Amazon Bedrock Knowledge Bases Exam Question Explained
This AWS AI Practitioner question is a managed RAG trap. If the scenario says S3 documents, chunking, embeddings, vector storage, retrieval, or citations, the answer is Amazon Bedrock Knowledge Bases.
Short answer
The correct answer is B. Amazon Bedrock Knowledge Bases.
Amazon Bedrock Knowledge Bases is the correct answer because the scenario describes the complete managed RAG workflow. The team has source documents in Amazon S3, wants Bedrock to chunk the documents, create embeddings, store vectors, and retrieve relevant passages at query time. That is exactly what Knowledge Bases provides. You connect a supported data source, sync it into a knowledge base, select the embedding and vector-store configuration, then call Retrieve or RetrieveAndGenerate from the application. The key exam phrase is not just "Bedrock"; it is the sequence of document ingestion plus retrieval. Agents are tempting because an agentic app might later use a knowledge base, but Agents answer the orchestration question: which tool or API should I call next? Guardrails answer the safety question: what content should be blocked or filtered? Provisioned Throughput answers the capacity question: how do I reserve model invocation throughput? None of those services owns the chunking, embedding, vector storage, retrieval, and citation loop.
The Question
A team wants a managed way to connect Amazon Bedrock to their documents in Amazon S3, chunk and embed them, store the vectors, and handle retrieval at query time without writing custom pipeline code. Which feature fits best?
Why B is correct
Amazon Bedrock Knowledge Bases is the correct answer because the scenario describes the complete managed RAG workflow. The team has source documents in Amazon S3, wants Bedrock to chunk the documents, create embeddings, store vectors, and retrieve relevant passages at query time. That is exactly what Knowledge Bases provides. You connect a supported data source, sync it into a knowledge base, select the embedding and vector-store configuration, then call Retrieve or RetrieveAndGenerate from the application. The key exam phrase is not just "Bedrock"; it is the sequence of document ingestion plus retrieval. Agents are tempting because an agentic app might later use a knowledge base, but Agents answer the orchestration question: which tool or API should I call next? Guardrails answer the safety question: what content should be blocked or filtered? Provisioned Throughput answers the capacity question: how do I reserve model invocation throughput? None of those services owns the chunking, embedding, vector storage, retrieval, and citation loop.
Why the other options are wrong
Amazon Bedrock Agents is for tool use and multi-step orchestration. An agent might call APIs or decide which action group to invoke, but it is not the feature that chunks S3 documents, creates embeddings, stores vectors, and runs retrieval.
Amazon Bedrock Guardrails is for input and output safety controls such as denied topics, harmful content filtering, and PII redaction. Guardrails do not build a searchable document index.
Provisioned Throughput is a model capacity option for predictable traffic and cost. It changes how model invocation capacity is reserved; it does not create a RAG pipeline over S3 documents.
The exam rule for Bedrock Knowledge Bases
For AIF-C01, think of Amazon Bedrock Knowledge Bases as the managed RAG answer. Retrieval-augmented generation means the application retrieves relevant chunks from private or domain-specific documents before the foundation model answers. That pattern is useful when facts are private, frequently updated, too large to fit in a prompt, or must be cited back to source material. Knowledge Bases packages the common RAG plumbing into a Bedrock feature: connect the data source, configure chunking, choose an embedding model, store the vectors, retrieve matching chunks, and optionally generate a grounded answer. The exam usually hides this in a service-selection question where every option is a real AWS feature. If the wording is about documents and retrieval, choose Knowledge Bases. If the wording is about action groups, API calls, or multi-step task execution, choose Agents. If the wording is about blocked topics, PII, harmful content, or runtime safety policies, choose Guardrails. If the wording is about reserved capacity for predictable traffic, choose Provisioned Throughput. Knowledge Bases does not fine-tune the foundation model; it supplies retrieved context so the model can answer from current data.
Bedrock Knowledge Bases exam trap table
| Scenario clue | Choose | Avoid | Why |
|---|---|---|---|
| S3 documents, chunking, embeddings, vector store, retrieval, citations | Amazon Bedrock Knowledge Bases | Agents, Guardrails, Provisioned Throughput | The scenario is asking for managed RAG infrastructure, not tool orchestration, safety filtering, or reserved model capacity. |
| Multi-step workflow, action groups, API calls, choosing which tool to invoke | Amazon Bedrock Agents | Knowledge Bases | Agents orchestrate actions. They may use a knowledge base, but they are not the document-ingestion and vector-retrieval feature. |
| Denied topics, harmful content, PII redaction, input/output safety policy | Amazon Bedrock Guardrails | Knowledge Bases | Guardrails control runtime safety behavior. They do not chunk documents, embed them, or retrieve source passages. |
| Predictable traffic, reserved model capacity, throughput commitment | Provisioned Throughput | Knowledge Bases | Provisioned Throughput changes invocation capacity and cost behavior. It does not create a searchable document corpus. |
Ready to see how you'd score?
Take the free practice quiz and find out which AWS Certified AI Practitioner domains you need to focus on. No signup required.
Practice 5 similar questions
Same cert, same or adjacent domain. Use these after reviewing the explanation.
Related AWS Certified AI Practitioner Practice Questions
Quick FAQ
When should I choose Amazon Bedrock Knowledge Bases on AIF-C01?
Choose Knowledge Bases when the question describes private or S3 documents, chunking, embeddings, vector storage, retrieval, citations, or a managed RAG pipeline with low operational overhead.
Is Amazon Bedrock Knowledge Bases the same as Bedrock Agents?
No. Knowledge Bases retrieves relevant document chunks for grounded answers. Agents orchestrate tasks, call APIs, and use action groups. An agent can use a knowledge base, but the two features solve different exam clues.
Does Knowledge Bases fine-tune the foundation model?
No. Knowledge Bases is retrieval-augmented generation, not fine-tuning. It keeps the model unchanged and supplies retrieved source context at query time.