This foundational course introduces agentic artificial intelligence and the AWS services used to build applications that interpret requests, retrieve information, and perform actions through tools. Participants explore agent design, foundation models, prompting, knowledge integration, security, and operational considerations.

The course focuses on Amazon Bedrock Agents, with introductory coverage of Strands Agents and Amazon Bedrock AgentCore to explain the options for developing and operating agents on AWS. Service coverage aligns with the Amazon Bedrock Agents documentation and Amazon Bedrock AgentCore documentation.

Duration 3 Days – 21 hrs.

 

Objectives

• Explain agentic AI concepts and distinguish agents from chatbots and predefined workflows.
• Identify suitable business use cases for AI agents on AWS.
• Describe the roles of Amazon Bedrock Agents, Strands Agents, and Amazon Bedrock AgentCore.
• Configure a basic agent with clear instructions and an appropriate foundation model.
• Explain how retrieval-augmented generation connects agents to organizational knowledge.
• Define agent actions and connect tools using AWS Lambda and APIs.
• Apply foundational access controls, guardrails, and human approval boundaries.
• Identify common agent failures and basic cost, monitoring, and deployment considerations.

 

Target Audience 

• Application developers and software engineers beginning with agentic AI.
• Cloud engineers and solution architects exploring AI applications on AWS.
• Data and AI professionals transitioning into agent development.
• Technical consultants and technology specialists supporting AI initiatives.
• Technical team leads responsible for planning agentic AI solutions.

 

Prerequisites 

• Basic understanding of cloud computing and AWS services.
• Familiarity with the AWS Management Console.
• Basic knowledge of Python, APIs, and JSON.
• General understanding of generative AI and large language models.
• Familiarity with AWS Identity and Access Management (IAM), Amazon S3, and AWS Lambda is helpful.
• Access to an AWS training environment with the necessary service permissions and supported model access.
• No prior experience developing AI agents is required.

 

Course Outline 

Day 1: Agentic AI and AWS Foundations

Module 1: Introduction to Agentic AI

• Core concepts and characteristics of AI agents.
• Agents, conversational assistants, and predefined workflows.
• Agent components: models, instructions, tools, context, and memory.
• The agent loop: interpreting requests, selecting actions, and using results.
• Common business use cases and limitations.
• Selecting appropriate boundaries for autonomous behavior.

 Module 2: AWS Services for Agentic Applications

• Overview of Amazon Bedrock and foundation models.
• Model selection considerations: capabilities, latency, and cost.
• Introduction to Amazon Bedrock Agents.
• Introduction to Strands Agents as an agent development framework.
• Introduction to Amazon Bedrock AgentCore for deploying and operating agents.
• Supporting roles of IAM, Amazon S3, AWS Lambda, and Amazon CloudWatch.

Module 3: Agent Instructions and Basic Configuration

• Defining agent purpose, scope, and expected behavior.
• Writing clear instructions and providing relevant context.
• Configuring a basic Amazon Bedrock agent.
• Handling incomplete requests and requesting clarification.
• Understanding sessions and conversational context.
• Agent preparation, versions, and aliases.

  

Day 2: Knowledge and Tool Integration

Module 4: Connecting Agents to Organizational Knowledge

• Retrieval-augmented generation fundamentals.
• Introduction to Amazon Bedrock Knowledge Bases.
• Preparing source documents and organizing data in Amazon S3.
• Basic concepts of chunking, embeddings, and vector search.
• Connecting a knowledge base to an agent.
• Source attribution, relevance, and handling missing information.

 Module 5: Agent Actions and Tool Use

• Identifying tasks that require external tools.
• Defining action groups in Amazon Bedrock Agents.
• Describing functions, parameters, and API schemas.
• Using AWS Lambda to implement agent actions.
• Validating tool inputs and handling tool responses.
• Error handling and confirmation before consequential actions.

 Module 6: Designing Reliable Agent Workflows

• Breaking user requests into manageable steps.
• Combining knowledge retrieval with tool execution.
• Managing context, session state, and memory requirements.
• Setting completion conditions and execution limits.
• Handling failed actions, retries, and escalation.
• Introduction to single-agent and multi-agent patterns.

  

Day 3: Security and Operational Foundations

Module 7: Security and Responsible Agent Behavior

• IAM roles and least-privilege permissions.
• Protecting sensitive data and managing credentials.
• Introduction to Amazon Bedrock Guardrails.
• Prompt injection and untrusted content.
• Human approval and authorization for sensitive actions.
• Understanding the limits of model instructions and guardrails.

 Module 8: Introduction to Framework-Based Agents and AgentCore

• Core Strands Agents concepts: models, tools, and agent execution.
• Choosing between managed agent configuration and framework-based development.
• Overview of AgentCore Runtime.
• Introduction to AgentCore Memory, Gateway, and Identity.
• Understanding how agent code, models, and operational services fit together.
• Basic deployment considerations for a simple agent application.

 Module 9: Monitoring, Cost, and Solution Planning

• Understanding agent traces, logs, and tool execution records.
• Basic operational visibility with Amazon CloudWatch.
• Troubleshooting permission, retrieval, and tool invocation issues.
• Cost drivers: model usage, retrieval, compute, and supporting services.
• Managing resource usage and cleaning up training resources.
• Defining a foundational architecture for a business agent use case.