What challenges did Built Technologies face before AI implementation?

Real estate finance workflows are document-heavy and highly manual, with more than 250 document types, long and inconsistent layouts, scanned pages, nested tables, embedded images, handwritten annotations, and domain-specific terminology. Built needed over 95% confidence in classification and extraction workflows and a reusable capability that could scale across multiple agentic products and millions of documents.

What was the impact of the AI solution?

Workflows that previously took 3-9 days can now be completed in minutes per package. The architecture is designed to scale to 20 million documents per month, 300,000 documents per week, and more than 50,000 batch processing runs. The same document intelligence capability is reused across multiple agentic AI products across the real estate finance lifecycle.

Rank#303of 3,449

Source

ProductionEvidence: Medium50/100

Built Technologies builds AI-powered document intelligence on AWS for real estate finance agents

Use case typeDocument processingUpdated Jul 15, 2026

Built Technologies, a real estate finance software provider that processes over $500B in real estate projects, deployed an AI-powered document processing engine on Amazon Bedrock and the AWS Intelligent Document Processing Accelerator. The solution serves as a reusable foundation for agentic products across the real estate finance lifecycle, supporting draw review, loan agreements, insurance validation, underwriting support, asset management, compliance, and portfolio workflows.

Organization: Built Technologies
Industry: Finance
Location: United States
Published: July 2026

Reported outcomes

3-9%

workflow turnaround timeTime & speed

Strategic outcomes

Speed & agilityReduced document-processing cycle time to minutesScale & capacityDesigned for large-scale batch processing and high document volumeEcosystem & partnershipsCreated a reusable document intelligence foundation across the product ecosystem

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Workflow turnaround time: 3-9% decrease

AWS Machine Learning BlogJul 15, 2026Blog postInferred claimMedium evidence strength

Workflows that previously took 3–9 days can now be completed in minutes per package.

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Built Technologies
Provider: AWS
Maturity: Production
Linked source: AWS Machine Learning Blog

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 2 of 2

1Document processing
2Workflow orchestration

Real estate finance workflows are document-heavy and highly manual, with more than 250 document types, long and inconsistent layouts, scanned pages, nested tables, embedded images, handwritten annotations, and domain-specific terminology.
Built needed over 95% confidence in classification and extraction workflows and a reusable capability that could scale across multiple agentic products and millions of documents.

Built partnered with the AWS Generative AI Innovation Center and AND Digital to create a scalable document intelligence engine using the AWS Intelligent Document Processing Accelerator and Amazon Bedrock.
The multi-stage pipeline uses AWS Step Functions and AWS Lambda for orchestration, Amazon Textract for OCR and structural extraction, and Amazon S3, EventBridge, DynamoDB, and Amazon SQS for ingestion and concurrency management.
The solution performs OCR, document classification and splitting, dynamic schema generation, extraction, assessment with confidence scoring and evidence geometry, human-in-the-loop review, and rule-validation reasoning with citations.
AND Digital built a custom React-based user interface authenticated with Amazon Cognito for uploading documents, managing processors and schemas, reviewing results, and routing low-confidence items to subject matter experts.

Technologies

Amazon Bedrock AWS GenAI Intelligent Document Processing (IDP) Accelerator Amazon Textract AWS Step Functions AWS Lambda Amazon S3 Amazon EventBridge Amazon DynamoDB Amazon SQS Amazon Cognito AWS AppSync Amazon Nova Lite Anthropic Claude 3

Workflows that previously took 3-9 days can now be completed in minutes per package.
The architecture is designed to scale to 20 million documents per month, 300,000 documents per week, and more than 50,000 batch processing runs.
The same document intelligence capability is reused across multiple agentic AI products across the real estate finance lifecycle.

Architecture

A scalable multi-stage document intelligence architecture on AWS: documents enter Amazon S3 and emit EventBridge events, a Lambda-based queue sender writes to DynamoDB and SQS, a queue processor controls concurrency and starts AWS Step Functions, and the workflow runs OCR, Bedrock-based classification and splitting, extraction, assessment, and optional rule validation. Extraction is parallelized in a Step Functions Map state. The solution also includes a custom React UI authenticated with Amazon Cognito, plus AppSync subscriptions for real-time status updates.

Sources & evidence1

Evidence: Medium50/100Evidence strength