Amazon Bedrock for generative AI — AIF-C01

WHAT IT IS

Amazon Bedrock is a fully managed service that provides secure, enterprise-grade access to high-performing foundation models from leading AI companies, enabling you to build and scale generative AI applications.

Because it is fully managed, you interact with foundation models through APIs — you do not provision servers, manage GPU clusters, or operate model-serving infrastructure.

Mental model

Think of Bedrock as a managed switchboard for foundation models. Your application sends an API request; Bedrock routes it to the model you selected; the response comes back. The underlying compute, the model weights, and the serving infrastructure are all abstracted away. You own your application logic and your data — not the infrastructure beneath the model.

When to use it

The exam tests whether you can match a builder scenario to the right AWS generative AI service. The adjacent service most often placed beside Bedrock is Amazon SageMaker JumpStart, which also offers access to pre-trained models but in a different operational posture.

Dimension	Amazon Bedrock	Amazon SageMaker JumpStart
Infrastructure ownership	None — fully managed by AWS	You deploy models to SageMaker endpoints you manage
Primary access pattern	API calls; no model deployment step required	Deploy a model endpoint, then call it
Operational responsibility	Lower barrier to entry	More control; more operational overhead
Typical builder scenario	Build a generative AI application quickly with enterprise security	Experiment with, fine-tune, and host a wide range of open-source and proprietary models with full endpoint control
AWS exam category	Generative AI application service	ML platform / model hub

The blueprint (Task 2.3) explicitly names both services when listing AWS services and features to develop generative AI applications, so you should be able to distinguish them by their operational posture.

COMMON MISCONCEPTION

Bedrock is not a model you use — it is a service that gives you access to models.

Candidates sometimes treat "Amazon Bedrock" as if it were itself a foundation model (analogous to naming a specific model). Bedrock is the managed service layer; the foundation models (from Amazon, Anthropic, and other providers) are the resources the service makes accessible. This distinction matters: a question asking which AWS service provides managed access to foundation models points to Bedrock, while a question asking which model to use for a particular task points to the foundation models themselves.

A related trap: assuming that using Bedrock means you are training or building a model from scratch. The service is designed so you do not need to train a model or manage infrastructure to get started — that is precisely what "fully managed" and "lower barrier to entry" mean in the blueprint's language about the advantages of AWS generative AI services.

How it shows up on the exam

Task 2.3 asks candidates to identify AWS services and features for building generative AI applications and to describe the advantages of using those services. The cognitive target is recognition and classification — matching a scenario description to the correct service — rather than deep technical implementation.

Scenarios likely to involve Bedrock reasoning include:

• A builder who needs to add generative AI capabilities to an application without managing infrastructure
• A description emphasizing "enterprise-grade," "secure," or "fully managed" access to foundation models
• A scenario where the key advantage stated is lower barrier to entry, speed to market, or accessibility

Candidates who confuse the service layer (Bedrock) with the model layer (foundation models) may misread scenario questions that ask about infrastructure responsibility or about which resource is being accessed through an API call.

Related concepts

• Foundation models — the models that Bedrock provides access to; understanding what a foundation model is helps clarify what Bedrock does and does not do.
• Bedrock Knowledge Bases — a Bedrock capability that connects foundation models to your own data sources for Retrieval Augmented Generation (RAG) workflows.
• Fine-tuning foundation models — one of the customization approaches available within Bedrock for adapting a model's behavior to a specific domain or task.