Enterprise Linux environments face persistent challenges in configuration management at scale, including configuration drift, compliance violations, and the high manual overhead required to maintain consistent system states across large server fleets. Traditional Infrastructure-as-Code (IaC) tools such as Ansible, Puppet, and Chef provide automation frameworks but demand significant human expertise to author, validate, and evolve configuration playbooks, creating a critical bottleneck in operational efficiency. This paper presents the AI-Assisted Configuration Management Framework (AICMF), an end-to-end system that integrates large language models (LLMs) with existing IaC pipelines to automate playbook generation, semantic policy validation, and continuous configuration enforcement across enterprise Linux environments. The framework employs a fine-tuned transformer-based model augmented with retrieval-augmented generation (RAG) to interpret natural language configuration intents and produce syntactically correct, policy-compliant IaC artefacts. A continuous drift detection module performs real-time state reconciliation against defined baselines, triggering automated self-healing pipelines with tiered human-in-the-loop approval gates for risk-proportionate oversight. Experimental evaluation across a 1,000-node heterogeneous Linux testbed comprising RHEL 9, Ubuntu 22.04 LTS, and CentOS Stream 9 over a 12-week period demonstrates a playbook accuracy rate of 94.3%, a 78.6% reduction in mean-time-to-remediate compared to manual baselines, a drift detection latency of 47 seconds with a 3.8% false positive rate, and a CIS Level 2 benchmark compliance rate of 91.3%. These results establish that AI-assisted IaC substantially reduces operational overhead while improving system reliability, security posture, and auditability in enterprise Linux deployments.