Next-Generation AI Agents for Real-World Applications in the Apply AI sectors (RIA) (Partnership in AI, Data and Robotics)

Expected Outcome

While today’s AI agents still have limited capabilities, advances in model architecture, memory, reasoning, and autonomous behaviour are paving the way to unlocking their potential across economic sectors. The Apply AI Strategy[1]acknowledges this trend and the need to advance research on next-generation AI agents. Project results are expected to contribute to some of the following expected outcomes:

• Significant improvements in the autonomy, robustness and reliability of AI agents through advanced planning mechanisms, memory management, and reasoning capabilities.
• Innovative multi-agent frameworks and protocols demonstrating effective decentralized coordination and collaboration among multiple AI agents beyond the capabilities of individual agents.

Scope

Next-generation AI agents are autonomous systems powered by large AI language models (e.g., large language models or large multimodal models), that can plan, utilize tools and perform actions autonomously to achieve specified goals based on high-level instructions. The large AI model acts as the agent's "brain," capable of interpreting instructions, generating plans, and using tools. This capability enables agents to autonomously plan and adapt behaviour in real-time to accomplish complex, multi-step tasks. AI Agents hold significant promise in numerous applications areas such as data analytics and coding.

Effective AI agents require careful design, incorporating structured planning and reasoning methods to manage complex tasks, and be equipped with appropriate validation and monitoring techniques. Multi-agent collaboration frameworks further enhance capabilities by enabling structured interactions among multiple agents.

Key aspects in designing effective AI agents include robust planning, reasoning, and search mechanisms that allow agents to approach complex tasks by breaking them down into structured subgoals. Effective memory and state management are necessary to maintaining coherent long-term interactions, achieved through a balanced integration of short-term and external long-term memory solutions. Moreover, integrating external tools and APIs is essential for overcoming limitations in large AI models, enhancing agent performance in tasks requiring accuracy and reliability.

Potential research areas include enhancing AI agent autonomy through advanced self-planning and self-optimization capabilities, enabling agents to improve their decision-making and strategic planning. Other research directions include innovation in memory-augmented AI agents to facilitate robust long-term reasoning and lifelong learning; developing advanced multi-agent frameworks specifically tailored for collaborative agents, including research on AI agent frameworks based on mixed AI architectures, and advancing multimodal reasoning capabilities to enable real-world applications.

All proposals are expected to incorporate mechanisms for assessing and demonstrating progress, including qualitative and quantitative KPIs, benchmarking, and progress monitoring. When possible, proposals should build on and reuse public results from relevant previous funded actions. Communicable results should be shared with the European R&D community through the AI-on-demand platform.

Projects selected in this topic should link to the resources offered by the AI Factories, including the Data Labs. The results may be validated in the Testing and Experiment Facilities and further deployed via the European Digital Innovation Hubs (EDIHs) and will contribute to the Apply AI strategy.

This topic implements the co-programmed European Partnership on AI, data, and robotics (ADRA), and all proposals are expected to allocate tasks for cohesion activities with ADRA and the CSA HORIZON-CL4-2025-03-HUMAN-18: GenAI4EU central Hub. Proposals should also build on or seek collaboration with relevant projects and develop synergies with other relevant International, European, national, or regional initiatives.

Activities are expected to start at TRL 2 and achieve TRL 5 by the end of the project - see General Annex B.

[1] COM xxxx(2025) Apply AI Strategy - PLACEHOLDER