Software‑based legal assistance systems have traditionally relied on expert‑system rules, knowledge bases, and other symbolic reasoning techniques. This article reviews how such symbol‑centric document automation services can complement modern generative large language models (LLMs). Classical systems offer structured legal templates, traceable inference chains, and strict compliance checks—features often missing from free‑form text generation. Conversely, generative models excel at interpreting natural‑language requests, filling gaps, and producing diverse drafts, thereby improving user experience. We discuss the potential benefits of integration—accelerated drafting while preserving interpretability, reduced manual review effort—and the challenges, such as ensuring legal accuracy of model outputs, data privacy, and designing interfaces between symbolic and statistical components. To assess feasibility, we conducted preliminary experiments where an LLM identified and automatically corrected issues in texts authored by laypeople. The model captured roughly 78% of semantic errors and suggested plausible rewrites, yet fine‑grained compliance still required human verification. Overall, the synergy between expert systems and generative AI opens a promising avenue for legal document automation.
Review: The paper convincingly demonstrates both theoretical and empirical complementarity between symbolic reasoning and generative AI, laying groundwork for future interpretable and efficient legal drafting platforms.