Generation AI (operation confirmed version)

With the emergence of ChatGPT, generative AI has rapidly spread and is attracting attention as an innovation that will bring a leap in competitiveness to companies. In addition to conducting research and studies on the technical aspects of generative AI, we have been working quickly to establish its use in the workplace.

Examples of initiatives

In response to the emergence of generative AI, we quickly established a company-wide research team called "LLM Mavericks" to set up a research and investigation system and work to create an environment in which "generative AI" can be utilized company-wide.

We have launched a virtual organization to research LLM (Large Scale Language Models)!

Slack AI Chatbot

We created an environment where anyone can easily use generative AI from Slack, which is a common tool used for work. This quickly increased understanding within the company about what this much-talked-about generative AI is.

Web AI Chatbot

When using generative AI in business, there are concerns that input data may be used for re-learning. Therefore, we used the Azure OpenAI Service, which does not have such concerns, and provided an environment where generative AI can be used in business with an easy-to-use ChatGPT-like Web UI across the company.

Generative AI Usage Guidelines

In addition to providing infrastructure for using generative AI, we also developed guidelines outlining points to be aware of when using generative AI in business, and rolled them out company-wide.

Development and provision of RAG (Retrieval-Augmented Generation) systems

We developed an environment that allows even non-engineers with no prior experience with SQL to use the "company-wide integrated data infrastructure" that was built in-house using DWH and data lakes, by simply "writing in natural language what they want to search for in the data infrastructure."

The "normalized structured data" stored in the DWH is not compatible with generative AI if provided as is, so the database metadata within the DWH (what columns each table has and what data it contains) is vectorized and provided to the RAG system.

Generative AI implementation support service for enterprises

Based on these in-house efforts, we also provide a generative AI implementation support service that uses GPT, Azure, and Slack to accumulate and utilize internal data.

Generative AI Implementation Support Service for Enterprises | Saison Technology