Generative AI on AWS FAQ [Guide]

Generative AI is fundamentally changing business operations, from automating content and code creation to reinventing customer service and business intelligence. Amazon Web Services (AWS) provides the essential tools for implementing this powerful technology securely and at scale. 

This Frequently Asked Questions (FAQ) guide is your starting point. Authored by Ascendient Learning's (part of Accenture) AWS expert Myles Brown, this resource demystifies core concepts like Foundation Models (FMs), Retrieval Augmented Generation (RAG), and prompt engineering. You will receive clear explanations of key AWS services, including Amazon Bedrock and Amazon SageMaker. Whether your goal is to quickly deploy an intelligent chatbot or build a custom machine learning workflow, this guide offers the foundational knowledge to build and scale secure, enterprise-grade Generative AI applications on AWS. We also have an AI on AWS playlist with short videos with Myles breaking down some of these tools and concepts.

Please contact us if you have any more specific questions or if you'd like to delve deeper into any of these areas. We are here to help you tailor a Cloud training program that meets your exact needs. To view our courses you can browse our AI/ML on AWS training offerings.

Table of Contents:

  1. What is Amazon Bedrock? 
  2. How does Amazon Bedrock differ from Amazon SageMaker? 
  3. What are Foundation Models (FMs) and which ones are available on AWS? 
  4. What is Retrieval Augmented Generation (RAG) and how does it work with AWS? 
  5. How can I ensure my data is secure when using Generative AI on AWS? 
  6. What are the key use cases for Generative AI on AWS? 
  7. How do I get started with building a Generative AI application on AWS? 
  8. What is the difference between fine-tuning and RAG for customizing a model? 
  9. What is an AI agent on AWS? 
  10. What is Amazon Bedrock AgentCore? 
  11. How does prompt engineering work with Amazon Bedrock? 

  1. What is Amazon Bedrock?

    Amazon Bedrock is a fully managed service that provides a single API to access a wide range of high-performing Foundation Models (FMs) from Amazon and leading AI companies like Anthropic and Meta. It simplifies the process of building and scaling generative AI applications by providing a serverless experience, so you don't have to manage any infrastructure.

    Bedrock also includes a comprehensive set of capabilities for building enterprise-grade applications with built-in security, privacy, and responsible AI practices.

    For more information, watch our short video on Amazon Bedrock.

    To learn about developing application with Bedrock, check out Developing Generative AI Applications on AWS.

  2. How does Amazon Bedrock differ from Amazon SageMaker?

    Amazon SageMaker is a complete, end-to-end platform for machine learning (ML) development, providing deep, granular control over every step of the ML lifecycle, including training, tuning, and deploying your own models.

    Amazon Bedrock, on the other hand, is a higher-level service focused specifically on Generative AI. It offers a choice of pre-trained FMs, allowing you to quickly build applications without managing the underlying ML infrastructure. 

    While SageMaker is for deep, custom ML workflows, Bedrock is for fast and accessible Generative AI development.

    View the free video tutorial.

    To learn about SageMaker and building ML workflows using SageMaker, check out Amazon SageMaker Studio for Data Scientists and MLOps Engineering on AWS.

  3. What are Foundation Models (FMs) and which ones are available on AWS?

    Foundation Models are large-scale models pre-trained on vast amounts of data that can be adapted to perform a wide range of tasks.

    On AWS, Amazon Bedrock offers a choice of FMs from leading providers, allowing you to select the best model for your specific use case. This includes models from Amazon (like Amazon Nova and Titan), Anthropic (Claude), Cohere, Meta (Llama), Stability AI (Stable Diffusion), and others.

    This "model choice" philosophy is a core part of the AWS Generative AI strategy.

    To learn the basics of FMs, check out Generative AI Essentials on AWS.

  4. What is Retrieval Augmented Generation (RAG) and how does it work with AWS?

    Retrieval Augmented Generation, or RAG, is a technique that enhances a Foundation Model's response by retrieving relevant data from a private knowledge source and including it in the prompt.

    RAG allows the model to generate more accurate and contextually relevant answers based on your own data. On AWS, you can easily implement RAG using Knowledge Bases for Amazon Bedrock, which connect your FMs to your data stored in sources like Amazon S3.

    We have a short video on RAG for more information.

    This is a critical capability for building accurate, domain-specific AI applications. To learn about the basics of RAG, check out our course, Generative AI Essentials on AWS.

