How Associa transforms document classification with the GenAI IDP Accelerator and Amazon Bedrock

2026-02-05 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Cloud Computing & IT Infrastructure · Depth: Intermediate, long

Summary

Associa, North America's largest community management company, collaborated with the AWS Generative AI Innovation Center to develop a generative AI-powered document classification system. This system addresses the challenge of manually categorizing approximately 48 million documents across 26 TB of data, a process that was time-consuming and error-prone for its 15,000 employees. The solution, built using the Generative AI Intelligent Document Processing (GenAI IDP) Accelerator and Amazon Bedrock, automatically classifies incoming documents with high accuracy, aiming for operational efficiencies and cost savings. The development involved optimizing prompt input (full PDF vs. first page only), prompt design (multimodal with OCR vs. image only), and model choice (Amazon Nova Lite, Pro, Premier, and Anthropic's Claude Sonnet 4). The final system, utilizing Amazon Nova Pro, achieved 95% accuracy at an average cost of 0.55 cents per document by processing only the first page with combined OCR and image data.

Key takeaway

For AI Engineers and ML teams developing intelligent document processing solutions, prioritizing prompt engineering and model selection is crucial. Your evaluation should focus on optimizing input length and multimodal data integration, as demonstrated by the 95% accuracy and 0.55 cents/document cost achieved with first-page-only OCR+image input using Amazon Nova Pro. This approach can significantly reduce operational costs and improve classification accuracy, especially for ambiguous document types, directly impacting project ROI and system reliability.

Key insights

Optimizing prompt input and design significantly enhances generative AI document classification accuracy and cost-efficiency.

Principles

• First-page-only processing improves accuracy and reduces cost.
• OCR data combined with images enhances classification accuracy.
• Model choice balances accuracy, especially for "Unknown" types, and cost.

Method

The GenAI IDP Accelerator uses OCR (Amazon Textract) and generative AI (Amazon Bedrock) to convert unstructured documents into structured data, supporting scalable, modular document processing workflows.

In practice

• Evaluate prompt input (full vs. first page) for efficiency.
• Test multimodal prompts with OCR and image data.
• Benchmark multiple LLMs for accuracy and cost trade-offs.

Topics

• Generative AI
• Document Classification
• Amazon Bedrock
• Intelligent Document Processing
• Amazon Textract

Code references

• aws-solutions-library-samples/accelerated-intelligent-document-processing-on-aws

Best for: Machine Learning Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• Rocket Close transforms mortgage document processing with Amazon Bedrock and Amazon Textract — Generative AI, combined with OCR, can automate complex document processing with high accuracy and speed.
• How Ricoh built a scalable intelligent document processing solution on AWS — Combining generative AI, serverless architecture, and standardized frameworks can overcome document processing scaling limits.
• I Spent May Evaluating Different Engines for OCR — OCR is a routing problem; no single engine excels across all document types.
• Technical advances in document understanding — Document processing has evolved from basic OCR to sophisticated multimodal models that preserve structure and fine detail.
• Text Detection and OCR with Amazon Rekognition API — Amazon Rekognition API offers highly accurate cloud-based OCR, providing line or word-level text detection.
• Why Nepali OCR Is Brutally Hard — And How We Finally Solved It — A hybrid deep learning pipeline overcomes complex script challenges to automate document data extraction and form pre-filling.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.