Building an Agentic AI Workflow with Amazon Bedrock and AWS Step Functions

Agent-based AI systems represent the next step beyond traditional chatbots by reasoning about user goals, planning actions, and using tools to complete tasks automatically. With Amazon Bedrock, these systems can be built and managed easily through integrated agents and foundation models that connect with AWS services. Learning how to design and orchestrate these intelligent workflows helps developers create applications that can automate decisions, interact with AWS resources, and adapt based on real-time data.

In this hands-on lab, you will deploy an Amazon Bedrock Agent, connect it to an AWS Lambda function, and build an AWS Step Functions workflow that invokes the agent through the AWS SDK integration. You will then run the workflow and observe how the agent performs its reasoning and action steps while recording results in Amazon DynamoDB and Amazon CloudWatch.

Learning objectives

Upon completion of this beginner-level lab, you will be able to:

• Explain the concept of an agentic AI system and how it expands on traditional chatbot or LLM interactions.
• Describe how reasoning, planning, acting, and observing are implemented within an AWS workflow.
• Identify the AWS services that enable autonomous reasoning and action, including Amazon Bedrock Agents, AWS Step Functions, AWS Lambda, and Amazon DynamoDB.
• Build and analyze a simple agentic architecture that connects these services into a workflow.

Intended audience

• Cloud Architects
• AI Developers
• Solution Architects
• Technical Consultants new to agentic system design

Prerequisites

Familiarity with the following will be beneficial, but is not required:

• Amazon Bedrock
• AWS Step Functions
• AWS Lambda
• Amazon DynamoDB

The following content can be used to fulfill the prerequisites:

• Amazon Bedrock Orchestration using Knowledge Bases and Agents
• Streamlining Workflows and Tasks With Amazon Bedrock Agents