  5. How can I ensure my data is secure when using Generative AI on AWS?

    AWS prioritizes data privacy and security. With services like Bedrock, your data is never used to train the underlying Foundation Models. Your data remains in your AWS environment, with encryption in transit and at rest.

    AWS also provides features like Bedrock Guardrails, which allow you to define safety policies and block harmful or off-topic content in your AI applications. You maintain full ownership and control over your content.

    To learn about securing your Generative AI applications, check out Developing Generative AI Applications on AWS.

  6. What are the key use cases for Generative AI on AWS?

    Generative AI on AWS is being used to reinvent a wide range of business processes. Common use cases include:

    • Customer Service: Building intelligent chatbots and virtual assistants.
    • Content Creation: Generating marketing copy, product descriptions, and summaries.
    • Code Generation: Accelerating software development with AI-powered assistants like Amazon Q Developer.
    • Business Intelligence: Creating dashboards and reports from natural language queries.
    • Personalization: Delivering highly personalized experiences and recommendations.

    To learn about Generative AI use cases, check out Generative AI Essentials on AWS.

  7. How do I get started with building a Generative AI application on AWS?

    The fastest way to get started is by using the Amazon Bedrock console. The console provides an interactive playground where you can experiment with different Foundation Models, try out various prompts, and see the results instantly.

    You can then use the AWS SDKs to integrate the models and their capabilities into your own applications, starting with a few lines of code.

    To learn about developing application with Bedrock, check out Developing Generative AI Applications on AWS.

  8. What is the difference between fine-tuning and RAG for customizing a model?

    • Fine-Tuning: This involves taking a pre-trained FM and training it on a smaller, labeled dataset to adapt its internal parameters for a specific task or style. This is best for teaching the model a new skill or a specific tone.

    • RAG (Retrieval Augmented Generation): This technique doesn't change the model itself. Instead, it provides the model with new data as an input at inference time. This is best for keeping a model's knowledge up-to-date with your proprietary data without the need for expensive retraining.

    This quick video breaks down the differences between fine-tuning and RAG.

    To learn more about customizing models, check out Developing Generative AI Applications on AWS.

  9. What is an AI agent on AWS?

    In the context of AWS, an AI agent is an autonomous application built on Amazon Bedrock that can execute a multi-step task for a user. For example, a travel-booking agent could process a user request, search for available flights from your data sources, and then book a flight on their behalf.

    Bedrock Agents simplifies the development of these agents, allowing them to perform tasks, converse with users, and access company systems and data.

    To learn about building AI agents, check out Agentic AI Foundations.

  10. What is Amazon Bedrock AgentCore?

    Amazon Bedrock AgentCore helps developers securely deploy and operate highly capable AI agents at scale, regardless of the underlying framework or model. 

    It eliminates the heavy lifting of building and managing specialized agent infrastructure and works with popular frameworks like CrewAI, LangGraph, and LllamaIndex. 

    To learn about AgentCore, check out Agentic AI Foundations.

  11. How does prompt engineering work with Amazon Bedrock?

    Prompt engineering in general is the strategic art and science of crafting the most effective inputs ("prompts") to guide a Generative AI model toward a desired output.

    While the core concept is universal, on AWS, it becomes a hands-on skill practiced across multiple services.

    It's the critical first step in working with Foundation Models (FMs) on Amazon Bedrock. The Bedrock Playground serves as your primary prompt engineering sandbox, where you experiment with different models from providers like Anthropic (Claude), Cohere, and AI21 Labs. You'll quickly learn that a prompt that works for one model may not work as well for another, making the selection of the right FM as important as the prompt itself.

    Prompt engineering is also essential for:

    • Creating Bedrock Agents: You engineer a prompt to give the agent its identity, define its purpose, and specify the instructions it needs to follow to accomplish a task.

    • Implementing Guardrails for Amazon Bedrock: Prompt engineering is used to test and refine the rules you set to ensure the model responds safely and appropriately, without generating harmful or off-topic content.

    • Building RAG Solutions: The prompt engineer directs the model to integrate the information retrieved from your Knowledge Base for Amazon Bedrock to provide an accurate, domain-specific answer.

    Ultimately, on AWS, prompt engineering is the fundamental skill required to leverage the full power of the Generative AI services, tailoring the behavior of FMs to your unique business needs and applications. To learn about the basics of prompt engineering, check out Generative AI Essentials on AWS.