The Power of Generative AI in Financial Analysis

The financial services sector is undergoing a profound transformation, thanks to the rapid advances in generative AI. For investment professionals, particularly those working in credit markets, the ability to quickly access reliable and actionable intelligence is critical. Traditionally, analysing vast amounts of financial statements, industry reports, and market news was a labour-intensive process. Now, generative AI offers a new frontier for optimising financial workflows and reducing missed opportunities.

One company at the forefront of this revolution is Octus, formerly known as Reorg. Founded in 2013, Octus provides essential credit intelligence and data services to leading buy-side firms, investment banks, law firms, and advisory firms worldwide. Their new conversational AI product, CreditAI by Octus™, is redefining how financial institutions make informed decisions and manage risk. Learn more about them at Octus.com.

[Image Placeholder: Illustration showing AI revolutionising finance]

CreditAI: A Generative AI Solution Built on AWS

CreditAI by Octus™ utilises cutting-edge generative AI to provide real-time access to insights on over 10,000 companies, sourced from hundreds of thousands of proprietary articles. With features like targeted queries, sector-wide searches, chat history management, and follow-up suggestions, it significantly accelerates the credit analysis process.

Originally built using Microsoft Azure, Pinecone, and Amazon EC2, the system faced several operational challenges as it scaled. Issues such as escalating costs, fragmented DevOps across multiple cloud providers, and limitations in scalability prompted Octus to seek a better solution. This led to a major migration initiative towards Amazon Bedrock, AWS’s premier platform for generative AI applications.

Migrating CreditAI to Amazon Bedrock: A Strategic Move

Recognising the need for scalability, cost-efficiency, and technological agility, Octus selected Amazon Bedrock to support the next generation of AI for finance solutions.

Key migration goals included:

  • Scalability: Handling millions of documents with efficient embedding and retrieval.
  • Cost Optimisation: Moving away from GPU-based services towards multi-tenant serverless architectures.
  • High Performance: Achieving fast, accurate responses without latency penalties.
  • Zero Downtime: Seamlessly migrating production services without affecting users.
  • Operational Simplicity: Consolidating services under the AWS umbrella.

Through Amazon Bedrock Knowledge Bases, Octus streamlined their Retrieval-Augmented Generation (RAG) operations, employed OpenSearch Service for efficient vector search, and leveraged the Claude Sonnet LLM from Anthropic to deliver superior query responses.

Solution Architecture: How AWS Services Empowered CreditAI

CreditAI’s solution architecture is a fine example of how AWS services can be integrated to optimise generative AI applications:

  • Document Ingestion Workflow:
    • New documents uploaded to Amazon S3 trigger Lambda functions.
    • Amazon Textract extracts text and structure.
    • Data flows into Amazon Bedrock Knowledge Bases and OpenSearch for semantic search.
  • Q&A Workflow:
    • Users submit queries via a web app hosted on AWS Fargate.
    • Queries are processed through Kafka streams.
    • Claude Sonnet via Amazon Bedrock generates intelligent responses.

By utilising a serverless and event-driven model, Octus dramatically improved both the ingestion and querying pipelines of their application.

Using Generative AI for Financial Intelligence at Scale

One of the remarkable aspects of CreditAI’s new system is its scalability. With Amazon OpenSearch Serverless, Octus introduced advanced hybrid search capabilities, enabling them to seamlessly combine traditional keyword search with vector search for better query matching.

Additionally, client data isolation through multiple vector databases enhanced privacy and compliance standards — a crucial factor in the financial services industry.

To support multi-client environments, Octus developed:

  • An AuthZ access management service for fine-grained access control.
  • A Global Identifiers service to unify datasets across various domains.

This approach not only secured data but also streamlined access management, positioning Octus as a pioneer in AI for finance solutions.

[Image Placeholder: Visualisation of isolated knowledge bases and access management]

Achieving Cost Efficiency through AWS Innovations

Cost optimisation was a major goal for Octus. Moving from costly GPU instances to multi-tenant embedding services through Amazon Bedrock resulted in a 78% reduction in monthly infrastructure costs. Additionally, embedding times improved dramatically, with 20–30 millisecond turnaround times for document processing.

By choosing semantic chunking over fixed chunking strategies, Octus improved CreditAI’s response accuracy, grounding, and relevance. This approach allowed natural language queries to better align with user intentions, enhancing overall user satisfaction.

Enhancing Security and Response Quality with Bedrock Guardrails

Another crucial improvement came with the adoption of Amazon Bedrock Guardrails. These features enhanced content security, reduced moderation complexity, and improved system performance.

Guardrails allowed Octus to:

  • Simplify moderation through a unified framework.
  • Optimise performance by integrating checks into primary prompts.
  • Increase content control, reducing hallucinations and irrelevant results.

These enhancements helped maintain CreditAI’s high standard of delivering accurate and trustworthy financial insights.

Operational Excellence: Zero Downtime Migration

Using a combination of:

  • Terraform modules for IaC,
  • CI/CD pipelines for automated deployments,
  • SOC2-compliant security frameworks,

Octus achieved a zero-downtime migration to AWS. Their infrastructure costs dropped by 70%, with operational overheads cut by 20%. This streamlined environment not only benefited CreditAI’s reliability but also freed Octus’s engineers to focus more on innovation.

The Road Ahead: Expanding AI for Finance

Looking to the future, Octus aims to continue enhancing CreditAI with several initiatives:

  • Integrating reranking techniques for improved search result relevance.
  • Adopting the upcoming Amazon Bedrock RAG Evaluator for deeper performance insights.
  • Expanding to support large-scale structured data queries using text-to-SQL technologies.
  • Leveraging advanced parsing solutions to further cut ingestion costs.
  • Strengthening disaster recovery protocols for greater system resilience.

These strategic moves will help Octus maintain its leadership in generative AI applications for the financial sector.

Conclusion: AWS and Generative AI Driving the Future of Financial Analysis

The collaboration between Octus and AWS exemplifies how embracing generative AI can revolutionise financial services. By migrating CreditAI to Amazon Bedrock, Octus achieved:

  • 75%+ reduction in operating costs,
  • 250% increase in user engagement,
  • Zero downtime,
  • Improved scalability, reliability, and innovation speed.

In an industry where rapid and accurate decision-making is vital, solutions like CreditAI, powered by generative AI and AWS, represent the future of AI for finance.

Learn more about Amazon Bedrock and explore how you can transform your financial workflows today.