Building an intelligent brain platform for small and medium-sized enterprises using ChatGLM and Multi-Agent Systems.

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Enterprise Systems

Background:

• Large language models (LLMs) exhibit strong semantic understanding but face challenges in specialized SME domains, including overgeneralization and poor knowledge alignment.
• Small and medium-sized enterprises (SMEs) require tailored AI solutions to address specific business needs and integrate domain knowledge effectively.

Purpose of the Study:

• To propose and evaluate an
• Enterprise Intelligent Brain
• platform designed to enhance LLM capabilities for SMEs.
• To improve semantic adaptability and intelligent responsiveness of LLMs in real-world SME scenarios through a novel architecture.

Main Methods:

• Developed an
• Enterprise Intelligent Brain
• platform based on Chat Global Language Model (ChatGLM), incorporating a multi-agent coordination mechanism and enterprise knowledge graphs.
• Applied domain-specific fine-tuning to ChatGLM, integrated a multi-agent task allocation framework, and utilized knowledge graph reasoning for enhanced accuracy and contextual understanding.
• Identified core semantic demands of SMEs (policy consultation, customer service, business process execution) to construct a triadic system architecture (semantic parsing, task scheduling, knowledge support).

Main Results:

• The optimized system significantly outperformed baseline models, achieving a task completion rate of up to 99.904% and an average response time of 0.858 seconds.
• Demonstrated high context retention (up to 0.953) and user satisfaction (up to 4.767), alongside strong performance in knowledge invocation coverage and error recovery.
• Validated effectiveness using three public datasets: Baidu DuReader-Enterprise, E-commerce Dialogue Dataset, and Enterprise Knowledge Graph-Based Q&A Dataset.

Conclusions:

• The proposed
• Enterprise Intelligent Brain
• platform offers a practical and scalable framework for deploying LLMs in domain-specific SME contexts.
• The study provides technical solutions and theoretical insights for developing enterprise-grade semantic intelligence platforms supporting intelligent decision-making and service automation.
• The system's robustness in complex SME environments highlights its potential for intelligent business operations.