LG Chem: Creating Generative AI-Powered Services to Enhance Productivity
LG Chem: Creating Generative AI-Powered Services to Enhance Productivity
LG Chem noticed that their employees were spending a lot of time searching for safety regulations and guidelines so, with the help of Generative AI and Dataiku, they provided an AI service that helps them find that information quickly and accurately.
Working in the chemical manufacturing space, LG Chem is always at the forefront of technology and innovation. It was no different once Generative AI came on the scene, and the AI team — heavily focused on digital transformation (DX) technology to help the business — wanted to create AI services that use Generative AI to help the business.
We sat down with Sungbum Jun, director of LG Chem’s AI team (whose remit includes planning and developing LG Chem’s standard DX platform and AI services for its employees) to learn more about the Generative AI use case, the impact it has had on the business, and the role of Dataiku along the way.
Enter: AI Service for Searching LG Chem Knowledge
LG Chem noticed that their employees were spending a lot of time searching for safety regulations and guidelines (and it was difficult to find the answer), so they provided an AI service using Retrieval Augmented Generation (RAG) that helps employees find that specific information quickly and correctly.
Using Generative AI, the company developed a document and standard search service on top of the Open Source LangChain Framework (enabling flexibility and scalability), which they modified to fit their needs. The search engine is powered by RAG architecture, which has four main components:
A vector database
A Large Language Model (LLM), specifically the Microsoft Azure OpenAI API
Because LG Chem is a manufacturing company, adhering to specific IT security regulations was a challenge. Convincing the security team that the project doesn’t involve direct data exposure to the LLM but only utilizes it for responses required careful explanation. By clarifying potential misunderstandings during initial stages of implementation, the team was able to ensure smooth progress of the project.
Another important element of this Generative AI use case was the graphical user interface (GUI) the team built in order to enhance the usability of the LLM by the business. A well-designed GUI enables smoother interactions with the LLM and provides users not only with textual information but with visual elements. LG Chem’s GUI prioritizes intuitive usage and rapid familiarity for users new to the QA system by maintaining the user experience of the ChatGPT web browser while incorporating domain-specific visual features.
For example, it provides users with links to source documents alongside textual answers to their questions. Clicking on a source document link displays the referenced page or paragraph of that document in the web browser. This feature empowers users to quickly and accurately assess relevant documents searched to their needs without the added effort to search for relevant content in source documents. LG Chem expects these features to enhance user interest and satisfaction with the service.
Business Impact: Higher-Value Work for Employees, Increased Productivity Levels
This Generative AI use case is invaluable to LG Chem’s employees across the organization. The purpose and intent of the system are to leverage the model to handle repetitive queries that internal personnel have historically addressed. This includes interpreting and conveying information from internal documents (regulations, guidelines, manuals, cases, etc.) that could differ between individuals. The benefits and business impact of this use case include:
The ability to hedge against human-related risk and enhance operational efficiency by using LLMs to re-allocate employees to higher-value tasks, enhancing overall work productivity.
New employees can use the system to swiftly grasp and execute their tasks effectively, saving time.
Users receive responses that are not only much more accurate, but also user-friendly and of higher quality compared to rule-based chatbots. They can also obtain specialized responses tailored to their internal content rather than general responses from ChatGPT.
We have high expectations that utilizing Dataiku’s functionalities will allow us to effectively manage the backend and frontend aspects of our projects, leading to efficiency gains and scalability across multiple initiatives.
Sungbum Jun AI Team Director, LG Chem
Dataiku’s Role in Succeeding With Generative AI
To ensure the swift progress of each Generative AI project, it was crucial for LG Chem to establish a backend pipeline that involves embedding search documents and transforming them into a database. LG Chem leverages Dataiku to create the backend and frontend pipelines for their Generative AI use cases. Dataiku enables them to streamline processes, achieve standardization, and eventually apply similar approaches to numerous other projects.
When embarking on any Generative AI project, the most important element to LG Chem is focus on solving a business problem with DX technology. The second indispensable element is collaboration across teams — business experts, data scientists, data engineers, IT security, and platform engineers, for example. With Dataiku, LG Chem navigated the roadblock of difficulties sharing project statuses and processes across individuals and teams and used Dataiku’s robust collaboration features and capabilities to make sharing project goals and status updates easy.
The AI service for searching LG Chem knowledge has already been applied to Environmental, Social, and Governance (ESG) and Enterprise Resource Planning (ERP) systems. They made sure that when the IT systems (like ERP) or personnel changed, there were fewer interruptions in their work (thereby also reducing operational losses).
What Does the Future Hold for LG Chem and Generative AI?
With any Generative AI project, LG Chem focuses on four key elements:
Business value and strategy: Check that the project fits the business goal and DX team strategy.
Technical feasibility: Check that the project can be done within current technology and resources.
Data usability: The project has access to the system which has the right data.
User acceptance: Users can use the AI service well and accept the new technology.
Keeping this framework in mind, LG Chem is excited to expand to more Generative AI use cases such as a Q&A system based on IT systems and facility manuals (starting with LG Chem’s smart factory systems in Korea and eventually rolling out to America and China) and professional knowledge support to address IT security standards or legal problems. To make these incremental use cases happen fast, the team will start by making an AI development standard with templates, develop and expand the AI systems, and monitor the use cases and continually upgrade the system.
One of the most valuable use cases is expected to be in the realm of environmental safety. Due to ESG policies, LG Chem has introduced new processes and guidelines that need to be applied to various high-risk tasks. Remembering and responding to safety regulations for each of these tasks is challenging and personnel need to consult documents and then respond to queries, leading to delays in responses and work disruptions. Through the AI service being developed via this LLM project, LG Chem anticipates that users will soon be able to search for desired processes and guidelines directly, as well as access relevant documents. This should significantly reduce work delays caused by response delays, improving efficiency and effectiveness in managing these safety regulations.
JK Lakshmi Cement — Driving Innovation and Transformation With Dataiku
Learn how JK Lakshmi's data team uses Dataiku to improve and save time on reports, and make operational tasks more efficient.
In this fireside chat-style webinar, Lance Lambert, Director of Enterprise Business Intelligence at NXP Semiconductors and Kurt Muehmel, Chief Customer Officer at Dataiku, discuss NXP’s keys to success with their data initiatives.