Like Netflix for Legal AI Models: Reveal’s AI Model Marketplace Goes Deep
What is the genius of Netflix? Is it that they have great TV shows and movies? No, paid providers like HBO has had that for years. Is that their movies and television shows are on-demand? No, cable companies have had that for years as well. The genius behind Netflix is that for one subscription rate, not only do you have access to previously created movies and television shows, but Netflix is algorithmically producing new content based on what viewers actually want to watch.
Netflix content fits the “long tail” theory which says that there are audiences for just about everything, however, the distribution of the content via traditional methods capped the audience. Netflix creates create multiple series and movies to appeal to different audiences via their algorithmic analysis of the distribution of their audience base. They also release their new content periodically and allow for the discovery of new content to delight subscribers. This is the template that we at Reveal wanted to “borrow” when we created the industry’s first AI Model Library.
Out-of-the-Box, Pre-Trained Legal AI Models
Reveal’s AI Model Marketplace is a compilation of AI model solutions built by our team of data scientists. So when we say it’s like the ‘Netflix’ of AI models, we’re referencing the simplicity, diversity, and functionality that provides Reveal’s NexLP AI users access to out-of-the-box, pre-trained AI models.
We look to our clients and do not focus on “generic” review problems but try and create AI models that are highly important to the job function of our “long tail” audience. We believe there is a distribution problem with solutions, much like Netflix discovered. Reveal is solving this problem by using our deep learning NexLP AI technology and releasing machine learning AI models to serve the different audiences, practice groups, and teams that use our software.
The list of available AI models is available on our website and from the AI platform itself. We look at data to develop new AI Models that solve real challenges for our end-users. Not every Netflix user has the same taste in shows and movies, and we see that not every lawyer, investigator or knowledge worker is looking for the same types of stories in their data.
Our data science team has developed machine learning models that have a wide array of story coverage. We have models for investigations, privilege detection, HR, data clean-up and general litigation data challenges. There is an audience for every model and some clients may rely on these AI Models as part of their specific workflow.
Next-Generation Artificial Intelligence
The rush to create the “new Netflix” from every single streaming provider is a lot like the evolution of the AI marketplace.
Everyone in the eDiscovery industry claims to have AI. Of course, not all AI solutions are authentic AI. Authentic artificial intelligence will adapt the solution to the data and understand the context. That is why AI models from the AI Model Marketplace and Reveal’s NexLP AI technology are powerful examples of authentic artificial intelligence.Our next-generation authentic AI technology is built upon the work of Dr. Dan Roth, winner of the John McCarthy Award, given to top AI researchers. That foundation created an industry-first AI Model Marketplace. This new marketplace sets the stage for organizations across legal, compliance, risk, and eDiscovery to tap into the NexLP AI ecosystem to utilize the power of AI without the need for programming.
Everything a user needs is available in a one-shop-stop and can quickly be accessed and integrated without needing extra IT or data resources on the user’s end. This experience is only possible through our AI Model Marketplace.
Like Netflix, our list of AI models continues to grow, and most are included as part of your subscription. Our AI Model Marketplace offers 20+ ready-to-go models that can detect a range of compliance & ethics behaviors from fraud to discrimination in addition to sensitive data like PII and privileged conversations. Also, like Netflix, we worked hard on the user experience. Just like selecting a movie, you can read the title, description, and some facts about the AI model.
Increase Efficiency & Transparency
Now using these models may seem like a different workflow or paradigm shift than traditional review techniques. However, looking at the workflow closely and you will see it is not that different in workflow, but vastly different in technology.
In many traditional review scenarios, a review team may kick-off the project by thinking-through a variety keyword searches. They apply these keyword searches and based on the results may sample or review in a linear fashion all the documents. If the team is versed in technology, they may use TAR techniques as well. However, this workflow results in additional unnecessary documents being reviewed and a lack of transparency.
The use of AI models is very similar to the workflow above. However, it is much more efficient and has full transparency. The AI model has been built by a team of data scientists. Once the AI model is easily applied to your case, every document is analyzed by the AI and provided a probability score between 1% and 100%. In other words, the AI ranks how much signal each document has to what is critically important to you. The case team simply starts with the high scoring documents and works their way down to a cut-off point where the documents are no longer what they are looking to learn. Meanwhile, the AI applications are continuously learning in the background and will provide batches of new documents for the case team to evaluate.
Take for instance the Sexual Harassment AI Models included in the AI Model Library. If the case team’s job were to track down instances of suspect employee behavior, the workflow would be easy to perform.
Simple to Load & Apply
First, select the Sexual Harassment AI Models from the AI Model Library to load into your case. The loading of the models should take less than a minute or so. Next, the AI model will pre-score and provide analytics on the documents and people that have the highest probability of meeting the criteria. Then simply have your team interrogate the highest scoring documents. As the team is evaluating the documents, the AI is refining and learning what your team is looking to accomplish. The AI will then assist your case team in finding all the documents that match the criteria.
We have seen a surge in interest in our AI models to detect privileged documents from a collection of discovery material. For the problem of privilege detection, our data science team broke down the problem of privilege into discrete AI models. Independently, each of these AI models can be utilized on their own. However, when utilized together, these AI models comprise the privilege content detection solution.
This is much like when Netflix has an anthology series. Each episode can stand on it’s own, but when combined they make for a powerful series with a cohesive theme.
A Data Science Team at your Fingertips
In addition, the Reveal AI Platform provides the capabilities of a full-blown data science team with AI developers in your office, allowing your organization to build your own AI models and store them securely in your own private AI model library.Your models are only limited by your imagination and can be used on all matters where you want to remove the time-to-knowledge, create efficiencies, and leverage your organization’s expertise.
The recipe for creating great AI Models when using the Reveal NexLP AI technology is very simple. You need access to data, authentic AI technology, and a team that understands the problem trying to be addressed by an authentic AI solution. Reveal’s NexLP AI technology is built to replicate the natural language processing activities that a data scientist would go through to break down a problem and find the solution. Using this recipe, Reveal clients can build world-class AI models that can be re-used on every matter and across an organization.
As if creating your own library wasn’t enough, we’ve taken the process a step further by providing a method for Reveal clients to suggest specific AI Models from the Reveal team — a process designed to spur ongoing tech innovation across the legal and compliance industries.
Legal and compliance teams can now achieve outcomes of 5x faster review than traditional review strategies and significant cost reductions. We listen to our clients’ use cases and build new AI models based on their feedback. The data science team tries to emulate the long tail of Netflix and build AI models that are important while most software companies don’t approach such complex challenges.
For Reveal clients that provide trusted advice to outside users, we have heard positive feedback from our clients who use the full scope of the AI Model Marketplace as part of their business development presentation.
Typically, clients will include a screenshot (similar to above) of the AI Models developed by the Reveal data science team, as well as bespoke models created by partners. The feedback we hear is that the ultimate client is over-joyed to discuss solutions to business challenges, as opposed to discussing AI for technology’s sake. The ultimate client wants to know how AI can help immediately with them for their organization. The Reveal AI Model Library was developed to solve this immediate, time-to-knowledge issue. As a great side benefit, it helps with the business development aspect as well.
Reveal’s AI Models allow legal, risk and compliance teams access to technology that was not previously available without added integration on their end. The AI Model Marketplace streamlines the delivery and use of models developed by Reveal’s data science team and partners to both on-premise and cloud clients.
For more information, or to request a model, visit our AI Model Marketplace.