What was the impact of the AI solution?

Dramatically reduces manual labor required to draft STRs, accelerating compliance reporting timelines. Improves accuracy and completeness of STR documents by supplementing knowledge with real-time web data. Enables financial institutions to comply with regulations more efficiently, reducing risk of noncompliance penalties.

Source

Automating Suspicious Transaction Report Drafting for Financial Compliance with AWS Generative AI

Use case typeRisk assessmentUpdated Jun 13, 2026

AWS leverages generative AI large language models through Amazon Bedrock, integrated with knowledge bases and agent workflows, to automate drafting suspicious transaction reports (STR) for financial compliance. The solution uses Retrieval-Augmented Generation (RAG) with Amazon OpenSearch Service for vector search of relevant context, and AWS Lambda functions for web crawling and data enrichment stored in Amazon S3. Amazon Bedrock Agents orchestrate multi-step user interactions to gather data, query knowledge bases, and generate accurate and compliant draft STR documents. This managed generative AI solution significantly reduces manual effort, speeds compliance reporting, and improves report quality with up-to-date financial fraud information.

Organization: Financial institutions
Industry: Finance
Location: United States
Published: July 2025

Reported outcomes

Strategic outcomes

Risk & complianceAutomated suspicious transaction report draftingCustomer experience & trustImproved report accuracy and completenessRisk & complianceMore efficient regulatory complianceSpeed & agilityAccelerated compliance reporting timelines

Primary read

Use case focus

Showing 3 of 4

1Regulatory compliance automation
2RAG for knowledge-augmented generation
3Conversational AI agents

Financial institutions face time-consuming manual processes to draft suspicious transaction reports (STR) for regulatory compliance, risking delays and inaccuracies.

The solution uses Amazon Bedrock to access foundation models (e.g., Anthropic Claude 3.5 Haiku) and applies RAG via Amazon OpenSearch Service knowledge bases to ground responses in factual data.
Amazon Bedrock Agents enable conversational interactions to collect transaction details, invoke Lambda functions for web scraping, and generate structured STR drafts.
Data from scraped websites about fraudulent entities are indexed in OpenSearch and integrated into the report generation workflow.
The solution provides architecture and code implementation details for deployment using AWS Cloud Development Kit (CDK) or manual setup.

Technologies

Amazon Bedrock Amazon SageMaker Amazon OpenSearch Service AWS Lambda Amazon S3 Amazon Bedrock Agents Amazon Bedrock Knowledge Bases

Dramatically reduces manual labor required to draft STRs, accelerating compliance reporting timelines.
Improves accuracy and completeness of STR documents by supplementing knowledge with real-time web data.
Enables financial institutions to comply with regulations more efficiently, reducing risk of noncompliance penalties.

Architecture

The architecture involves Amazon Bedrock Agents interacting with Amazon Bedrock Knowledge Bases powered by OpenSearch Serverless for vector search indexing. AWS Lambda functions crawl and scrape websites to enrich knowledge base content stored in S3, orchestrated through agent action groups to gather user inputs and generate draft STR reports.

Sources & evidence2

Groundedness: 4/5

Primary source Source 2

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence