AI Agent
An intelligent system that autonomously sets goals and uses tools (APIs, the web, etc.) to perform complex tasks. Moving beyond simply generating answers, it repeats a reasoning, planning, and execution loop to solve problems on its own.
Detailed explanation
Why it matters for tool selection
AI agents are tools for 'problem-solving' rather than 'knowledge retrieval.' When businesses want to automate actual operations (reservations, payments, data processing) beyond simple consulting, distinguishing whether a tool is a model with simple chatbot functions or an 'agentic' system equipped with external API integration and self-correction capabilities is key to determining return on investment (ROI).
What to check when selecting a tool
- Tool-use capability: Can it call specific APIs or functions with the correct parameters?
- Planning and adaptation: Can it decompose complex goals into subtasks and propose alternatives upon failure?
- Long-term memory support: Does it remember past execution results or user preferences to reflect them in subsequent tasks?
- Guardrails and control: Are there features to limit the agent's permission scope and require human approval at critical steps?
Key use cases
If you ask a travel agent for a '3-day, 2-night trip to Jeju Island,' the agent autonomously performs the entire process: finding attractions via web search (Plan), querying flight/accommodation APIs (Act), revising the itinerary if the budget is exceeded (Observe/Iterate), and ultimately sending a booking confirmation email.
Commonly confused terms
LLM (Large Language Model)
The engine that serves as the 'brain' of the agent. While it understands language and predicts the next text, it cannot use tools or alter the external environment on its own.
RAG (Retrieval-Augmented Generation)
A technology that generates answers by 'referencing' external information. The agent utilizes RAG as one of its tools to make more accurate decisions.