project will then disseminate this knowledge throughout their organizations.

Expected outcomes

The project aims to provide an understanding of how systems of agents with diverse functionalities can collaborate and synchronize to solve complex tasks within innovation and production processes. This includes insights into how these agents can be managed across various use cases and how they can be shared within complex systems across different companies.

Specific goals:

• Implementation: Five organizations will have implemented MAS to strengthen their business or innovation processes.
• Knowledge sharing: Core components and shareable knowledge will be developed and disseminated.

The project will establish frameworks for how industry partners can mobilize and share lessons learned and generic solutions, aligned with AI Sweden’s mindset of "speed and boldness." This is central to ensuring a rapid transition, where the ecosystem shares general insights that each organization can then bring back to develop its own specific solutions. This collaborative approach also contributes to a more sustainable way of working—addressing economic, environmental, and societal aspects—by sharing risks, pooling development resources, and democratizing knowledge.

For more information about this project, please reach out to: