AI-Driven Operating System for Intelligent Desktop Automation

Modern desktop operating systems require significant command-line knowledge and manual configuration, creating accessibility barriers for many users. This paper presents an AI-driven operating system assistant prototype that integrates a locally-deployed Large Language Model into a minimal Linux environment to enable natural language control of system operations. The proposed architecture employs a modular design with a user interface, a central Orchestrator for secure command execution, and a finetuned LLM that translates user intent into structured system actions. All processing occurs offline on consumer hardware, preserving privacy while maintaining responsiveness. The prototype was evaluated through desktop automation scenarios including file management, project organization, and version control workflows, demonstrating the feasibility of using compact, domain-adapted language models for intuitive desktop interaction. The results establish a foundation for future research into intelligent, offlinecapable computing assistants.