In our first chapter of this Legal AI Series, we explored what legal AI is and how it is revolutionizing and optimizing the industry. In this chapter, we’ll take a closer look at the challenges that law firms face during the discovery stage.

For attorneys, “discovery” is the critical, information-gathering phase of a case—a time that’s spent collecting and reviewing evidence that will eventually become the building blocks for future arguments.

In times of yore, this process was primarily physical—it involved gathering actual documents, pouring over paper and ink photos, interviewing live witnesses, and combing through objects archived in evidence storage lockers.

These days, however, building a case isn’t so much about reviewing what you can see, taste, feel, and smell, as it is about what you can’t. It’s about combing through huge amounts of unstructured material that only exist as 0s and 1s inside a database somewhere. We’re talking, of course, about the millions of photos, texts, emails, videos, and tweets that are recorded and exchanged all over the world, every minute, of every day.

Currently, experts estimate that over 80–90% of the world’s data is unstructured—a number that’s expected to grow in years to come. This upward trend is slightly problematic for attorneys, who now find themselves drowning in unprecedented amounts of new material in the form of CCTV footage, Zoom calls, jailhouse audio, social media photos, smartphone videos, and so much more. This electronically stored information (ESI) can be incredibly valuable to a budding case. However, the sheer volume that needs to be reviewed as a part of “electronic discovery” (or “e-discovery”) is making it increasingly difficult for attorneys to manage, demanding a solution.


In this blog, we’ll cover some of the biggest hurdles that ESI presents to document review, including:

  • A lack of qualified talent
  • How expensive it can be
  • Fast turnaround times
  • Inaccuracies within the data
  • And concerns regarding security and compliance.

The Five Main Problems with Legal Document Review

The ever-growing monster of unstructured, ESI data has created a lot of unforeseen issues for certain industries. In the legal sphere, they present as five main challenges, which begin with finding a qualified workforce, and go on to include exorbitant costs, tight timelines, data review errors, and, of course, security concerns.

Here’s a closer look at each of these.

Hurdle 1: Finding a Qualified Workforce

The hurdles presented by so much unstructured data begin with finding a qualified workforce. Often media files—especially the kind contained in CCTV footage, attorney-client jail calls, and even personal smartphone videos—contain sensitive, confidential information that not everyone is privy to. Hence, firms cannot simply hire just anyone to conduct e-discovery.

Instead, case managers must retain licensed attorneys who are in good standing with the Bar Association. This can be a challenge, especially when you consider that large-scale reviews often require additional abilities, such as foreign language proficiency, or certain technical skills, to translate documents, videos, and other media that might pop up during discovery.

The pool of candidates who meet these high standards is often small, and the competition for their time is intense. Thus, the ever-increasing load of ESI content only amplifies one of the biggest challenges that every large-scale document review already faces: workforce availability.

Hurdle 2: Exorbitant Expenses 

The cost of paying workers to sift through a Mt. Everest of data would have been expensive enough, even without a skilled workforce. With them, however, the bottom line of each billable hour increases astronomically, and this is before you even consider the tremendous annual increase in video, photos, and other ESI content, which can easily skyrocket the time needed for e-discovery.

The cost of providing the space, equipment, and necessary amenities to house an army of this size creates a heavy burden on clients, who are already paying steep legal fees for representation—costs that only continue their dangerous, exponential climb when time, another important commodity, isn’t used wisely.

Hurdle 3: Tight Timelines

Time (or rather, the lack thereof), is another huge issue for legal document review. While the amount of unstructured ESI might be increasing, the court’s flexibility isn’t, and this puts attorneys in an impossible situation.

In order to meet tight deadlines, time has to be used efficiently, and to do that, reviewers need clarity from the onset of document review. Unfortunately, it’s difficult for legal teams to provide clear, concise direction without first knowing what evidence exists—the kind of evidence they don’t have until discovery is well underway.

It’s a chicken-before-the-egg conundrum that wastes precious seconds—time that is further squandered when teams lack the kind of software that can adequately sort, filter, and organize the growing mountain of complex media files being included for review.

Hence, as discovery reveals new insights, legal strategies inevitably shift, requiring additional reviews, more time, extra costs, and, what’s worse, a greater room for error.

Hurdle 4: Data Errors

In any large-scale review, a certain level of error is inevitable. One way attorneys attempt to reduce this margin is to conduct e-discovery in levels, with each file passing through several stages of review before getting a full pass. Unfortunately, while multi-tiered reviews can decrease the margin of error in some ways, they actually contribute to the problem in others.

In addition to the obvious (i.e. the increase in time and expense), multi-tiered reviews can’t function properly under a disorganized legal team—which, as we just discussed, most are at the beginning of a case. And in the absence of clear, concise direction, reviewers are left with too much room for interpretation, resulting in inconsistent coding, and wildly varying perceptions of what is and isn’t relevant. Problems that only increase, the more audio files, photos, and videos you add to the mix.

At a minimum, these miscommunications mean more rounds of review from the already time-strapped legal team. At most, they can result in missing a crucial smoking gun, or even the release of information that should have been redacted as privileged or confidential.

Hurdle 5: Security Concerns

Failing to redact privileged or confidential information can have serious consequences—not just for a single case—but for the fate of an entire firm. Risks have evolved from the improper release of text documents to include things like identity leaks, photos of a home address, license plates, conversations, private activities, and more. And, unfortunately, these aren’t even the only hazards that can hamper the success of document review.

Confusing, hard-to-use software, inadequate firewalls, off-site reviewing, and even the rogue, disgruntled employee are all examples of security risks that can compromise the integrity of a lawsuit.

In a world where saving, stealing, and exchanging information is easier than ever before, law firms are searching for methods that not only provide answers to these important security questions but that also address the other pressing problems that the increase of ESI presents to the e-discovery process.

Overcoming the Hurdles of Legal Document Review

We might not be able to touch, taste, feel, or smell electronically stored information, but that doesn’t mean this content isn’t having a very real impact on legal document review—an impact that can’t just be ignored. On the bright side, the human ingenuity that pushed us into this technological era is more than capable of meeting the challenges that it caused.

Artificial Intelligence (AI) has proven to enhance and augment human capabilities to overcome the most modern challenges. By adding this technology to the eDiscovery world, we can overcome many of the hurdles we have outlined here. But before we can dive into how it solves these problems, we need to first have a clear understanding of artificial intelligence—such as what it is, what it can and can’t do, and more—which we’ll explore in the next blog.