The study reframes agentic AI for enterprise teams, shifting the focus from model selection to architectural decisions. This involves considering factors such as training budget allocation, modularity preservation, and the trade-offs between cost, flexibility, and risk. The researchers identified two primary dimensions within the agentic AI landscape: agent adaptation and tool adaptation. Agent adaptation involves modifying the foundation model that underpins the agentic system by updating its internal parameters or policies.

The rapid growth of agentic AI has led to a proliferation of tools and frameworks, making it difficult for developers to determine the best options for their specific needs. Agentic AI refers to AI systems capable of autonomous action to achieve specific goals. These systems often combine large language models with planning, memory, and tool-use capabilities.

The framework is expected to provide clarity and direction for developers, enabling them to make more informed decisions about the design and implementation of agentic AI systems. By categorizing frameworks based on their strengths and weaknesses, the researchers hope to streamline the development process and accelerate the adoption of agentic AI across various industries. The researchers suggest that understanding the trade-offs between agent and tool adaptation is crucial for optimizing the performance and efficiency of agentic AI systems.