Scott Stevenson, Spellbook’s co-founder and CEO, said he went into legal software because he was stunned by how expensive the legal fees were while launching a previous startup.
But it wasn’t until late 2021, when Microsoft’s GitHub released Copilot, a software development aid that uses OpenAI large language models to autocomplete code, that Stevenson started thinking about how to apply that technology to legal software.
LLMs seem particularly well-suited to take on legal work. New models from companies like OpenAI and Anthropic are able to consume lengthy documents and quickly offer summaries and insights.
That capability, combined with fine-tuning that uses legal-specific data, makes the technology a powerful tool for attorneys.
Businesses are only recently figuring out how to turn these new advances into products for lawyers and the general public.
Companies like DocuSign, for instance, have troves of data that they can use in conjunction with services like ChatGPT to create new consumer products that were not possible a year ago.
DocuSign earlier this month announced “Agreement Summarization,” a tool that gives users the ability to upload complicated contracts and have it explained in plain language. Essentially, it translates “legalese” into English.
Another example is the human resources startup Deel. According to a person familiar with the matter, the company is working on a generative AI chatbot-like assistant, trained on Deel’s global data, that can answer compliance questions the same way a legal expert might.