Meet Cascade: Our Next-Gen AI for Confident Contract Drafting

When you’re building a new product, you need to start with a real understanding of the problem that you’re solving and for which users. Otherwise, you may end up with a product that doesn’t have a use case for anyone.

This is where our co-founders' experience as former Magic Circle lawyers makes such a difference - it means that we can start with that genuine understanding, and build their experience into the solutions.

This is what allowed us to build something as exciting as Cascade, our newest generative AI product. Cascade helps lawyers instantly detect and understand the knock-on effects of edits made in legal contracts, and solves a problem in a way not seen before. 

Legal contracts are a complex system of definitions, clauses and obligations, interwoven through nested cross references. An edit in one place in the contract can have an unexpected and unintended impact across multiple pages or get buried deep down in the document. Understanding the effects of edits is already complicated enough without adding the pressures lawyers face, racing against time (and sometimes fighting sleep) when navigating all this information.

These are highly valuable contracts. Missing a critical dependency during drafting or review on a contract that could be worth millions… well, that's got to create anxiety, right?

This was reflected when we interviewed lawyers during our customer research. We had 'oh that’s cool' moments, when they realised the peace of mind and confidence it could give them when they introduced changes to documents. We also learned more about how our solution can keep them focused during high-stakes negotiations. In addition, interviewees were excited by the time-saving and productivity gains the product offered throughout the workflow.

Building, Testing, Iterating

An amazing part of the building process is how collaborative it is. With Cascade, we had a cross-functional team consisting of Product Managers, AI and Product Engineers and Designers, working closely together with a defined understanding of the problem space, validated by users. The clarity that understanding gave us meant we were all working from the same page to build the best possible solutions. 

The first challenge was building a system that can understand legal text. Legal contracts have nested structures, so processing and understanding them correctly means breaking the contract down to its composite parts of definitions, clauses, etc. This is where the tried and tested algorithm that powers Definely Draft was so useful - it meant we could build a system that understands legal text, fully within context, with accuracy.

Building the reasoning capability of the system was another challenge. It had to not only identify the provisions affected by the initial change but also be able to explain the output. To ensure lawyers can rely on Cascade with total confidence, we had to evaluate the accuracy of the system, a problem that’s common to any AI and Machine Learning development.

These challenges are both specific and complicated. This is why we've worked extensively with a legal professional throughout development, to feed back on the approaches we’ve taken. 

We worked closely with our customers from the start. Some of them have seen the product since it was a mere Figma prototype, and helped us shape the final product’s design and direction by providing valuable feedback and insights as we developed it. 

Sam Dixon, Chief Innovation Officer at Womble Bond Dickinson, UK was one customer we worked with. He said, "I've worked closely with the Cascade team throughout the product’s development and iteration. Their innovative approach to solving this challenge continues to impress and surprise me."

We’ve been excited for it to go live, because we’re proud of the result. To our knowledge, there is no other solution in the market comparable to Cascade when it comes to the deep understanding of the document that then allows us to flag knock-on effects through a contract and provide the reasoning behind it.

What’s Next?

Interoperability is something that is a real difference-maker for us in terms of building around workflows. One thing I love to tell people is that Cascade does not work alone but as part of the Definely Suite. Let’s say a user finds an Undefined Term from Draft's highlighting and decides to Insert Definition using Vault. This newly inserted definition may have consequential changes across the document bringing the user to Cascade. Finally, the user will use Proof to proofread any cross-reference issues.

Using an intelligent agentic system to connect these tools in an even more powerful way is very much on our radar and a core part of our roadmap moving forward. 

The future of Cascade is becoming a key part of our agentic framework, orchestrated by Enhance to help lawyers do their work faster, better and with confidence. It’s an exciting proposition, especially considering the early positive feedback we’ve seen since the launch.