Agentic Super Intelligence (ASI)

As AI systems become increasingly capable, the next frontier lies in developing intelligence that can autonomously reason, plan, learn, and collaborate to solve complex real-world problems. Achieving this requires a shift from standalone models to integrated systems that can continuously adapt, improve, and operate effectively across diverse domains and environments.

The Agentic Super Intelligence (ASI) pillar explores the foundations of open-ended, self-improving, introspective, and real-world-grounded intelligence. Rather than treating superintelligence as a single model or an achieved capability, we focus on developing increasingly capable agentic systems that can learn, reason, plan, act, and collaborate across domains. These systems are designed to continuously acquire knowledge, adapt through experience, and assess their own capabilities and limitations while remaining controllable, reliable, and trustworthy.

By grounding AI in an understanding of both the physical and societal world, we aim to enable effective human-AI collaboration, support complex decision-making, and advance towards more capable forms of intelligence that augment human expertise and address real-world challenges.

Our research advances the foundations of agentic intelligence through the following areas:

Research Focus

Human Digital Twins & Trustworthy Interaction

Developing AI systems that can understand, adapt to, and collaborate effectively with people over extended interactions. By modelling human expertise, goals, preferences, and feedback, we aim to create trustworthy human-AI partnerships that enhance decision-making while preserving human oversight and control.

Omni-Tasking & Cross-Domain Generalisation

Advancing AI systems that can operate across diverse tasks and domains, transfer knowledge between contexts, and compose specialised capabilities to solve increasingly complex problems. This research seeks to enable flexible and adaptable intelligence that extends beyond narrow task-specific performance.

World Models for Prediction, Planning & Anticipation

Building AI systems that develop rich representations of the physical and societal world, enabling them to forecast outcomes, evaluate alternatives, anticipate risks, and make informed decisions under uncertainty.

Scalable Agentic AI and Organisational Intelligence

Developing coordinated teams of AI agents that can collaborate, delegate, communicate, and learn collectively. Our goal is to enable organisational-level intelligence, where multiple agents work together as a coherent and continuously improving system.

Verification, Alignment & Self-Correction

Creating AI systems that can verify their own reasoning, monitor their behaviour, detect failures, and correct mistakes before taking action. By embedding verification and alignment into the decision-making process, we aim to ensure that increasingly autonomous systems remain reliable, trustworthy, and aligned with human goals.

Application

We apply and evaluate our research across a range of complex, high-impact domains that place different demands on intelligence, autonomy, and human-AI collaboration. These application areas serve as real-world testbeds for developing, refining, and validating a common foundation of agentic capabilities, ensuring that advances in ASI can be adapted safely and effectively across diverse sectors.

Materials Science, Surfaces and Catalysis

Medtech

Developing AI systems that support clinical decision-making, longitudinal human-AI interaction, and safety-critical recommendations, enabling more effective and trustworthy healthcare delivery.

Hub of the Future

Creating intelligent systems that operate within complex physical-social environments, supporting mobility, infrastructure, public services, and adaptive urban operations through anticipation, planning, and coordination.

Manufacturing

Enabling autonomous planning, process optimisation, and intelligent operations across manufacturing environments, improving efficiency, resilience, and adaptability while maintaining safety and reliability.

Digital Services

Building AI agents that support knowledge work, organisational workflows, and service delivery, enabling more effective collaboration, tool use, and decision-making across complex digital ecosystems.

Semiconductor

Developing AI systems that enhance engineering design, analysis, verification, and scientific discovery, helping to address increasingly complex semiconductor research and development challenges.

Collaborate With Us

The rapid emergence of general-purpose and agentic AI creates new opportunities for collaboration across research, industry, and the public sector. Drawing on strengths in self-correcting agents, self-teaching language models, multi-agent learning, embodied AI, and causal reasoning, the ASI pillar provides a platform for advancing the next generation of AI systems. We welcome partnerships through joint research, innovation programmes, and talent development initiatives to translate frontier AI advances into broad societal and economic impact.