Exclusive Content:

Haiper steps out of stealth mode, secures $13.8 million seed funding for video-generative AI

Haiper Emerges from Stealth Mode with $13.8 Million Seed...

Running Your ML Notebook on Databricks: A Step-by-Step Guide

A Step-by-Step Guide to Hosting Machine Learning Notebooks in...

“Revealing Weak Infosec Practices that Open the Door for Cyber Criminals in Your Organization” • The Register

Warning: Stolen ChatGPT Credentials a Hot Commodity on the...

How Ricoh Developed a Scalable Intelligent Document Processing Solution Using AWS

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:

  1. Select the Appropriate Processing Pattern: Choose a framework that balances complexity and control.
  2. Combine OCR with FMs: A hybrid approach maximizes scalability and accuracy.
  3. 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.

Latest

Trade Desk Considers Partnership with OpenAI’s ChatGPT as CEO Increases Stake

Key Insights on The Trade Desk: Exploring New Opportunities...

Robotics Companies Receive Funding Surge Following Chinese New Year TV Gala

Humanoid Robots Shine at China's Lunar New Year Gala,...

Editorial Message | Natural Language Processing

The Human Touch in the Age of AI: Exploring...

‘Generative AI is for the Underprivileged’: Viral Video Suggests AI May Increase Creativity Gap

The Cultural Divide: Can Generative AI Democratize Creativity? Insights...

Don't miss

Haiper steps out of stealth mode, secures $13.8 million seed funding for video-generative AI

Haiper Emerges from Stealth Mode with $13.8 Million Seed...

Running Your ML Notebook on Databricks: A Step-by-Step Guide

A Step-by-Step Guide to Hosting Machine Learning Notebooks in...

VOXI UK Launches First AI Chatbot to Support Customers

VOXI Launches AI Chatbot to Revolutionize Customer Services in...

Investing in digital infrastructure key to realizing generative AI’s potential for driving economic growth | articles

Challenges Hindering the Widescale Deployment of Generative AI: Legal,...

Integrate Amazon Quick Suite Chat Agents into Enterprise Applications

Streamlining Conversational AI Integration: Overcoming Challenges with Amazon Quick Suite Embedded Chat Enhancing User Experience with In-App Conversational AI Seamless Deployment of Secure Embedded Chat Solutions Solution...

Creating a Custom Model Provider for Strands Agents Using LLMs on...

Bridging the Gap: Creating Custom Model Parsers for Strands Agents on Amazon SageMaker Navigating Response Format Incompatibilities Understanding Strands Custom Parsers Implementation Overview Step 1: Install ml-container-creator Step 2:...

Techniques and Implementation of Time Series Cross-Validation

Mastering Time Series Cross-Validation: Techniques and Implementation What is Cross Validation? Understanding Time Series Cross-Validation Model Building and Evaluation Importance in Forecasting & Machine Learning Challenges With Cross-Validation in...