Blog Post

eDiscovery Software Hasn’t Replaced Humans, But They Need to Work Together

October 20, 2020

Seeing what software is capable of today is enough to turn anyone into a Luddite. Automation, artificial intelligence (AI) and advanced software have already made millions of jobs obsolete. There aren’t any typists or switchboard operators anymore. Manufacturers, drivers and many other human-based roles are rapidly disappearing. Are paralegals and lawyers next? For many in the legal industry, eDiscovery software serves as the herald of an upcoming wave of automation. Document review and eDiscovery — essential components of litigation prep — can still be performed manually, but doing so takes up huge chunks of billable hours or requires expensive outsourcing. Increasingly, legal teams are instead turning to eDiscovery software that uses AI to accelerate the document review process.

How AI-powered eDiscovery  software works

The most common AI method used to support eDiscovery is known as predictive coding, or technology-assisted review (TAR). Under this approach, an expert takes a random sample from their total document set and codes the documents within for responsiveness or unresponsiveness. This seed set can then be used to teach the eDiscovery software what features make a given document potentially responsive to the matter. Based on this initial seed set, the software then searches through the document set, flagging potentially responsive documents.

Where humans fit in with eDiscovery software

This technology not only allows for human expertise, but requires it. eDiscovery software with predictive coding capabilities are often referred to as a “garbage-in, garbage-out” systems. If the seed set doesn’t consist of accurately coded documents, then the output won’t either. That’s why eDiscovery software requires senior, expert lawyers to develop this seed set, to check output as eDiscovery is ongoing and to correct the course if the software is producing erroneous results.

This means that rather than replacing its human counterparts, AI-powered eDiscovery software requires humans. Legal predictive coding literally cannot identify responsive documents without being first trained by a lawyer on what constitutes responsiveness for the matter at hand.

At the same time, it’s likely that lawyers will find themselves reliant on eDiscovery software and AI. In 2010, the world produced 2 zettabytes of data in a year. By 2020, that number had exploded to 59 zettabytes. By 2024, the world is forecasted to produce 149 zettabytes in a year. If we keep producing more and more data, litigation document sets are going to increase commensurately. We’re already at the point where manual review is an incredibly time-consuming process that can be feasibly performed for matters with large datasets; it may be the case that document review without AI assistance becomes effectively impossible.

What about the future?

AI — and especially AI in the legal industry — still has a long way to go. For the foreseeable future, legal AI will likely have the greatest impact in eDiscovery software and will likely still require expert supervision.

Attempts at more sophisticated legal AI have been made, such as NextLaw Lab’s ROSS, a legal research software based on Watson (IBM’s Jeopardy!-winning AI). However, solutions like ROSS only serve as virtual assistants to lawyers, facilitating research, identifying overturned or criticized cases, providing intelligent answers to questions and so on. They’re not defining strategy, making judgements or providing nuanced advice to clients.

There are some impressive solutions out there, but there aren’t any that can actually replace a living, breathing legal professional. For now, these tools only augment legal professional’s capabilities.

Of course, AI solutions may become so sophisticated in the coming decades that they really do replace lawyers — but by that time, every other human profession may be automated as well. By the time AI becomes sufficiently advanced so as to do the work that legal professionals do today, the world will be so radically different it’s not worth worrying about from a practical perspective.

In the meantime, legal professionals can enjoy the benefits that AI-powered eDiscovery software and other legal technologies offer without worrying that they’re automating themselves out of a job.

If you’re curious about what the use of AI in eDiscovery looks like in practice, read about our work for Vedder Price, where we selectively reduced a document set of half a million potentially relevant documents down to just 14,000.