Prepare and test your skills
Prepare and test your skills
Free sample question, answered and explained below. The Generative AI Leader practice bank has 2,191 more like it. Try 10 free: instant score, no account.
An enterprise is building an intelligent customer service assistant using generative AI on Google Cloud. The assistant must handle complex customer billing inquiries that require performing step-by-step reasoning, executing queries against an external billing database API to fetch real-time transaction records, and evaluating the retrieved data before formulating a response.
Which advanced prompting technique is best suited for combining multi-step reasoning with external tool execution?
Few-shot prompting is an in-context learning technique where several input-output example pairs are included in the prompt to demonstrate the desired format, style, or task execution before the actual query.
While ReAct agents can utilize few-shot examples to learn how to interact with tools, few-shot prompting on its own refers to static demonstration examples rather than an action-oriented agent framework.
Zero-shot prompting is the practice of presenting a model with an instruction or task description directly without providing any demonstration examples, intermediate reasoning instructions, or external tool integrations.
Zero-shot prompting is effective for baseline tasks, but as task complexity increases to include external tools and multi-step deduction, structured prompting frameworks like ReAct become necessary.
Chain-of-Thought (CoT) prompting is a technique that encourages large language models to break down complex problems into a sequential series of intermediate reasoning steps before arriving at a final answer.
Chain-of-Thought prompting forms the foundation of ReAct's "Reasoning" component, but ReAct extends CoT by coupling those reasoning traces with dynamic "Action" invocations.
ReAct (Reasoning and Acting) is an advanced prompting framework that synergizes reasoning traces and task-specific actions in large language models. Rather than operating purely in isolation or generating answers without verification, ReAct enables models to alternate between thinking (reasoning) about what to do next and executing (acting) concrete operations such as interacting with external APIs, databases, or search tools.

Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated