Agentic Architectural Patterns for Building Multi-Agent Systems

3

The Spectrum of LLM Adaptation for Agents: RAG to Fine-tuning

In the previous chapter, we explored the critical considerations for selecting and deploying LLMs, establishing them as the cognitive engines that drive the reasoning and decision-making capabilities of AI agents. We saw how an LLM's capacity for understanding, planning, and tool use is fundamental to an agent's ability to perform its designated functions effectively.

However, the journey to creating a truly proficient AI agent rarely ends with selecting a powerful, general-purpose LLM. The initial wave of generative AI focused on monolithic frontier models, but the landscape is evolving toward a distributed society of intelligent agents. To unlock business value, these agents must ...