Overcoming Document Processing Challenges with Generative AI: A Case Study from Ricoh
Transforming Enterprise Workflows through Serverless Architecture and Standardized Frameworks
Customer Overview
Challenges with Complex Document Processing Workflows
Solution Overview
Architecture Details
Results and Outcomes
Best Practices and Lessons Learned
Getting Started
Conclusion
Acknowledgments
About the Authors
Overcoming Document Processing Limits: A Scalable AI Solution at Ricoh
Co-written by Jeremy Jacobson and Rado Fulek from Ricoh
Enterprises often grapple with the unyielding complexities of document processing, particularly in industries like healthcare where regulatory compliance and accuracy are paramount. Ricoh has stepped up to this challenge by harnessing generative AI, serverless architecture, and standardized frameworks, creating an innovative solution that revolutionizes document processing.
The Innovative Framework
Utilizing the AWS GenAI Intelligent Document Processing (IDP) Accelerator, Ricoh developed a reusable framework that significantly reduces customer onboarding time from weeks to merely days. The results are impressive: the system’s processing capacity is projected to grow sevenfold, accommodating over 70,000 documents per month, while decreasing engineering hours per deployment by over 90%.
A Closer Look at Ricoh USA
Ricoh USA, Inc., a global technology leader, operates in over 200 countries and serves notable players in the healthcare sector—processing hundreds of thousands of critical documents each month. From insurance claims to clinical records, the challenges of reliance on custom manual engineering became a bottleneck, stunting expansion efforts. Each customer implementation required unique development, making scalability an elusive goal.
Compliance Meets Cutting-Edge Technology
Meeting stringent compliance standards such as HITRUST, HIPAA, and SOC II is no small feat. These regulations often constrict the agility required for rapid AI development and deployment. To combat these hurdles, Ricoh leveraged foundation models available through Amazon Bedrock and combined them with Amazon Textract. This strategic move empowers customers to reap the benefits of AI automation without jeopardizing compliance.
The Standardized Multi-Tenant Solution
One of the primary objectives of Ricoh’s initiative was to establish a standardized, multi-tenant solution for automated document classification and extraction. By addressing the common pain points in complex document workflows, Ricoh transformed its document processing from a tedious custom-engineering endeavor into a scalable, repeatable service.
Key Functional and Non-Functional Requirements
The Intelligent Business Platform’s functional requirements included:
- Capturing data attributes from unstructured documents
- Assigning confidence levels to those attributes to indicate when human review is necessary
Additionally, non-functional requirements focused on performance, accuracy, cost optimization, and operational efficiency. For instance, the system must handle traffic spikes, provide low processing costs, and ensure quick customer onboarding through configuration rather than coding.
Addressing Complex Document Workflows
Ricoh’s prior approach involved traditional optical character recognition (OCR) augmented by multimodal AI models capable of interpreting both text and images. However, the advent of multimodal FMs on Amazon Bedrock revealed limitations, particularly with complex workflows comprised of multi-part documents.
Partnering with AWS for Scalability
By collaborating with the AWS Generative AI Innovation Center, Ricoh implemented the GenAI IDP Accelerator, which enables production-grade document processing solutions. Selecting Processing Pattern 2—a combination of Amazon Textract for OCR and Amazon Bedrock for intelligent classification—has allowed Ricoh to tackle complex, multi-part healthcare documents effectively.
Security and Compliance
To comply with HIPAA, encrypted data at rest and in transit is paramount. The integration of AWS security services ensures data integrity and confidentiality, while operational controls satisfy HITRUST certification requirements.
Results and Outcomes
The outcomes of Ricoh’s implementation have been overwhelmingly positive:
- Onboarding Time: Reduced from 4–6 weeks to 2–3 days
- Monthly Throughput: Increased from ~10,000 documents to >70,000 documents
- Engineering Hours per Deployment: Reduced by over 90%
These advancements allow Ricoh to lower costs for healthcare customers while maintaining high accuracy rates—achieving extraction accuracy levels that typically hit 98–99%.
Best Practices and Lessons Learned
Ricoh’s rich experience has led to the identification of best practices for deploying IDP solutions:
- Select the Appropriate Processing Pattern: Choose a framework that balances complexity and control.
- Combine OCR with FMs: A hybrid approach maximizes scalability and accuracy.
- Integrate Confidence Scoring: This ensures effective human-in-the-loop workflows, critical for healthcare.
Getting Started with IDP Solutions
For organizations looking to streamline their document processing mechanisms, the AWS GenAI IDP Accelerator stands as an optimal choice. Engaging with AWS experts can help tailor your implementation strategy effectively.
Conclusion
Ricoh’s strategic implementation of the AWS GenAI IDP Accelerator illustrates how enterprises can transcend traditional scaling limits by integrating generative AI with serverless architecture and compliance frameworks. The outcome is a faster, more accurate, and operationally efficient document processing solution that meets regulatory standards without compromise.
To learn more about building your own IDP solution, explore the GenAI IDP Accelerator today.
Acknowledgments
Special thanks to key contributors, including Bob Strahan, Guillermo Tantachuco, Saeideh Shahrokh Esfahani, Mofijul Islam, Suresh Konappanava, and Yingwei Yu.
About the Authors
- Jeremy Jacobson is a lead developer and solutions architect for AI at Ricoh USA.
- Rado Fulek specializes in building secure and scalable document processing platforms.
- Earl Bovell, Vincil Bishop, and Jordan Ratner from AWS also contributed significantly to this endeavor.