arrow_back Back to Use Cases

Contracts AI for Financial Trading

Noah Waisberg • May 30, 2022 • 2 minute read

In the United States and elsewhere, public companies have filing requirements for certain contracts. Because of these requirements, millions of agreements are publicly available. Contracts are among many businesses’ most important documents. They contain a lot of data. Some of that data may enable traders to better evaluate a company’s value.

One problem with contracts as a data source is that there are a lot of them and they can be dense. A few services use experienced lawyers to do thorough analysis of loan covenants (e.g., [Covenant Review](Covenant Review), Xtract Research), and many use their data to make better decisions. As Xtract says in an article about syndicated loans:

Institutional investors, banks, CLO’s, hedge funds and top law firms use our reports to spot off-market terms, understand loopholes and to quickly digest the material provisions.

These services can be super useful, but have three important limitations for traders:

  • They only cover the documents they cover, and this tends to be limited to thousands of agreements.
  • The data is human-created, meaning that it does not come out immediately.
  • Many firms subscribe to these databases, meaning that they all have access to the same information.

Contracts AI can be used on whatever documents users put into it (including ones that may not be publicly available). It is AI, so it works rapidly. And—while then underlying tech itself is available to anyone who uses it—results depend on what documents you run it on and what you ask the AI to find. Our sense is that traders have used Contracts AI in limited situations (e.g., we knew of a law firm customer of our predecessor business (Kira Systems) who used Kira’s Contracts AI to help a client better evaluate contracts for trading purposes), but not heavily yet, offering room for traders to see something that others don’t. Embeddable Contracts AI has relevant limitations too:

  • It spits out data in API form, as opposed to in an easily digestible database. It may not give data in the exact form you would like.
  • Contracts AI makes mistakes, though—depending on what you’re looking for and which AI you’re using—can be pretty accurate (i.e., >90% recall with good precision).

For traders who are reasonably technically competent and willing to put in the work to get unique information, Embeddable Contracts AI may offer an opportunity to get an edge.