The Geometry of Agentic Design
Published in
6 min readAug 26, 2024
Agentic design intentionally structures an AI agent’s capabilities, interactions, and decision-making processes.
It’s the blueprint that determines how an agent will perceive, reason, and act. As AI systems grow increasingly complex and autonomous, understanding agentic design principles becomes essential.
It’s no longer just an academic exercise — it’s a necessity for anyone at the forefront of technology.
Agentic design is built on ten key components, each shaping how AI agents perceive, reason, and act:
- Input Modality: Defines how agents process information, from text to multi-sensory data.
- Access to Underlying Models: Determines the AI “brains” powering agent intelligence.
- External Capabilities: Enables interaction with tools, databases, and APIs.
- Role-Playing: Assigns specific functions within multi-agent systems.
- Goal Type: Sets objectives, from simple tasks to complex adaptive goals.
- Planning: Formulates strategies to achieve defined goals.
- Action: Executes plans and interacts with the environment.
- Reflection: Evaluates performance and incorporates feedback.
- Learning Capability: Improves agent performance over time.
- Non-Functional Aspects: Addresses autonomy, control, and ethical considerations.