The company Avtomakon has developed an AI agent, the Project Analyst Assistant, to automate functional modeling in ERP implementation projects. The tool uses a combination of algorithmic analysis and local LLM processing to build a project framework within two hours of receiving a database copy.
The system extracts metadata from 1C ERP databases to identify objects and functions, which account for approximately 95% of the required data. The AI agent, powered by Ollama and Qwen 2.5, then maps these functions to existing business processes stored in the company's internal ERP-tools database.
This approach allows analysts to skip manual data entry and focus on specific enterprise requirements and implementation nuances. The company reports that this methodology reduces the functional modeling phase duration by at least one month, enabling faster project timelines.
The developer plans to enhance the agent to suggest functional variations and previously implemented customizations based on historical project data. The current system operates entirely within the internal network to ensure data security and avoid token costs.
