Generative AI platform
Amazon Bedrock Agents
Orchestrates multi-step tasks with foundation models and action groups.
Key points
- Uses a foundation model to reason through user requests and decide which actions to call.
- Action groups connect an agent to APIs, Lambda functions, or other business logic.
- Can use Knowledge Bases so the agent has access to relevant private context.
- Maintains orchestration for multi-step tasks such as gathering information, invoking tools, and returning a final response.
- Works with IAM and service configuration to control what actions the agent can perform.
When to use it
- Choose Agents when a generative AI app must take actions, not just answer questions.
- Use Agents for booking workflows, ticket triage, data lookup, or guided business processes.
- Use Agents with Knowledge Bases when the workflow needs both retrieval and API execution.
Exam tips
- Agents orchestrate tasks; Knowledge Bases retrieve context; Guardrails apply safety controls.
- An action group is the exam clue that the model needs to call external tools or business APIs.
- Least-privilege IAM matters because agents can invoke real actions through configured integrations.
- For a simple grounded Q&A bot with no actions, Knowledge Bases alone may be the better fit.