eDiscovery Leaders Live: Jason Bentley of K2 Discovery
Jason Bentley, Senior Sales Engineer at K2 Discovery, joins George Socha, Senior Vice President of Brand Awareness at Reveal, for ACEDS #eDiscoveryLeadersLive.
With more than 20 years in litigation support and eDiscovery, Jason is a proven problem solver with the ability to understand diverse technical challenges and deliver solutions which consistently exceed expectations. Jason has been a leading force in many areas of the litigation support industry including paper discovery, EDD (starting in 1999), data processing, document review and production services, project management, and the management of teams across all aspects of eDiscovery software development and deployment.
Jason gave up an in-depth look at how K2 Discovery takes advantage of Reveal’s AI capabilities to help clients to speed up the review process and find content they otherwise might not locate. He used the lens of the construction defect matters they work on and the mid-sized firms they assist to give specific examples and detailed explanations.
Key Highlights
-
- [1:11] Introducing Jason.
- [1:47] Jason’s background.
- [2:43] How K2 Discovery uses AI in construction defect cases.
- [3:54] Using AI models to find contracts.
- [5:35] Using AI models: Finding change orders and finding prices and fees.
- [6:33] What it is like using AI models: selecting and deploying AI models.
- [7:54] What it is like using AI models: Enhancing models via supervised learning.
- [8:21] How long it takes for AI models to run.
- [9:16] Efficiencies gained by using AI models.
- [9:51] Determining which models to use.
- [10:57] Finetuning and grouping models.
- [12:18] More about supervised learning.
- [13:36] The value of working with multiple issue codes (and models) at the same time.
- [14:55] An iterative learning process starting with 4-10 documents tagged.
- [16:26] Their mix of clients and the matters they handle for them.
- [17:13] Helping level the playing field for midsized firms.
- [18:00] How they help clients in this process.
- [19:12] AI-driven image labeling and its effectiveness.
- [20:57] The ease of producing documents using Reveal.
- [22:14] Helping midsized firms with AI-driven transcription.
- [23:40] Helping clients move to a Reveal with ease.
- [26:50] A team approach to supporting clients.
- [28:37] Great client retention rates.
Key Quotes
- “A lot of my current cases are construction defect type cases…. We design our own workflows to help our clients get to the heart of the case very, very quickly. We’ll use all the different prefabricated AI models [in Reveal]. We will use supervised learning where we are creating issue codes and letting the system guide them through what they need to look at next…. What we find is it’s a very effective way for our clients to get their handle on the case very early on and make the important decisions that are often needed to be made early on.”
- “We’ll use each of those prefab models to group those types of documents together and present them to the attorneys or paralegals…. It saves them weeks and weeks of time often because they are not iterating through documents one after another in a linear fashion.”
- “We will have a meeting with the attorneys involved, learn about the issues in the case, understand what they’re looking for. Then we’ll apply the models that make the most sense to that, group them together, and have them look at them. As they continue to find important facts in the case via the data they look at, we continue to build the AI models in a different manner called supervised learning.”
- “[To get started with supervised learning,] you need anywhere between four and ten documents, depending upon on you set it up, that need to be coded with a certain issue tag.”
- “[With image labeling] they’re able to say, here’s the 5,000 photos, here are the 50 that have to do with piping, and they are able to look through the 50, find exactly what they’re looking for, and then move on with their day.”
Connect with Jason