KEY RESULTS

• Operational Efficiency: Reduced manual effort, cutting costs - improving productivity
• Accuracy: Achieved >95% accuracy - 30 different document types
• Customer Experience: Faster, more accurate information delivery for both internal and external stakeholders
• Scalability: Future-proofed solution supporting growth
• Data-Driven Insights: Enabled smarter decision-making through advanced analytics and enriched data outputs
  

THE CHALLENGE

A global Financial Services organisation faced significant operational inefficiencies due to the manual extraction of business-critical information from a high volume of large, complex documents. Manual data extraction required substantial human effort, leading to increased operational costs and delays in delivering critical information.  Furthermore, variability in human data extraction led to inconsistent outputs, compromising data accuracy, reliability and risks in decision making.  The documents used varied significantly by type, format, and layout, making standard extraction methods ineffective. As document volumes and business requirements grew, manual processes became a scalability bottleneck, limiting future growth.

THE SOLUTION

i-Volution Platforms partnered with the client to deliver a cutting-edge, automated data extraction solution leveraging Generative AI (GenAI) and Cloud technologies. The solution involved:

Input-Data Provisioning:

• Integrated AWS Textract on Cloud to automatically classify the diverse set of document types, to then parse and extract the relevant target information from each document type.

Model Implementation:

• Deployed and configured AI models from the Claude family on Cloud: Claude 3 Sonnet for primary processing, complex analysis and data extraction, Claude 3 Haiku for refined quote extraction for speed and efficiency.

Output-Model Improvement:

• Iterative evaluation and refinement to enhance model performance and accuracy.

Platform Integration:

• Seamlessly integrated the solution with data sources, workflows, and business applications.
• Cross-referenced LLM outputs with data from business applications to enrich results.
  

Related Projects