Skip to content Accessibility statement
Centre for Assuring Autonomy
HomeCentre for Assuring AutonomyGuidanceINSYTE

INSYTE: A Classification Framework for Traditional to Agentic AI Systems

The INSYTE (INtelligent SYsTEms) classification framework allows for the detailed classification of AI-enabled systems. It can be applied to a broad spectrum of systems, ranging from traditional rule-based systems to cutting-edge embodied AI and agentic systems, in any sector.

INSYTE classifications are visually represented on a radar chart, giving an immediate overview of the system as well as offering considerable insight into its individual characteristics. You are welcome to generate these radar charts (“INSYTE patterns”) for your system using our free online tool. 

The INSYTE framework considers the essential characteristics of an AI-enabled system across eight key dimensions grouped into four categories: system design (underspecification and adaptiveness); functionality (breadth and depth); operating environment (diversity and dynamism); and independence from human operational control (intervention and oversight). These eight dimensions allow for the distinctive characteristics of agentic AI systems to be represented, whilst still being relevant to digital and embodied systems enabled by more traditional AI techniques.

Central to the INSYTE framework is a tool-supported process to help you determine, for each of these eight dimensions, at what level a given AI-enabled system should be described. This process culminates in an “INSYTE pattern” - the visually informative pattern that the system yields on the framework’s eight-axis radar chart. INSYTE patterns represent the combination of the eight characteristics the system instantiates and to what level. In this way, they allow for the complexity of a given system to be considered in the round.

Download the INSYTE paper

A spider diagram with eight points starting at 11 o'clock on the diagram. 1. Underspecification 2. Adaptiveness 3.Breadth of Functionality 4. Depth of Functionality 5. Environmental Diversity 6. Environmental Dynamism 7. Intervention Independence 8. Oversight Independence

INSYTE is: 

  • A classification system for traditional to agentic AI.
  • Designed to support cross-stakeholder communication.
  • Designed to facilitate safety engineering and assurance.

Download the INSYTE Paper

Use the free, online INSYTE tool

Please use the following citation when you include an INSYTE pattern in reports, papers, or presentations:
Zoe Porter, Radu Calinescu, Ernest Lim, et al. (2025). INSYTE: A Classification Framework for Traditional to Agentic AI Systems. ACM Transactions on Autonomous and Adaptive Systems. (August 2025). https://doi.org/10.1145/3760424

Using CfAA's foundations to build INSYTE

The INSYTE framework was developed using a systematic, multi-stage methodology. A preliminary version emerged from insights we gained from our Assuring Autonomy International Programme’s 25 demonstrator projects. We supplemented these findings by considering the key system characteristics highlighted in existing autonomy-based classification frameworks, such as the Levels of Autonomy. To ensure that INSYTE’s dimensions are appropriate for state-of-the-art AI systems, we further refined the framework through consulting the growing body of literature on agentic AI. The INSYTE framework was then evaluated in two rounds by academic and non-academic stakeholders with different levels of experience, different disciplinary backgrounds, and drawn from a wide range of application domains. It has been validated with worked examples of applying INSYTE to systems currently being researched and developed by our authors and research collaborators.

Evaluator feedback indicates that the INSYTE framework has significant potential as a tool to foster cross-stakeholder transparency, understanding, and communication. Its foreseeable uses further extend to supporting decision-making during the design, development, pre-deployment, and post-deployment phases of the system lifecycle. INSYTE’s nuanced, systematic ‘whole system’ perspective can also inform reasoning about safety and risk, thereby providing a platform for robust safety assurance as well as enhanced approaches to regulation and certification.