A decision framework guides an AI agent from a user request to a sequence of actions. Memory retains context and tools provide access to external systems; the framework decides how to reason about the request, which tools to use, and when to use them. Without that guidance, an agent may choose the wrong tool or perform tasks in the wrong order. section 6.2 section 6.2.5
Flight and hotel request
Suppose a user asks for a flight from New York to San Francisco next Tuesday and a hotel near Union Square. The agent must identify two subtasks, retrieve flights with a flight-booking tool, search with a hotel-booking tool, and present the results. It should finalize bookings only after the user approves them; booking the hotel before confirming the flight could be the wrong sequence. section 6.2.5
Two approaches are identified:
- Rule-based decision-making. section 6.2.5
- AI-driven reasoning, including ReAct: reason about the next step, act, observe the result, and continue. section 6.2.5 section 6.5
A sound action sequence does not by itself guarantee an accurate result. For production reliability, tool validation, cross-referencing, and guardrails provide additional checks. section 6.5
