What challenges did Octus face before AI implementation?

Investment professionals needed to process vast volumes of research, financial statements, and proprietary intelligence quickly in order to make timely credit decisions. The previous architecture relied on multiple cloud providers, expensive GPU instances, and operationally heavy custom embedding services, creating cost, latency, maintenance, and scaling issues. Octus needed to maintain zero downtime while scaling document ingestion and query throughput for production use.

What was the impact of the AI solution?

The migration cut monthly infrastructure costs by 78.47% and per-question costs by 87.6%. Octus reported zero downtime during migration and 99% uptime overall. The system increased successfully answered questions by 27% and boosted user engagement by 250%. Octus nearly doubled the documents available for Q&A in days instead of weeks and reduced DevOps/SRE maintenance burden by 20%.

Source

Evidence: Low35/100

Transforming financial analysis with CreditAI on Amazon Bedrock: Octus’s journey with AWS

Use case typeInvestment researchUpdated Mar 10, 2025

Octus (formerly Reorg) migrated its flagship CreditAI conversational chatbot to AWS using Amazon Bedrock, Amazon OpenSearch Service, Amazon Textract, Amazon S3, AWS Fargate, Amazon ECS, Amazon MSK, and Amazon Bedrock Guardrails. The application supports natural-language queries over proprietary credit intelligence and streamlines research workflows for investment professionals.

Organization: Octus
Industry: Finance
Location: United States
Published: March 2025

Reported outcomes

+250%

user engagementCustomer experience

−78.5%monthly infrastructure costs−87.6%per-question costs+27%question success rate+99%uptime+100%documents available for Q&A

Strategic outcomes

Cost efficiencyLower maintenance burdenSpeed & agilityFaster addition of document sourcesRisk & complianceUnified on one cloud provider

Catalog median for customer experience deployments: +25% across 53 reported metrics. Compare benchmarks →

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Monthly infrastructure costs: 78.5% decrease

AWS Machine Learning BlogMar 10, 2025Blog postExplicit claimLow evidence strength

our monthly infrastructure costs are now 78.47% lower

Normalized claim

Per-question costs: 87.6% decrease

AWS Machine Learning BlogMar 10, 2025Blog postExplicit claimLow evidence strength

our per-question costs have reduced by 87.6%

Normalized claim

Question success rate: 27% increase

AWS Machine Learning BlogMar 10, 2025Blog postExplicit claimLow evidence strength

There has been ... a 27% increase in questions answered successfully

Normalized claim

User engagement: 250% increase

AWS Machine Learning BlogMar 10, 2025Blog postExplicit claimLow evidence strength

a 250% boost in user engagement

Normalized claim

Uptime: 99% increase

AWS Machine Learning BlogMar 10, 2025Blog postExplicit claimLow evidence strength

99% uptime overall for the whole application

Normalized claim

Documents available for Q&A: 100% increase

AWS Machine Learning BlogMar 10, 2025Blog postInferred claimLow evidence strength

We have almost doubled the number of documents available for Q&A ... in days instead of weeks

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Octus
Provider: AWS
Maturity: Unknown
Linked source: AWS Machine Learning Blog

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 3

1Investment research
2RAG infrastructure
3Workflow automation

Investment professionals needed to process vast volumes of research, financial statements, and proprietary intelligence quickly in order to make timely credit decisions.
The previous architecture relied on multiple cloud providers, expensive GPU instances, and operationally heavy custom embedding services, creating cost, latency, maintenance, and scaling issues.
Octus needed to maintain zero downtime while scaling document ingestion and query throughput for production use.

Octus migrated CreditAI to Amazon Bedrock Knowledge Bases and AWS Fargate, using Amazon Textract for document extraction, Amazon OpenSearch Service for vector search, and Amazon Bedrock Guardrails for response moderation.
The new design uses semantic chunking, multi-tenant embeddings, isolated knowledge bases for client data, and centralized infrastructure frameworks with Terraform, CI/CD, monitoring, and SOC2 controls.
The chatbot provides conversational question answering with source attribution across more than 10,000 companies and hundreds of thousands of proprietary intel articles.

Technologies

Amazon Bedrock Amazon Bedrock Knowledge Bases Amazon OpenSearch Service Amazon Textract Amazon S3 AWS Fargate Amazon ECS Amazon MSK Amazon Bedrock Guardrails

The migration cut monthly infrastructure costs by 78.47% and per-question costs by 87.6%.
Octus reported zero downtime during migration and 99% uptime overall.
The system increased successfully answered questions by 27% and boosted user engagement by 250%.
Octus nearly doubled the documents available for Q&A in days instead of weeks and reduced DevOps/SRE maintenance burden by 20%.

Architecture

CreditAI uses Amazon S3 event notifications to trigger Lambda-based ingestion, Amazon Textract for document extraction, Amazon Bedrock Knowledge Bases for managed embeddings and retrieval, Amazon OpenSearch Service for vector search, AWS Fargate for the web app and Q&A handler, Amazon MSK for streaming between services, and Amazon Bedrock Guardrails for content filtering and hallucination controls. Octus also uses isolated knowledge bases per client, semantic chunking, Terraform-based IaC, CI/CD, and SOC2-oriented monitoring and compliance controls.

Sources & evidence1

Evidence: Low35/100Evidence strength