Successful AI Requires Time and Learning – Not Magic

As much as people might believe it to be true, Artificial Intelligence (AI) is not a magic wand. But it is a pretty awesome bag of tricks.

In working with customers in law firms and legal departments, iManage has found that many organizations know what kind of knowledge they want to pull out of their documents but generally don’t understand how to use AI as a process to make it happen.

AI can be easily applied to with some their specific knowledge management projects–like due diligence reviews and extracting provisions from leases, for example–with little to no training required.

We’ve found, however, that customers want to go beyond these use cases and utilize this emerging technology in a transformational way beyond a single project. They want to unlock knowledge and insights from their legal documents and contracts at an enterprise level, in a repeatable way, and across many projects and teams.

Sure, the interface for our RAVN AI products is easy to set up and use. But the magic can only happen after they take the time to develop the different extraction rules that are required help the solution learn what to look for in the groups of documents or contracts they want it to review—as well as determine which ones should be triggered for each use case.

Though we have over 70 out-of-the box extraction models, customers still may find they would be better served developing their own or supplementing one of ours with more customized models they created themselves.

Our iManage AI University (AIU) service was born out of the realization that while we already had AI training programs in place, customers often required more in-person, in-depth knowledge to ensure they were utilizing the right rules within RAVN to get the results they desired.

The whole impetus around AIU is to merge our technology with our customers’ expertise and documents so that they move forward with a real AI project in a significant way. This provides so much more value than simply teaching customers which button to press.

There’s always a light-bulb moment when a firm realizes that taking the time to develop their own AI models and rules will be more effective and give them a competitive advantage in how they can better serve clients.

Learning to facilitate machine learning

During the 2-day training session, we visit the customer on site to help them learn how to create the extraction models they need for a specific use case or help them think about what rules-based approach they would require to achieve their goals.

But before we initiate an AIU, there’s a few things we require the customer to do. This allows iManage and the customer to be prepared to ensure that together we can help the technology learn what we want it to do.

In addition to having RAVN installed and running, we require customers to designate and book time with strategic stakeholders on their team to attend before we begin an AIU. These stakeholders include:

  • Partner representatives
  • Legally trained staff
  • Data and knowledge managers

We ensure these people are in the room during an AIU so we can validate their use case and ensure they are going to have the time and resources to work on it properly.

We also require they have identified and uploaded the documents they want to work with—as well as the datapoints they want to extract from those documents.

While this prep work could take a few days—or longer in some cases—it gives customers a better understanding of what’s involved in facilitating machine learning and how it must be managed. We’ve had prospective customers ask for us to run an AIU for them just so they can better understand how the process works and could be administered in their organization.

We also ensure all this prep work is taken care of a few weeks in advance of the AIU so we too can have the time to fully understand a customer’s use case so we can suggest the most optimal extraction techniques for their project.

Discovering how AI can work

During the AIU, we help customers create and utilize various models to discover all sorts of things in their documents, such as patterns, similarities, and complexities that machine learning can pick up. Sometimes that leads to the realization that the organization needs RAVN to review other groups of documents to return the results they desire.

We provide a suite of AI-powered tools and train customers on the best fit for each one. As they gain confidence in the technology, we are continually impressed at the innovative applications and results that customers discover for themselves after project 1.

One example was a recent LIBOR repapering projectd for a major investment and insurance company. After creating models to review nearly 20 data points on commercial mortgage documents, another group wanted to do a similar LIBOR review for bond prospectuses. This new group was able to reuse some of the models from project 1 for project 2. Even better, since some key data managers had attended the initial AIU they were able to start a new project with new team members seamlessly.

On the second day of an AIU, we reserve time for user feedback of what might be needed in another module or use case. That allows us to incorporate their insights into future product enhancements so we can further help teach RAVN how to help our customers. We also explore how the data we’ve extracted can be utilized within other tools, like creating a Clause Bank within iManage Insight so that knowledge can be leveraged for other projects and matters.

If you’re interested in an AIU, or learning more about how RAVN can work in your organization, check out more about it here.