The goal of CTU and Cisco is to jointly explore how agent-based AI can move towards true collective intelligence, enabling agents and humans to share context, build on their knowledge, and solve complex problems together. By combining academic research with practical expertise, this collaboration aims to lay the foundation for systems that not only work together but also truly think together, representing a significant step towards a new generation of artificial intelligence.
CTU Rector Prof. Michal Pěchouček has been involved in the field of artificial intelligence for many years and is one of the pioneers in this field. In his role, he wants to support the use of AI not only in teaching and research, but also plans to establish research and technology partnerships in the field of AI with large technology companies. One of these is the recently signed cooperation with Cisco. “I am pleased to continue our historic collaboration with this technology company. In the field of AI Agents and Agent Computing, CTU has been among the world leaders for almost a quarter of a century, and cooperation with Cisco can therefore bring great results,” notes Prof. Michal Pěchouček.
“AI agents can already independently solve specialized tasks and manage corporate data. They can coordinate, but they cannot yet share intent or context,” says Vijoy Pandey, CEO of Outshift, part of Cisco’s Emerging Technology & Incubator group. “Connecting agents and enabling them to share context is an important first step, but the more challenging task is helping them think together. This is the technological hurdle we are focusing on and will explore further as part of this collaboration.”
A multi-agent system works similarly to a team of human experts solving a task. The result is systems capable of handling complex problems significantly more effectively than a single model. Cisco is a pioneer in a system called the Internet of Agents. It is a set of standards and tools that allow agents to collaborate, communicate securely with each other, and even find each other the best agent for a specific task.
The collaboration will focus on four interconnected areas:
- Long-term agent performance – designing systems that continuously improve and learn from their past experiences, including through test-time training techniques.
- Multi-agent vs. single-agent – systematic research into which types of tasks and domains multiple agents working together produce demonstrably better results than a single isolated model.
- Agent coordination and shared intent – methods by which agents share a common understanding of a goal and better coordinate.
- Memory in multi-agent systems – how to teach large language models to effectively decide what to remember and what to forget under conditions of limited token budgets.
CTU contributes to the collaboration with top researchers, laboratories with a strong background in the analysis of long-term behavior of multi-agent systems and effective training of large language models, and in the field of reinforcement learning. This is an approach where AI systems improve through feedback from their own decisions.
“The Czech Republic has an exceptionally strong technical background and world-class science and research, and CTU is the best proof of this,” says Zuzana Švecová, CEO of Cisco Czech Republic. “Therefore, Cisco has long invested in cooperation with academia and supports top-notch research and education. I believe that the results of this collaboration will be felt not only by the scientific community, but also by companies and institutions that will use agent-based AI in practice.”