Revolutionizing Document Management: How Associa Utilizes Generative AI for Efficient Document Classification
Revolutionizing Document Management: How Associa is Utilizing Generative AI
A guest post co-written by David Meredith and Josh Zacharias from Associa.
In today’s rapidly evolving digital landscape, efficient document management is vital for operational success. Associa, North America’s largest community management company, is pioneering a transformative approach to document classification that not only enhances productivity but also reduces costs. With approximately 7.5 million homeowners and a staggering 48 million documents across their expansive network, Associa faced significant challenges with their existing document management system. This post delves into how Associa is leveraging the innovative capabilities of AWS’s Generative AI to streamline document processing.
Understanding the Challenge
Associa oversees an extensive array of documents, managing around 26 TB of data across more than 300 branch offices. Their employees faced the daunting task of manually categorizing and organizing incoming documents—a process fraught with inefficiencies and errors that created significant bottlenecks. Each day, countless hours were spent on manual tasks, detracting from the company’s overall productivity.
Recognizing the need for improvement, Associa partnered with the AWS Generative AI Innovation Center to develop a cutting-edge document classification system. This generative AI-powered solution aligns with Associa’s vision of achieving operational efficiencies in document management by automating the classification process.
The Solution: Generative AI-Powered Document Classification
Built using the Generative AI Intelligent Document Processing (GenAI IDP) Accelerator, this cloud-based solution utilizes advanced OCR technology and generative AI to convert unstructured documents into structured, organized data. The system is designed to integrate seamlessly into existing workflows, radically transforming how employees interact with document management systems.
Key features of the solution include:
- High Accuracy: The system automatically categorizes incoming documents with remarkable precision.
- Cost Efficiency: It significantly reduces operational costs related to manual document classification.
- Scalability: The accelerator adapts to high volumes of document processing seamlessly.
Workflow Optimization
To ensure the efficacy of their new system, Associa undertook comprehensive evaluations, focusing on three critical areas of document classification:
- Prompt Input: Using full PDF documents versus first page only.
- Prompt Design: Employing multimodal prompting with OCR data versus using the document image alone.
- Model Choice: Selecting among different models, including Amazon Nova Lite and Anthropic’s Claude Sonnet 4.
This meticulous evaluation framework led to striking findings that shaped their final approach.
Evaluation Outcomes
Prompt Input
Initial evaluations showed that using full PDF documents yielded an overall classification accuracy of 91%. However, upon implementing a first-page-only approach, accuracy surged to 95%, while reducing classification costs significantly from 1.10 cents to 0.55 cents per document. The first-page approach not only proved to be more cost-effective but also improved the accuracy of classifying unknown document types, enhancing operational efficiency.
Prompt Design
Experimentations in prompt design indicated that relying solely on the document image—without OCR data—resulted in an overall classification accuracy of 93% with a minimal cost. However, this approach inadequately classified unknown documents, reaffirming the importance of OCR data for maintaining high accuracy across all document types.
Model Choice
The final selection of the Amazon Nova Pro model resulted in a balanced trade-off between performance and cost efficiency, achieving 95% accuracy at an average classification cost of 0.55 cents per document. This configuration proved to be ideal for the variety of document types Associa handles.
Conclusion: A New Era in Document Management
The collaboration between Associa and the AWS Generative AI Innovation Center has led to a groundbreaking generative AI-powered document classification system that is set to revolutionize how they handle operational tasks. As Andrew Brock, Associa’s President of Digital & Technology Services, states, "The solution developed by AWS improves how our employees manage and organize documents, ensuring a significant reduction in manual effort."
This case study showcases how generative AI can address complex challenges in document management, ultimately leading to notable improvements in operational efficiency and substantial cost savings.
For organizations looking to enhance their document management systems, the GenAI IDP Accelerator offers a promising solution. Those interested can explore the GenAI IDP Accelerator GitHub repository or reach out for collaboration opportunities.
Acknowledgments
We extend our gratitude to Mike Henry, Bob Strahan, Marcelo Silva, and Mofijul Islam for their invaluable contributions and guidance in this endeavor.
About the Authors
David Meredith is the Director of Employee Software Development at Associa, with nearly 20 years of experience in the residential property management industry.
Josh Zacharias is a Software Developer at Associa, leading architectural efforts for internal software solutions.
Monica Raj, Tryambak Gangopadhyay, Nkechinyere Agu, Naman Sharma, Yingwei Yu, and Dwaragha Sivalingam represent a team of esteemed experts at the AWS Generative AI Innovation Center, driving innovative AI solutions across various industries.
Explore how generative AI can reshape your organization’s document management processes today!