Cost-Effective Strategy for Changing FAQ Answers

AIF-C01 By islever Editorial Last updated

Describe design considerations for applications that use foundation models. Understand the capabilities and limitations of generative AI for solving business problems. Choose effective prompt engineering techniques.

Short answer

D Use Retrieval Augmented Generation with prompt engineering so agents get current FAQ answers from an updated knowledge base without retraining the model.

Question

A company deployed an AI/ML solution to help customer service agents respond to frequently asked questions. The questions can change over time. The company wants to give customer service agents the ability to ask questions and receive automatically generated answers to common customer questions.
Which strategy will meet these requirements MOST cost-effectively?
  1. Fine-tune the model regularly.
  2. Train the model by using context data.
  3. Pre-train and benchmark the model by using context data.
  4. Use Retrieval Augmented Generation (RAG) with prompt engineering techniques. Correct Answer

Key Takeaways

The question tests choosing between RAG and model-training approaches when knowledge is dynamic; the trap is reading "regularly" as a cue for fine-tuning (A) when it actually signals unacceptable recurring cost.

For customer service FAQs that change over time, Retrieval Augmented Generation combined with prompt engineering (D) is the most cost-effective strategy, because knowledge is refreshed in the index instead of by retraining the model.

Common mistake: Option A, fine-tuning the model regularly, is the most tempting wrong answer: fine-tuning adjusts behavior and style rather than reliably injecting new facts, and repeating it for every FAQ change is far more expensive than updating a retrieval index.

Expert Analysis

Why D is correct The scenario has one decisive property: the FAQ knowledge changes over time, and the requirement is to answer automatically at the lowest cost. Retrieval Augmented Generation (RAG) satisfies both halves. At query time the system retrieves the current FAQ passages from a vector or keyword index and injects them into the prompt, so the answer reflects today's content without touching the model weights. Prompt engineering then shapes the retrieved context into a clear, on-tone answer for the customer service agent. This is exactly what option D describes, and it is the pattern AWS positions as the default for knowledge-intensive, frequently updated enterprise Q&A.

The adjacent decision: A, B and C Option A, regular fine-tuning, would make sense only if the goal were to change the model's style, format or domain behavior — not to refresh facts. Fine-tuning is expensive per run, requires curated examples, and still does not reliably inject new factual knowledge, so re-running it whenever the FAQ changes is the least cost-effective choice here. Option B, training the model with context data, has the same flaw and additionally implies prohibitive compute and data-labeling cost for a general-purpose foundation model. Option C, pre-training plus benchmarking with context data, is a one-off evaluation activity for a model you build or select; it neither updates live knowledge nor gives agents real-time answers, and benchmarking is an evaluation step, not a delivery strategy.

What would flip the answer If the question instead asked how to make the model consistently adopt a fixed brand voice or a specific output schema that never changes, fine-tuning (A) would become the right answer and RAG would be unnecessary. Likewise, if the retrieved knowledge base were absent and only a static document set existed, prompt engineering alone could suffice. The word "regularly" combined with "change over time" is the trap: it tempts you toward A, but it actually disqualifies A on cost grounds.

Operational consequence Choosing A or B means repeating costly training jobs for every FAQ edit, with latency and downtime for agents, while stale answers persist until retraining completes. With D, updating an entry in the knowledge base propagates immediately; getting it wrong burns budget and leaves agents serving outdated answers.

Official Reference

These links open the official vendor documentation in a new tab.

Exam Strategy

When a question stresses changing knowledge plus "most cost-effective," shortlist retrieval-based options before any training option: RAG updates a data source, while training and fine-tuning imply recurring cost and latency. Reject any option whose cost scales with how often the content changes. Confirm the choice by checking that the option also covers answer generation, not just evaluation or one-time preparation.

Frequently Asked Questions

Why is RAG more cost-effective than fine-tuning for frequently changing FAQs?

RAG retrieves up-to-date passages from an external knowledge base at query time, so updating an FAQ entry costs only a re-index. Fine-tuning requires collecting examples, training compute and redeployment each time the content changes.

What role does prompt engineering play alongside RAG in this solution?

Prompt engineering frames the retrieved FAQ context and instructs the model on tone, length and format, improving answer accuracy and consistency without any model retraining.

When would fine-tuning actually be the right choice instead of RAG?

Use fine-tuning when you need stable changes to behavior, style, tone or output structure that persist across all queries. If the need is to keep facts current, RAG is preferable.

More AIF-C01 FAQ →

Community Votes

D
100%

100% of anonymous learners picked answer D. Votes are pick records left by other test-takers — they are not the verified answer.

Discussion 5 contributions

Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.

From public community threads 5

  1. Rcosmos 👍 4 Answered D
    se a Geração Aumentada de Recuperação (RAG) com técnicas de engenharia rápidas. A abordagem RAG combina modelos de geração de linguagem com um sistema de recuperação de informações. Isso permite que os agentes obtenham respostas geradas automaticamente com base em perguntas e dados atualizados sem precisar ajustar ou treinar o modelo continuamente. É uma solução eficiente, especialmente quando as perguntas dos clientes mudam frequentemente, pois utiliza técnicas para buscar informações de fontes relevantes em tempo real.
  2. teeinwma53 👍 3 Answered D
    Fine-tuning or training the model again costs a lot of money and time. RAG lets the model search for the latest information from a database or document before generating an answer. So when something changes, you only need to update the data, not retrain the model. Prompt engineering helps make the model’s responses clearer and more accurate.
  3. Aamee 👍 2 Answered D
    Exam tip: When a question mentions knowledge that changes frequently and asks for the most cost-effective solution, the answer is often RAG rather than fine-tuning or retraining.
  4. 65703c1 👍 2 Answered D
    D is the correct answer
  5. NareshGondi 👍 1 Answered D
    Frequently asked questions change over time. RAG pulls the most up-to-date FAQ data from an external knowledge base

Reader comments & corrections

No reader corrections yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

How to use a custom trained model in Amazon Bedrock?

S-grade

How to Increase Image Specificity in Amazon Bedrock Stable Diffusion?

S-grade

How to Quickly Deploy and Consume Foundation Models in a VPC on AWS?

S-grade

Which AWS Service Sends Email Notifications When ISV Compliance Reports Are Available?

S-grade

Practice All AIF-C01 Questions

Access 100 questions with complete answers and detailed explanations.

View Full AIF-C01 Practice Test →

← Back to AIF-C01 Study Guide