Let There Be Light – the Primary Function of AI in Legal Research

Imagine a library that tried to save money by relying on candles instead of electricity. Any dollars saved come at the expense of knowledge lost. Without adequate light, the contents of a library are as inaccessible as if the doors were closed.

In the world of legal information, light comes to the “library” through indices, key number systems, topic digests, abridgments, and more. Digitization, electronic access, multi-field search, hyperlinks and boolean logic add more light, but are still merely candles.

Candles are discrete tools designed to provide light only within a limited range of where the candle is placed. Compare that to light derived from electricity.

Electricity is a network. Electricity is a power source. Electricity can be context-aware. Electricity can be responsive. Electricity can communicate. And a networked, powered, context-aware, responsive and communicative system can shine light in precisely the right places, at precisely the right time and with precisely the right brightness.

Artificial intelligence as applied in legal research is as transformative to knowledge access as bringing electricity to libraries.

To understand what’s possible, we don’t need to look far for examples.

Open up Google in private or incognito mode on your preferred browser and enter the title of the latest superhero movie (I typed “joker”) and among the top results will be show time listings from nearby theatres. Use a more generic word like “statute” and you are equally likely to see results connected to your location. And when the Google results provide a portion of site (e.g., show time listings), instead of merely a link to a site, you are getting an answer premised on a contextual understanding presumed to have a strong relationship to your query.

In these simple examples, Google, a powerful source of networked information, is looking for context to determine which among the tens or hundreds of millions of results it should present on the first page in order to be responsive to your needs. Context includes your location, whether you’ve searched from desktop or mobile, which sorts of results have generally been in demand, and the presumed preferred interpretation where the search term is capable of multiple meanings.

Outside of private or incognito browsing, Google will also take into account things like your recent search behavior. The more contextual information available, the more your engagement with information is understood, the more opportunity AI has to discern context in identifying and shining light on a set of results that is most relevant to you in that moment.

Just as Google strives to return the results and answers you want, AI-powered and context-aware legal research is also a reality, and it has been advancing at a rapid pace for many years now. Natural Language Processing (NLP) is used to discern the meaning and significance of text to support search queries, and machine learning models are built on and extrapolate from those candle-powered discovery techniques referenced at the top of this article to make sense of that text in the context of your query or research purpose. Three examples:

  • ROSS Intelligence – (US only) – “Highlight specific language in a case and ROSS will use artificial intelligence to find other cases discussing the same point of law…. ROSS returns case cards showing the information that is most relevant to your query.”
  • CARA from Casetext – (US only) – “Just drag and drop a complaint or brief from your matter, enter a few search terms, and let CARA A.I. find you cases and other authorities that share the context of your matter, including facts, legal issues, and jurisdiction”
  • Vincent from vLex* – (Canada, US, Spain, Mexico, 5 other countries and very soon in the UK) – “Just upload a judgment, brief, legal memorandum or any other legal document and in a matter of seconds, Vincent finds and analyzes cited case law, extracts significant legal concepts, and uses dozens of additional criteria to generate the most relevant resultsfrom among vLex’s extensive libraries.”

Each of Westlaw, Lexis and Bloomberg have also announced this year an intent to roll-out (US only at this point) similar context-aware capabilities in their legal research tools, thereby validating the importance of this approach to AI-powered legal research. The common selling point offered by all providers is that AI-driven and context-aware searching reduces search time, surfaces more relevant results, and avoids errors and omissions due to poorly crafted searches. Put differently, the function of AI in legal research is to help you do the right thing faster, better and with greater confidence.

Returning to the candle analogy, many of us were trained on candles and the experts among us can string together an impressive array of candles. But without electricity (i.e., without AI), we are limiting our visibility and knowledge access to the reach of the candle. Why would we be satisfied with that when we can plug in and let there be light!


*Disclosure: I’m with vLex and serve as interim GM for North America and as CEO of vLex Canada


  1. “But without electricity (i.e., without AI), we are limiting our visibility and knowledge access to the reach of the candle.” But aren’t you also limiting your knowledge access to the reach of the “system”? Are these systems designed only to point the user to what’s available on that particular system? There are six different systems listed in the article each presumably with a different reach or scope. Each system may only present a partial view of the picture and not the entire picture, the search may have to continue onto another system. One system may better cater to a certain practice area than another. With the availability of varying systems with varying content is AI being used as effectively and efficiently as it could be or merely adding cost to legal services when the high price of such services is an already contentious issue?

  2. Hi Verna.

    Yes, the fact that each platform I mentioned is presently limited by the content scope of the platform is a barrier to knowledge access. There are very straightforward ways of resolving that.

    “APIs” – or “Application Programming Interfaces” are a pervasive form or technical bridge across knowledge sources than can allow one “app” to search and deliver information present and viewable on another. To solve the problem you identify, it takes nothing more than one publisher to allow another access to their search API.

    Imagine if CanLII had access to the LexisNexis Canada content set through a search API. A CanLII visitor could expand the scope of their search results to include the full Lexis collection. How that user then gains access to the discovered document (content comes to user, or user sent to Lexis site) would be a matter of commercial arrangement between them.

    One physical world version of this is where a patron in Library A searches a catalogue to learn that the resource sought resides in Library B. In this scenario, the patron either trudges over to Library B, or maybe Libraries A and B have an inter-library loan arrangement so the resource can be physically delivered to the patron. Another physical world version is any circumstance in which a resource mentioned

    There is nothing new in collection scope limitations.

    There is nothing new in the idea that adding resources has the potential to increase costs.

    What is new, however, is how much easier and less expensive it is to overcome those limitations through networked (including by API) and AI-supported systems.

    In terms of cost implications, the US experience has shown that the arrival of AI-driven and context-aware search methods is driving the cost of access to content and advanced tools down.

    The average price across monthly and annual plans from Ross, Casetext and vLex in the U.S. is around $75 per month for complete Federal and State primary law collections.

    This competition has driven the legacy players to not only reduce their prices and increase their discounts, but it caused them to actually begin publicly listing prices:

    If you are experiencing limitations due to scope or price increases from a vendor, it’s not because of AI.

  3. Hi Colin, “Artificial intelligence as applied in legal research is as transformative to knowledge access as bringing electricity to libraries.”
    There is another product in AI applied legal research.
    Alexsei (Canada only – Ontario, Alberta, and BC) creates legal research memos using AI based on lawyer inputs.

    Once electricity happens, the network quickly, quickly grows.
    Cheers, Shaunna
    (Disclosure: I work with Alexsei to build client relationships)

  4. Great article Colin. What Google does a great job of is figuring out a user’s search intent based on the device they are using. A mobile user will be served up results and content blocks that make sense to the search intent inherent in the mobile user experience – quick, easy answers to buy something, go somewhere, get a quick factual answer/read news. The desktop user has a different search intent – do more in depth research, buy something for delivery, see more visually detailed analysis, etc… The effect of AI on legal research is it really the value it brings to figuring out a user’s search intent based on the device they are using and serving up answers and content blocks appropriately. Something that goes way beyond just having responsive design on a website and that can only get better and better with pulling together multiple content sites. The ultimate answer for the legal researcher/user depends on the context and the search intent.