Build a legal document conflict detector (isaacus.com)

A new project has launched a demo for a legal document conflict detector that identifies discrepancies between a reference contract and an input contract, utilizing advanced AI models. Traditionally, this process required tedious manual comparisons of legal provisions, but the emergence of generative AI technologies has provided a faster albeit imperfect solution. The challenge with existing generative models is their limited recall and susceptibility to hallucinations, leading to incomplete and sometimes inaccurate analyses during document reviews. This project aims to address those shortcomings by implementing a two-step process: using a classifier to identify deviations and gaps between document provisions, and reserving generative AI for final summaries and recommendations. The significance of this innovation lies in its potential to significantly streamline the legal review process by improving recall and precision in conflict detection. The system categorizes findings into 'deviations'—provisions covering the same topic but with differing legal effects—and 'gaps'—provisions present in one document but absent in the other. Essential steps include normalizing and enriching documents into clauses, embedding them for semantic analysis, and utilizing a universal classifier for scoring conflicts. By enhancing automation while maintaining clarity, the demo provides a useful tool for legal professionals, ultimately fostering more efficient contract management and review processes.