Automation Document Legal Legal oo leh AI

Template-based automation has existed for decades. Tools such as HotDocs, Clio Draft (formerly Lawyaw), Gavel (formerly Documate) and the open-source Docassemble turn a Word document into a template with variables, conditional sections and repeating blocks. A user answers an interview, and the software assembles the document the same way every time. The output is only as good as the template, but it is predictable. The same answers produce the same text, and a change can be reviewed once and then reused. Generative AI adds a different capability. Instead of choosing among pre-written paragraphs, a model writes new text: a statement of facts, a tailored letter, a first-draft motion or suggested redlines. Tools such as Spellbook and Harvey work this way, as do AI features in mainstream practice software. This helps when no two documents are alike. The downside is that the output varies from run to run and can contain confident errors. In Mata v. Avianca (S.D.N.Y. 2023), lawyers were sanctioned after filing a brief with case citations that ChatGPT had made up. The case is widely cited as a warning about unverified AI drafting. Templates remain safer in four situations: when a court or agency prescribes a form, when negotiated or approved language must not change, when volume is too high to review every sentence, and when you need to prove exactly what logic produced a document. AI suits first drafts of narrative sections, turning facts into prose, and adapting approved language to new facts, with review. The common misconception is that AI makes templates obsolete. Many firms combine the two. AI pulls data from intake documents to fill template variables, or it drafts inside clearly marked sections of an otherwise fixed template, so the controlled parts stay controlled.