Legal AI Demos - An Introduction

Legal AI Actions

AI is too broad a term to get any real idea of what it can do. We don’t go around talking about electricity in an all-encompassing sense. Instead, we talk about how electricity can be used through various different electrical appliances. In the same way, rather than speaking about AI in a general vague sense, we can think and speak about the particular actions we want to try and perform using AI. We can call these AI Actions.

Surprisingly, there are only a handful of different AI Actions that capture how AI can be used in the legal space. We can come up with a list of Legal AI Actions:

1. 🏷 Label - AI can be used to label text in sentences, clauses and documents
2. ⛏ Extract - AI can be used to extract information from text
3. 🗂 Organise - AI can be used to organise collections of text
4. 🔗 Compare - AI can be used to compare passages of text
5. ✍️ Draft - AI can be used to draft a passage of text
6. 🔎 Find - AI can be used to find information from a corpus of text
7. 📝 Summarise - AI can be used to summarise a passage of text

Legal AI Demos

We’ve put together a collection of demos to show these different Legal AI Actions in action 😉. Here’s our Senior Legal AI Chef 🧑‍🍳 Jordan giving a whizz through the demos.

Try out the demos for yourself and let us know what you think!

You can find the demos here.

When Worlds Collide

A common difficulty with doing Legal AI is that there are two separate worlds. There is the world of the domain experts; the legal professionals. Then, there is the world of the technologists; the AI experts. They both live independent lives, have their own fraternities and speak entirely different languages.

When these worlds collide, we struggle to communicate as each tries to adopt the nomenclature of the other. Legal professionals end up speaking about Large Language Models as though they are sacred. AI experts end up speaking about contract review as though that is all legal professional ever do.

What happens if we step back and peel away all of the excess. All of the sales and marketing talk. All of the use case definitions. All of the talk about technical solutions. What are we left with?

Bridging The Gap

The common thread that remains across the domain and technology are the underlying actions legal professionals have to do on a day to day basis. When we get the bus, we don’t really care about the specification of the engine. We’re just worried whether it will complete its action and get us to where we need to be on time.

The Legal AI Actions create a bridge across the domain and the technology. It’s an intuitive way of capturing the aspects of legal work that won’t change over time. You don’t need to be an expert in technology or law to understand these basic actions. Can you think of scenarios where legal professionals have to label text? Or extract information? Or organise documents?

A Shared Foundation

AI can be defined as a technique that uses machine to replicate the problem-solving and decision-making capabilities of the human mind. The Legal AI Actions capture the fundamental capabilities that we seek to replicate with our models. The Legal AI Actions are units of intelligence that can be composed together to create more advanced workflows.

The Legal AI Actions provide a shared foundation and a common basis that can be built upon. There will always be advancements in technology. Today we have Large Language Models, tomorrow we will have something new. What’s constant is the underlying action that legal professionals need to do. As technology develops, there will be better ways of completing those very same actions.

Our Approach

An AI model is a process not an event. An AI model is an end-product. It's a result of a process brining together different components in a particular way and combining them. This process isn't magic. In fact, it's actually repeatable!

Making an AI model is a lot like cooking and we can make a pretty cool analogy!

A 👩‍🍳 chef (data scientist) combines the 🥒 ingredients (data) according to a recipe 📜 (algorithm) and 🍳 cooks the meal (trains the model) in the 🔪 kitchen (computing environment). Once the 🥧 meal (AI model) is ready, it can be 🛎 served (deployed) to a 💁🏻customer (user) ready to be 🍽 eaten (used) and enjoyed 😋 (meeting the user's needs).

To create the demos, we’ve sourced publicly available 🥒 ingredients (data) in the CUAD dataset and worked with open-source tools and models like spaCy, HuggingFace and scikit-learn to craft the recipes 📜 and prepare these meals 🥧. To 🛎️ serve the meals we’ve put together streamlit applications and hosted them on HuggingFace Space’s so everyone can 🍽️ try for themselves 😋 enjoy! The models have not been optimised for performance. The purpose of these demos is to help you understand what can be done with Legal AI.

White-glove Legal AI Services

simplexico offers white-glove Legal AI Service around education, design and development. We’re on a mission to help legal professionals step into a future of collaboration with AI.

We have the 🧑‍🍳 Legal AI Chefs and the 📜 recipes. You have the 🥒 ingredients. Rather than sending your precious 🥒 ingredients out into the 🌲 wilderness, why not bring in a 🧑‍🍳 Legal AI Chef who can tailor a 🥧 dish for your tastes? Best of all, no 🥒 ingredients have to ever leave the 🔪 kitchen!