ADAP for Public Sector: Agentic AI Document Analysis and Processing

ADAP is a modular, AI-enabled document processing tool that helps federal, state, local, and higher education public-sector teams reduce the burden on employees. It supports the many demands of releasing information to the public by helping users read, understand, extract data from, and find specific information in large volumes of documents. Built on AWS, ADAP stores documents securely in Amazon S3, uses AWS Textract to convert scanned and image-based documents into machine-readable text, and processes documents through Amazon Bedrock (Claude and Titan models) for AI-driven detection, classification, and generation. Batch operations — such as processing large document sets for redaction or report generation — are orchestrated through a serverless AWS Lambda pipeline, while interactive capabilities like document chat call Bedrock directly, drawing on a ChromaDB vector store for semantic search. Access to all of this is governed by least-privilege AWS IAM roles. In one tool, employees can redact information for FOIA requests, summarize documents for understanding, query documents on specific topics, or generate reports based on the user's context and needs. Employees can also sanitize documents by removing sensitive information such as methods and sources, rewrite content in the author's tone, and support declassification while allowing human review at each step. Initial text-based capabilities include:

• Redaction – Automatically detect and redact a broad range of sensitive information, including Personally Identifiable Information (PII), Protected Health Information (PHI), financial data, classification markings, and other regulated or custom-defined sensitive content. Helps speed up FOIA processing while preserving human review.
• Sanitization and Declassification – Remove methods and sources from intelligence reports, rewrite content in author's tone, and support declassification after sanitization with human review.
• Document Chat – Query large documents using natural language in a chat interface.
• Report Generation – Generate a report from documents and data.
• Summarization – Summarize large documents.