Overview
Agentic workflows move away from 'zero-shot' prompts toward iterative processes. Instead of asking an AI to do something in one go, you give it the tools and the autonomy to work through the problem step-by-step.
Key Characteristics
- Iterative: The model can review its own work and make corrections.
- Tool Use: The model can call external APIs, run code, or search databases.
- Planning: The model breaks down complex goals into manageable sub-tasks.
- Reflection: The model evaluates its progress and adjusts its strategy.
Benefits
Significantly improves the performance of LLMs on complex tasks like coding, research, and data analysis.