Organizations today possess vast internal knowledge accumulated over decades, but this information is often scattered across multiple systems, making it time-consuming for employees to find what they need. While conversational AI tools like ChatGPT offer powerful search capabilities, they pose serious security risks for enterprises handling confidential data.

This case study demonstrates how SHIFT ASIA developed a secure, enterprise-grade AI chat platform for a major systems integrator, combining conversational AI convenience with robust security and RAG (Retrieval-Augmented Generation) technology. The solution achieved a 60% reduction in information search time while maintaining complete data sovereignty.

Client Overview

A systems integrator (SIer) company affiliated with a major IT enterprise group, with decades of experience ranging from infrastructure development to digital business solutions.

Client Challenges

Over several decades of providing system integration services, the client had accumulated vast internal knowledge assets, including technical documentation, project histories, and organizational regulations. However, this information was scattered across various locations within the organization, requiring employees to spend considerable time searching for it.

While demand grew from staff to create an environment where desired information could be instantly retrieved using conversational AI similar to ChatGPT, Gemini, or Claude, the client faced several significant challenges:

Information Leakage Risk

When using public AI services, there was concern that confidential information entered could be used as training data and potentially leak outside the organization.

Balancing Security and Convenience

Handling sensitive data from universities, government agencies, and healthcare institutions requires extremely high security standards.

Limited Budget for Implementation

As an internal tool, budget constraints necessitated finding a development partner who could deliver high quality at a reasonable cost.

Client Requirements

To promote internal DX and improve operational efficiency, the client required an internal AI platform that met the following specifications:

Robust Security to Protect Confidential Information

A secure environment that completely eliminates information leakage risk, including ensuring that input data is not used for AI model retraining.

Highly Convenient Generative AI Environment

A system supporting a wide range of operations through natural conversational interfaces like ChatGPT.

RAG (Retrieval-Augmented Generation) Implementation Leveraging Internal Knowledge

A mechanism enabling responses that reference organization-specific data, such as internal regulations and project documentation.

Flexible Management Functions and Cost Control

Capabilities for usage restrictions and budget settings at the project and user levels, with easy administrator visibility into usage patterns.

SHIFT ASIA’s Solution

To address these challenges, we proposed and developed a closed AI chat platform exclusively for the client using Azure OpenAI Service (AOAI) and other technologies.

The key elements were establishing a secure environment and implementing RAG (Retrieval-Augmented Generation) technology. This enabled the AI to reference proprietary internal data and provide accurate, contextually appropriate responses like an experienced internal assistant.

The main features of the AI platform developed in this project include:

1. Strong Security for Worry-Free Use Without Information Leakage Concerns

We established a secure environment where input data is not used for AI model retraining. Enterprise-grade security enables safe usage even for operations containing confidential or personal information, without concern for data leakage risks.

2. Instant Answers from Internal Knowledge via RAG

Internal regulations, organizational charts, project documentation, and other materials are integrated with the platform. RAG technology enables the AI to search and reference internal data to generate responses simply by uploading data.

Questions like “What are the specifications for the XX feature in the OO project?”, “What are the contract terms for the client?” or “Tell me about the latest security regulations.” Receive immediate AI responses with source references. This has dramatically reduced inquiry response costs and idle time waiting for answers.

3. Closed Environment Limited to Invited Users Only

Beyond company-wide shared information, “closed environments” accessible only to specific members can be created. Only invited users can interact with the AI based on specific project documentation, making it useful for highly confidential project management.

4. All-Purpose AI Assistant from Administrative to Development Tasks

The AI platform handles everything from general office tasks, such as email composition and meeting minutes, to engineering work, including code generation and debugging. It functions as a “reliable partner” for all employees regardless of role.

5. Flexible Budget Allocation and Usage Visualization

Token usage and costs can be monitored in real-time per member or project. Budget caps can be set per group, enabling planned AI operations while preventing unexpected cost overruns.

Achieved Results

The AI platform deployment achieved significant results, reducing internal information search time by approximately 60%. Particularly, the dramatic reduction in search time for internal rules and past project documentation enabled employees to redirect time previously spent on information gathering toward core business tasks.

Additionally, as confidence spread throughout the organization that “using AI won’t leak information externally,” AI adoption became established company-wide. AI began to be used routinely across diverse operations, including meeting minute creation and code reviews, significantly improving operational productivity.

The project also delivered substantial cost benefits. By leveraging SHIFT ASIA’s offshore development resources, development costs were significantly lower than those for domestic development. The combination of reasonable implementation costs and cost savings from improved operational efficiency enabled the recovery of implementation expenses in a short period, achieving a high return on investment (ROI).

Tools, Technologies, and Languages Used

• RAG (Retrieval-Augmented Generation)
• Platform: Microsoft Azure
• AI Model: GPT-4 (Azure OpenAI Service)