Legal AI works best alongside rules-based engines, not as a replacement, Litera says
Litera's testing reveals that AI alone is insufficient for accurate legal document review, advocating for hybrid systems combining AI and rules-based engines.Summary
Litera, a legal software company, conducted benchmarking tests comparing traditional rules-based engines with leading AI models like Gemini, Claude, and ChatGPT on complex legal documents. The results demonstrated that while AI excels at natural language understanding, it falls short in accuracy and reliability when handling non-text elements and long-form documents. Accuracy rates dropped significantly on longer contracts, reaching as low as 40%, which is unacceptable in legal work where precision is critical. Therefore, Litera argues that a hybrid approach—integrating the precision of rules-based engines with AI’s strengths—is optimal. AI can effectively orchestrate workflows and assist with tasks like test case generation, but should not replace established systems where accuracy is paramount. Expert supervision of AI outputs is also crucial to mitigate risk. Ultimately, success in legal technology isn’t measured by AI adoption, but by achieving positive client outcomes through the appropriate application of tools.
(Source：{{author}})