My favorite Legal AI Agents
So many bots to build, so little time.
But what are legal AI agents?
Agents are defined differently depending on who is doing the defining. For legal work, I like this definition:
A bot that produces an outcome for a specific legal or consulting workflow.
Important to distinguish with AI assistants. An agent is not an assistant.
This means that it produces the outcome based on structured inputs with a very high degree of accuracy, rather than only providing help or speeding up while a human has to lead.
Building specialised legal AI agents is exciting work.
I am listing out here legal AI agents that excite me the most (in no particular order). I have tried to be more general in the descriptions so that more of you see where it intersects with your work.
A bot to:
Review a set of documents, and prepare a structured memo, note or opinion.
Compare how two drafts stack up on specified parameters.
Review instructions and/or documents and produce a first draft of an agreement.
Identify whether a document is complying with a checklist of requirements.
Research on materials available online and produce a report, note or memo.
Identify the most serious and least serious issues in a document.
Restrict your opinion or note to basing itself on a specified set of documents or database.
Listen in on meetings and organise outcomes and follow ups.
Help a layman draft a highly accurate legal document in a guard railed manner.
Review, summarise and brief on background and action points before meetings, appearances and other IRL interactions.
Are there any others that come to mind?
Which ones have you used, experienced or liked?
I think it can be used as a check and balance, a self-interrogation, you can give it all your best arguments *against* yourself privately and so "push your envelope" of correctness & accuracy all the more. And if it's critical I might get up to 5 agents to argue amongst themselves, as they all speak slightly different languages.
I've instructed my day to day Google/basic search AI, Meta on WhatsApp, to only speak in English, strictly not 'murican. For an English Law practitioner, AI speaking 'murican is riddled with errors and pretentiousness/verbosity, even fakeness, for its own sake. Because that's what the murican AI data, murican lawyer stuff, does.