Robert Ambrogi Blog
UniCourt, a company that provides access to litigation data and analytics, today launched into beta a first-of-its-kind product that combines LLMs and APIs to allow its customers – which include law firms, insurers, legal tech companies, and others – to extract any data points they need from UniCourt’s collection of over 1 billion dockets and documents and deliver the data to wherever they want, such as to a data lake or to software such as Foundation.
Called UniCourt DEEP – for Docket Extraction and Enrichment Platform – the product enables users to uniquely leverage what the company says is the largest normalized and structured database of litigation data in the United States, covering over 40 states and more than 3,000 state and federal courts.
By combining this comprehensive docket database with advanced generative AI features, UniCourt DEEP allows users to create customized docket and document views, allowing firms to extract specific data points and insights tailored to their unique needs and use cases.
Docket data extracted through DEEP can be directly integrated into popular applications such as Litera Foundation or pushed to data warehouses such as Snowflake, Salesforce DataCloud, Azure Synapse, and Microsoft Fabric.
This allows firms to incorporate litigation data directly into their existing workflows and applications, UniCourt says, pushing the specific docket data they need to the location where they need it.
UniCourt says that one of DEEP’s defining features is unparalleled real-time access to court data, with 100% uptime. “This reliability is crucial for use cases where up-to-date information is essential and sets a new standard in the legal tech industry,” the company says.
DEEP provides this customization by enabling the use of LLM prompt engineering techniques on-platform so users can find and extract exactly what they are looking for from dockets or documents.
“Our aim is to allow the subject matter experts to leverage their knowledge in our AI tools to easily find and structure exactly what they are looking for,” said UniCourt founder and CEO Josh Blandi. “It is now super simple for firms, insurance carriers, and legal tech companies to build applications and their own custom AI models on this data.”
A Need for Custom Data
Ahead of today’s launch, I was briefed on the product and given a demonstration by Blandi, Akshay Kumar, vice president of product management, and Rob Lynch, chief operating officer and former head of product.
In a nutshell, UniCourt DEEP takes unstructured data from courts and other sources and converts it into a highly structured litigation knowledge graph built around dockets, documents, attorneys, law firms, parties, judges, and courts.
The product evolved, Lynch said, from the realization that some law firms wanted more specific data points than the standard data being provided through UniCourt’s APIs, which includes categories such as case, law firm, attorney, case type, judge, party and court.
“So what DEEP does is it puts the power in their hands without needing a huge engineering team to get all of the power out of our data set, without needing to put a team of engineers on it, or for us having to build a product,” said Lynch. “It just puts the power in their hands to say, find what you want, extract it, and push it where you need it.”
“For example,” Blandi added, “you could identify, ‘I want this attorney across all the jurisdictions he’s ever filed cases in. I want to load a set of documents with a certain judge.’ Right. And identify certain documents and identify things inside those documents.”
Before the development of DEEP, customers would often come to UniCourt with highly specific data needs that UniCourt would then build for them. For example, UniCourt’s data would already have tagged all personal injury cases, but a user might want to know if, across those cases, injuries from burns in hotel kitchens are trending upwards.
“Now when folks ask that question,” Lynch explained, “we can say, ‘Look, you have a platform here that you can use and you can extract that information right down into the document level, and you can answer those questions using it, structure it, and then push it where you want it.’”
What they realized, Blandi said, is that having the customer’s subject matter expert on the platform, using UniCourt’s AI tools, and able to develop and iterate and validate prompts, is much more efficient and precise than having it done by UniCourt’s own engineers, who are not SMEs.
A Video Example
This video shows how DEEP allows users to customize their views of court data. In this example, the user is looking for all cases involving chest injuries where more than $500,000 was awarded.
They start using the structured fields of Case Type to select Personal Injury cases and Court to search for New York cases and then further refine the search to find just complaints and judgments.
That produces a list of cases, from which the user can then use AI to surface details from within those documents and dockets.
The user prompts, “Find cases with chest injuries. Show judgment awarded.” DEEP then shows the injuries and award amounts. Those results can then be piped into other applications, either on a one-time or ongoing basis.
Coming Out of Alpha
The product so far has been in a private alpha development phase that enables the data to be customized and viewed within Snowflake. The product is now moving into a public beta stage, and new pipelines will allow the data to connect into additional platforms such as Foundation and Salesforce.
At this point in its development, the product is limited to verdict data from personal injury cases, which UniCourt extracts and layers on top of its structured knowledge graph.
During this next phase, UniCourt hopes to get more large law firms to sign on to test the product. The company hopes to launch the production version in the fall.
Kumar said the power of the product is really a combination of two things. One is the data UniCourt already has. “What we have, literally in the world right now, I would very frankly say [is] the best structured knowledge graph around a litigation data set.”
But the second piece is the AI element, “which can dig deeper into terms and pieces of knowledge than a straight-up tagging could ever do.”
For those attending ILTACON this week, UniCourt will be demonstrating DEEP there in their booth in the exhibition hall.