As part of a small talented team at AI-startup McCarthyFinch, I supported product discovery and concept validation.

Alongside understanding user needs through various interviews and concept prototypes, I learnt about the nuances of both law and data science as disciplines.

Training the Algorithm

The team began working on an algorithm to identify the various elements in a legal document as a first step into automating some of the repetitive and mundane parts of the lawyers job. Documents were to be broken down and tagged by lawyers. A quick functional interface was put in place.

Report builder

The first use-case chosen to test the algorithms was creating due-diligence reports. These are a summary of all the legal obligations an entity may be under. A good example would be when a company is being bought by someone else.

After some interviews with lawyers to understand how and why they are created today, and shadowing them on how some of them were working through the process, I moved to wireframing and paper-prototyping ideas as a way to share and get early feedback. Once we were comfortably clear on what the interface was supposed to do, the final UI was designed and developed.

While developing the report builder, we realised that lawyers were heavily conditioned to use MS Word as their primary tool. Any text editing functionality we would build will be compared to that, and most likely we will not be able to support it.

Our goal as a product team was to build and showcase that AI algorithms can aide add value to the legal workflows, and that needed to come with a strong enough text-editing capability.

Building smaller word plugins that leveraged our algorithm was one of the ways forward to make this a more viable product. Some of them are now live on the word app store.

Building on these already formed habits again, we designed a service to initiate an automated first draft via email.

Just like they would email someone in the firm to kick-start the document and then review it.