How AI Solves One of the Biggest Pain Points in Doc Review

October 25, 2023



Sarah Moran
Sarah Moran

Most legal teams would say that reviewing the same documents multiple times across multiple matters—aka “repeated review”—is a costly but necessary part of eDiscovery. With legacy tools and methods, that may be the case. You start each new matter from scratch, regardless of how it overlaps with previous or concurrent matters.

But today’s tools enable a different approach. Artificial intelligence, concept clustering, and even traditional TAR models can be deployed in ways that capitalize on past work and reduce human review hours.

In separate posts, we explained how starting small is the key to breaking out of the repeated review cycle and shared three ways technology can help. Now we’ll dive deeper into technology and focus on AI specifically.

So, how can modern AI capabilities help reduce repeated review?

Reuse past coding decisions instead of repeating eyes-on review

Modern AI tools keep a record of how documents are coded. The first time you use AI on a matter, it will record the coding decisions for every document in the corpus. When any of those documents come up in future matters, the AI will show you how they were coded previously. If a document was coded multiple times, AI reflects the whole history.

This is a true game changer for classifications like privilege and other sensitive information. Relevancy may change from matter to matter, along with strategy, scope, and countless other details. But privilege is much less variable, and other types of sensitive info will always be sensitive. If a document is considered PII or PHI one time, it should be coded that way every time.

With AI, you can make sure that’s true. By taking a document’s coding history into account, attorneys can code sensitive information consistently every time, so they don’t risk inadvertently sending it on to production. They also don’t spend time and money redoing someone’s previous work.

These benefits are more than theoretical. Firms and corporations are using modern AI to reuse coding right now, in live matters. For example, a global pharmaceutical company took this approach on a set of 5 matters and was able to reuse a total of 26,000 previous coding decisions.

Beyond repeated review: increase intelligence and accuracy over time  

The power of modern AI to leverage past decisions can also give teams an edge when approaching new documents. This comes from modern AI’s ability to “get smarter” the more data it ingests.

With each matter, modern AI refines and adds to the massive repository of rules that inform its analysis and recommendations. It learns, for lack of a better word, what counts as privilege and other classifications—including all the nuance that surrounds how you define those things for your data specifically. So it gets better and better at making those classifications as time goes on.

It’s not unlike the institutional knowledge that review attorneys develop when they work for the same client regularly. Clients like to utilize the same review teams again and again precisely because of that knowledge. Attorneys who are familiar with the nuances of your data and decision making can work much faster and more accurately than attorneys who are coming in fresh.

Modern AI accumulates and integrates that knowledge in much the same way—with potentially stunning results. After using modern AI on every matter over 3 years, the eDiscovery director at a global tech company saw it shrink privilege review on a new matter by almost 90%.

“We expected nearly 190k documents would be subject to privilege review using our typical workflow; with Lighthouse AI and outstanding outside counsel, our actual results were just over 24k.” —Director of eDiscovery, global tech company

An important detail here is that you don’t have to retain data from past matters in order for modern AI to learn from it. Once a matter is complete, modern AI will automatically update its rules and rubrics for decision making, and you can delete the matter. This is impossible to do with a traditional TAR model, which needs files on hand in order to extract and apply learnings from them.

Be aware, not all AI platforms have the same capabilities

Today, AI is everywhere in the eDiscovery marketplace. But it’s not all the same AI.

The inner workings vary quite a bit—from those that use machine learning, which has been around since the 1970s, to those that use modern innovations like deep learning and large language models. The applications and benefits of different AI offerings are also very diverse.

To find AI that helps reduce repeated review, ask questions like:

  • Is this tool able to reuse coding?
  • Can the model tell me the historical coding of a document?
  • Does it refine its model and analysis over time?
  • Does it do this without retaining data?

Also, remember that AI is one of many options for reducing repeated review. Some options require no technology at all, just some forethought and strategy for how to make the most of each review cycle.

For a closer look at how repeated review holds you back and what you can do about it, explore our deep dive on the subject.

About the Author

Sarah Moran

Sarah is a Director of Marketing at Lighthouse. Before coming to Lighthouse, she worked for a decade as a practicing attorney at a global law firm, specializing in eDiscovery counseling and case management, data privacy, and information governance. At Lighthouse, she happily utilizes her eDiscovery expertise to help our clients understand and leverage the ever-changing world of legal technology and data governance. She is a problem solver and a collaborator and welcomes any chance to discuss customer pain points in eDiscovery. Sarah earned her B.A. in English from Penn State University and her J.D. from Delaware Law School.