Generative AI vs predictive AI
Generative AI creates new content (text, images, code), while predictive AI analyzes historical data to anticipate future outcomes. In law, predictive AI is used by Predictice to estimate litigation outcomes. Generative AI is used to draft documents and summarize court decisions. The two approaches are complementary.
The distinction between generative AI and predictive AI is fundamental to understanding today's legaltech landscape. Generative AI, embodied by LLMs such as GPT-4, Claude or Mistral, creates new content: it drafts pleadings, summarizes court decisions, and generates draft contracts. Predictive AI, by contrast, analyzes patterns in historical data to anticipate future outcomes.
In the French legal sector, the two approaches coexist and complement each other. Predictice (acquired by Doctrine in July 2025) is the flagship example of predictive AI applied to law: by analyzing millions of court decisions, the platform estimates the odds of a successful appeal, likely compensation ranges, and case law trends. On the other side, generative tools assist lawyers with drafting, research and summarization.
The future lies in the convergence of these two approaches. The most advanced legaltech solutions already combine predictive analytics and text generation: estimating the likely outcome of a dispute (predictive), then generating a reasoned defense strategy accordingly (generative). This complementarity gives legal professionals a complete picture, from strategic analysis to document production.
Related terms
An LLM (Large Language Model) is an artificial intelligence model trained on massive volumes of text to learn the statistical stru…
Predictive justice refers to the use of AI algorithms to analyze past court decisions in order to anticipate the likely outcome of…
AI-powered legal automation refers to using artificial intelligence technologies to automate repetitive, low-value tasks: reading …
AI-assisted legal research uses semantic search and NLP to transform how legal databases are queried. Unlike keyword search, seman…
A foundation model is a large AI model trained on massive unlabeled data through self-supervised learning, adaptable to a wide ran…