Multi-agent systems can strengthen an organisation’s combined knowledge and capability by bringing together specialised agents that operate collaboratively and continuously. If designed well, such systems can create added value, improve operational outcomes and contribute to greater cost efficiency. At the same time, many companies still find it difficult to benefit from AI in practice. A major reason is trust: organisations need confidence in AI-based systems before granting them autonomy. This project is therefore shaped by the specific needs of Swedish industry.
Project focus
The work centres on several key questions that must be addressed if multi-agent systems are to be deployed in a robust and useful way:
- Trust and use
Users need to place confidence in the systems at the right level. Too little trust can limit adoption and value, while too much can result in flawed outputs being accepted without sufficient scrutiny. - Evaluation over time
Methods are needed to assess both the quality of individual agents and the behaviour of the full system as it evolves. - Security and delegation
The project addresses issues such as attacks, trust escalation across agent boundaries, permissions, delegation, and the risks of false consensus and inter-agent influence.
Purpose and expected value
The project will contribute knowledge about how systems of agents with different capabilities can be coordinated in innovation and production processes. This involves both developing the individual components and understanding how they can be combined into complete systems within organisations.
By taking part in the development and evaluation of core capabilities, participating companies build skills, improve their ability to manage risks and strengthen their readiness to use multi-agent systems in practice. The knowledge developed in the project will then be disseminated within the participating organisations.
The project also aims to:
- Support implementation
Five organisations will have implemented multi-agent systems to strengthen business or innovation processes. - Develop transferable knowledge
Core components and knowledge that can be shared will be developed and disseminated.
Societal benefit and sustainability
The project will also create ways for industry partners to mobilise, share lessons learned and develop generic solutions together. By evenly distributing the risks, resources and lessons learned, the project may contribute to sustainable AI-development, from a economical, environmental and societal perspective. A collective knowledge repository makes it possible for more actors to share and use multiagent systems, which reduces the risk of unnecessary parallel development.
