IntermediateAWS Generative AI Developer project · about 12 hours

Bedrock agent that calls business APIs

Create a Bedrock agent that uses action groups to look up and update orders.

The real-world brief

An online shop wants an assistant that can check order status and start returns.

What you will build

  • A DynamoDB orders table
  • Lambda action groups with an OpenAPI schema
  • An agent that chooses the right action
  • Confirmation before any change

Skills you will practise

AgentsTool useAPI designSafety

Tools

Amazon Bedrock AgentsAWS LambdaAmazon DynamoDBOpenAPI schema

Project tasks

The detailed steps, checklists and trainer review open in the SOTT student portal for enrolled students.

1Data and actions Steps in portal
2Agent Steps in portal
3Confirm changes Steps in portal
4Trace Steps in portal

How companies use AWS Generative AI Developer

Insurance: Faster answers from policy documents. Agents searched long PDFs to answer customer questions. A retrieval-augmented assistant answers with citations to the exact clause. Handling time fell and answers became consistent.

Retail: Product descriptions at scale. Writing descriptions for thousands of new products took weeks. A generative AI pipeline drafts descriptions from product data for a human to approve. New products go live in days, with editors reviewing every text.

IT services: Helpdesk triage. Engineers spent hours sorting and routing tickets. An AI model classifies, summarises and routes tickets, with humans handling exceptions. Tickets reach the right team faster.

Questions

Can I add the Bedrock agent that calls business APIs project to my resume?

Yes. Once your trainer approves it, list it under Projects with the tools you used, one line on the problem it solves and a link to your repository and demo video.

How long does this project take?

About 12 hours of hands-on work, and it fits best after week 8 of the AWS Generative AI Developer (AIP-C01) course.

Where do I get the step-by-step tasks?

Enrolled students get every task with steps, "done when" checks, the submission form and a trainer review in the SOTT student portal.