Rally offers software solutions to law firms, HR departments, and other organizations. An important part of its service is facilitating the process of legal document distribution, signature execution, and tracking. Its latest feature for law firms will facilitate the development of contracts before the signature process.
Last week, the company announced a new product called Spellbook that promises to help lawyers draft contracts three times faster. Beyond the standard GPT-3 training, Spellbook has also been “trained on thousands of contracts.”
Filling in the Gaps…Faster
Spellbook is, first and foremost, about filling in the gaps in agreements. A blog post announcing the solution summarizes how to think about the product.
While this impressive new technology is still in its infancy, there are some very obvious applications in the legal field and it’s only a matter of time before its use becomes standard operating procedure. For example, generative AI can be used to assist in drafting:
Contracts
Trial briefs
Legal research
Pleadings
Discovery requests
Deposition questions
Marketing materials for your firm or business
The tech industry is moving to support these use cases and it’s moving fast. Tools like Alexsei are already helping lawyers use AI to efficiently turn research questions into comprehensive memos and we’ve made incredible progress with Spellbook to aid you in drafting contracts.
All of this means that, as a lawyer, you will soon need to be skilled in using generative AI to remain competitive. Fortunately, it will also allow you to be more productive and efficient in your work. For example, if you can use generative AI to reduce the time you spend drafting contracts, you will be able to spend more time on other tasks, such as meeting with clients or preparing for trials. It will help avoid costly mistakes and can act as a useful second set of eyes when drafting or reviewing documents. In other scenarios, it can act as a muse, providing inspiration for things you hadn’t thought of.
From Creator to Editor
The stakes are higher for contracts than they are for generating SEO-optimized blog posts. That is why lawyers aren’t likely to be removed from the contract drafting process anytime soon. The idea behind solutions like Spellbook is to save time in generating the initial legal text and shift the primary activity to making intelligent requests as prompts and becoming an expert editor. It’s like having a cost-effective research and writing assistant that shows up through software automation instead of via an hour-long commute.
Many knowledge-worker jobs have not changed much over the last generation. Word processing, document management, and search technologies have all helped professionals become more productive. However, tasks like contract drafting are essentially creative processes conducted by trained humans. Before deep learning-based AI hit the scene, most of these tasks were impervious to automation.
Generative AI solutions like GPT-3 have changed the equation by offering a strong start along with in-context augmentation of existing work. Expect to see knowledge worker skill requirements change as these tools become more common.
Novel Text-to-X Applications
Spellbook is a novel application of large language model capabilities. Because it is technical in nature, it immediately seemed very similar to the code generation solutions from GPT-3 and GitHub CoPilot, even though it is generating prose. I later learned that Rally and Spellbook co-founder Matt Mayers says it was CoPilot that inspired the company to create the solution.
Spellbook also has a similar value proposition to CoPilot: make highly skilled and highly paid professionals more productive.
The Rally solution then takes this a step further by suggesting new ideas that the lawyer did not explicitly express and offering advice on terms that may merit negotiation or gaps in the legal protections enumerated in the document. You could imagine another beneficial outcome of standardizing certain types of legal language used across contracts.
Large language models such as GPT-3 were a pure novelty when first introduced to the public in 2020. People had fun generating mostly coherent essays, poetry, and HTML code. With specialized training, new applications have emerged to help professionals with their jobs. Marketing content development and software code generation have proven to be popular use cases. Now we will see if contract writing and other efficiency-oriented solutions take hold as well.