AWS re:Invent 2025 - From Documents to Decisions: Unleashing AI-Powered Business Automation (SMB204)

From Documents to Decisions: Unleashing AI-Powered Business Automation

Overview

Arit, a global management and technology consulting company, has developed a novel solution called Doxy AI to address the challenge of extracting and leveraging data locked within business documents. This presentation showcases Arit's journey in solving this problem, the technical architecture of their solution, and the significant business impact they have achieved for their clients.

The Challenge: Data Locked in Documents

• Arit's clients, primarily healthcare payers and financial services organizations, struggle to efficiently extract data from various types of business documents, such as provider contracts and vendor agreements.
• The current approaches, relying on manual effort and rule-based extraction, are not scalable and suffer from inconsistencies and inaccuracies.
• This data lock-up prevents organizations from effectively utilizing the information for downstream analytics, insights, and business process automation.

Arit's Journey: From Manual to Generative AI

1. Manual Approach: Prior to 2020, Arit's consultants manually processed documents, achieving only 100 documents per person per week.
2. Rule-Based Extraction: In 2020, Arit implemented a rule-based approach using AWS Textract, which improved efficiency but still struggled with varied document formats and templates.
3. Generative AI Solution: In 2024, Arit leveraged the advancements in generative AI to develop their Doxy AI solution, a metadata extraction and interpretation tool built on large language models.

Doxy AI: The Generative AI-Powered Solution

• Doxy AI is a comprehensive platform that integrates various AWS services, including:
  • AWS Textract for document processing and text extraction
  • AWS Bedrock for leveraging large language models (e.g., Anthropic Sonnet 3.5, 3.7, 4.0)
  • AWS Lambda, ECS, and Snowflake for the overall solution architecture
• The platform uses prompting and domain-specific ontologies to extract structured metadata from business documents, overcoming the limitations of rule-based approaches.
• Doxy AI achieves 99% accuracy in metadata extraction, a significant improvement over the previous 50-60% accuracy.

Business Impact and Results

• Doxy AI has enabled Arit to process 500,000 documents per week, a 5x increase from the manual approach.
• The solution has resulted in a 97% reduction in manual effort, allowing Arit's consultants to focus on more strategic and value-added activities.
• Arit's clients have realized $330 million in direct and indirect savings, including:
  • Identifying overpayments in healthcare claims through contract data reconciliation
  • Avoiding the need for additional human resources to manually process documents
• Arit is now launching a SaaS platform based on Doxy AI, making the solution accessible to a wider range of clients.

Conclusion

Arit's journey from manual document processing to an AI-powered solution showcases the transformative potential of generative AI in addressing real-world business challenges. By leveraging AWS services and large language models, Arit has developed a scalable and accurate platform that unlocks the value of data trapped in business documents, enabling its clients to make more informed decisions and drive operational efficiencies.