What challenges did Vitech face before AI implementation?

Fragmented documentation across multiple internal platforms. Low productivity due to inefficient retrieval and information overload. Inconsistent information access without a single source of truth.

What was the impact of the AI solution?

Reduced inquiries to product teams by over 50%. VitechIQ is used by about 50 users daily. The system handles roughly 2,000 queries per month. The knowledge base expanded to more than 35,000 API documents, totaling up to 3 GB.

Source

Evidence: Low35/100

Vitech uses Amazon Bedrock to revolutionize information access with an AI-powered chatbot

Use case typeStaff assistantUpdated Jun 13, 2026

Vitech is a global provider of cloud-centered benefit and investment administration software. The company’s product documentation was scattered across Confluence and SharePoint, creating low productivity and inconsistent access to a unified source of truth. Vitech built VitechIQ, an internal AI-powered chatbot for employees to access documentation more efficiently. The chatbot uses Amazon Bedrock, Amazon Bedrock Knowledge Bases, Amazon Titan Embeddings, Amazon Aurora PostgreSQL with pgvector, Amazon S3, Amazon EC2, Elastic Load Balancing, CloudWatch, and LangChain. The solution provides source attribution and private connectivity through Bedrock VPC endpoints.

Organization: Vitech
Industry: Tech & Comms
Location: United States
Published: May 2024

Reported outcomes

−50%

quantified impactOther quantified impact

Strategic outcomes

Customer experience & trustCreated a single source of truthNew product / capabilityBuilt an internal AI chatbotRisk & complianceProvided source attribution and private connectivityScale & capacityExpanded the knowledge base

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

Normalized claim

Quantified impact: 50% decrease

AWS Machine Learning BlogMay 30, 2024Blog postInferred claimLow evidence strength

Reduced inquiries to product teams by over 50%.

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

Customer: Vitech
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 4

1Internal Knowledge Assistant
2Document Search
3Employee Productivity

Fragmented documentation across multiple internal platforms.
Low productivity due to inefficient retrieval and information overload.
Inconsistent information access without a single source of truth.

Built VitechIQ as an internal AI chatbot for documentation retrieval.
Used Amazon Bedrock and Amazon Bedrock Knowledge Bases to implement RAG from ingestion through response generation.
Stored documents in Amazon S3 and embeddings in Amazon Aurora PostgreSQL with pgvector.
Used Amazon Titan Embeddings for semantic search and Amazon Bedrock for summarization and Q&A.
Hosted the Streamlit application on Amazon EC2 behind Elastic Load Balancing and secured traffic with Bedrock VPC interface endpoints.
Used LangChain for orchestration, chunking, prompt templates, and conversational memory.

Technologies

Amazon Bedrock Amazon Bedrock Knowledge Bases Amazon S3 Amazon Titan embeddings Amazon Aurora PostgreSQL Amazon EC2 Elastic Load Balancing CloudWatch LangChain

Reduced inquiries to product teams by over 50%.
VitechIQ is used by about 50 users daily.
The system handles roughly 2,000 queries per month.
The knowledge base expanded to more than 35,000 API documents, totaling up to 3 GB.

Architecture

VitechIQ uses Amazon Bedrock for LLM inference and Amazon Bedrock Knowledge Bases for RAG. Product documentation is stored in Amazon S3, chunked and embedded with Amazon Titan Embeddings, and searched via Amazon Aurora PostgreSQL compatible edition with pgvector. The Streamlit front end runs on Amazon EC2 behind Elastic Load Balancing, with Amazon Bedrock VPC interface endpoints providing private connectivity and CloudWatch capturing logs and runtime metrics.

Sources & evidence1

Evidence: Low35/100Evidence strength