AI Can Further Yield Cost-Efficient Document Review
Ediscovery and Legal Tech
Statistical modeling and unsupervised machine learning techniques have been around for decades. Using the internet and other enormous data sets have offered new ways to provide wisdom to the masses using a common body of knowledge.
Wikipedia was a not-for-profit repository of crowd-sourced articles that initially attempted to provide this wisdom; Google attempted a more egalitarian way of organizing website information to let the user decide what was important to them without human curation.
Ediscovery companies have been known to be sitting on a massive corpus of enterprise data from all the matters that they handle, oftentimes creating full replicas of an entire company’s data including emails, user-generated documents, and collaboration chat. This data is not internet data however, but proprietary and confidential in nature.
Chat GPT
Imagine the combination of #ChatGPT, the enterprise multi-cloud, and in-place #ediscovery.
What if ediscovery productions could be fulfilled by one employee at the company using AI working with one attorney at outside counsel? It doesn’t work that way today. The need for an army of document reviewers and attorneys is often the reason for this being the costliest component of the litigation process.
It’s time to ideate software with more efficient, accelerated workflows because #ChatGPT doesn’t try to break ground on new economies (e.g., cryptocurrencies, virtual worlds), but it creates new economies of scale on the time returned to the knowledge worker on normal everyday activities.
Adoption tends to be much faster with the latter type of invention. That’s why #ChatGPT reached 1MM registered users in an historic six days, faster than any other software platform.
There’s much less time spent explaining what’s happening underneath the hood as a new shiny toy and more time illustrating the immediate benefits without a ton of technical jargon.
Benefit:
#ChatGPT offers a single text input box much like the simplified GUI of other search engines, but rather than inputting keywords using Boolean search syntax, the user can employ natural language chat to provide the desired context. The results are returned in an intuitive way, as if the machine was having a friendly dialogue with the user. You can literally ask #ChatGPT to explain the results to you as if you were five-year old, if you like.
Who benefits?
Users who formulate thoughts in their native language very concisely benefit the most. The habit we’ll all need to shake is sloppy keyword queries. I think we all know the amount of time wasted trying to get a search engine to read our minds.
A New Breed of Ediscovery Software
#OpenAI, the creators of #ChatGPT, are funding startup ventures. This means the tech that they employ can be used for different domain-specific applications. The training data can be enterprise specific data used for input, and the chat engine can be leveraged by API to drive the instruction set for document review without all of the upfront searching, tagging, organization, and filtering steps. I should acknowledge that there is AI software available to automate this upfront data classification as well and has been available for many years.
If designed correctly using the latest #ChatGPT technology, the tagging, organization, and filtering can be performed with a simple natural language query on demand. For example, the human reviewer can ask, “Please withhold all attorney-client privileged communication into an isolated review set and generate a privilege log in Excel and read out the document count. Create a priority batch of documents between our priority list of key custodians who communicated with one another by email between Jan 2016 and Dec 2019 and read out the document count. Tag those as Relevant by default.”
And all of the reporting can be confirmed by the AI through chat reply. For example, the Chat Bot can respond, “There are no untagged documents remaining for review. All tagging conflicts have been resolved, and the production is ready for export.”
This doesn’t intuitively feel too farfetched now that #ChatGPT use cases have flooded social media and we’ve all seen the power of this technology firsthand.
Jerry Bui is Managing Director of Digital Forensics within FTI Consulting’s Technology segment focused on forensic technology and risk & compliance issues (all opinions his own). Jerry is a Certified Fraud Examiner and has over 20 years of experience in digital forensics, ediscovery, automated risk assessments, dashboard compliance monitoring, and investigative analytics. Jerry’s team provides evidence acquisition, expert witness, and strategic consulting services to law firms and corporations. Connect with Jerry on LinkedIn, Twitter and TikTok.
The Digital Forensics Future (DFF) podcast is also available on the platforms below.
