ai and computer

AI: Everything’s an Exception in eDiscovery

Share this article

Many service providers will tell you that leveraging AI in eDiscovery is the equivalent of using an easy button. Too many repetitive tasks? AI will automate them. Improve on previous cases? Predictive analytics to the rescue. Hunting for small amounts of data among millions of documents? A simple task with AI.

While we might not have an easy button, AI isn’t as difficult to deploy as most people think. You just have to be aware that use cases for AI aren’t identical. There are different wrinkles, and subtle nuances that distinguish each one in every matter.

Finding the benefit

The most important question to ask is: What benefits will AI provide for your particular use case? While it’s possible to leverage AI in most eDiscovery cases, you need to be certain the result will yield value. If you don’t have the internal resources to confidently identify opportunities where AI is likely to be beneficial, it helps to bring in a trusted advisor or partner. You can also educate yourself about potential use cases that might apply to you, and then look for a solution.

The maturity of your organization’s legal operations directly affects decision-making regarding whether to use AI internally or bring in a consultant. Outside help might be necessary to dig down into the specific requirements of your use case and determine whether AI helps you achieve the desired outcome more effectively and at a lower cost. 

Calculating cost: It’s not what you might think

When deciding whether to deploy AI in a particular use case, keep in mind the value is in the outcome that the technology provides, not what the technology costs. Technology in eDiscovery is rarely the biggest line item. Cost is typically driven by time to insights or time to relevance.

The time spent trying to find answers or responding to requests has the largest impact on cost (attorney review time). The higher the time, the higher the cost. For example, inefficiencies such as straight linear review, overbroad or narrow search parameters, or the slow education of matter topics and themes have a large impact on cost. An upfront investment in technology that reduces time to insight or time to relevance has tremendous downstream-cost savings affecting the larger portion of an eDiscovery budget. You can gain additional savings by investing in a platform that offers AI features at no additional cost.

AI tends to be most cost-effective in cases involving large volumes of data, tight deadlines, high levels of data complexity, and repetitive tasks prone to human error. For example, one of my clients – an Am Law 200 firm – was faced with reviewing more than 600,000 documents in a complex construction case. The firm chose to deploy technology-assisted review (TAR). Doing so allowed staff to reduce the number of documents for review by more than 90%, meet the deadline, and complete the review for 50% under budget.

Another multinational company I worked with was required to search 8 terabytes of data for a Foreign Corrupt Practices Act (FCPA) investigation and produce relevant data within three weeks. That company used AI alongside ECA tools to cull the number of documents by 98% at lightning-fast speed.

In yet another use case, a global organization that was my client faced an internal investigation in which 60,000-plus multilingual documents needed to be searched and culled in both English and Japanese on a strict two-week timeline. The company’s legal service provider used active learning technology to reduce the initial review set by 50% in a single day. It ultimately achieved a 97% overall cull rate and completed the project on deadline and $200,000 under budget.

Typically, most organizations turn to external eDiscovery providers in situations where they believe the provider can manage the process at a predictable cost, negotiating hard to get the biggest bang for their buck. Historically, clients negotiate with eDiscovery providers to reduce unitized costs (e.g. hosting, first pass review, etc.) to the lowest rate. This “race to the bottom” methodology distracts from increased efficiencies and achievement of business outcomes, and breeds vendor-hopping – which leads to change-management nightmares. AI is often a key factor that enables providers to come in on time, under budget, and with defensible results.

While some providers impose additional charges for using the AI options in their platform, a decision about whether to use AI should never be based on the cost of the technology alone. If you’re uncertain whether using AI is appropriate in a particular use case, ask a trusted partner.

A trusted partner will be an invaluable resource to align the benefits of AI to the client’s business outcome. Even if you don’t understand exactly how the technology works, you can ask your partner to supply explanations and validation to demonstrate positive results.

How can we be sure AI is working? It has to help us achieve the outcome we want. This might mean meeting an “impossible” deadline, reaching a target cull-rate, or a certain level of accuracy. It could also mean gaining early insight into the data that helps you budget accurately and set legal strategy, or perhaps establishing process improvements that make eDiscovery more efficient.

Two areas where AI often proves especially valuable are investigations and compliance. That’s because the technology can quickly find whether certain kinds of information exist in very large bodies of data and in diverse data types. But it can also be indispensable in major litigation where parties are obligated to use reasonable efforts to find relevant material.

In many cases, you’ll need to align AI work for your particular matter, and that’s where a trusted adviser comes into play, especially in complex cases. A good provider won’t just offer a cookie-cutter AI solution; the best should be able to tailor the workflow to ensure you achieve a successful outcome.

Start with low risk and build from there

Many legal departments believe they can benefit from AI, but get overwhelmed and don’t know where to start. Having a good first experience is important; it’s essential you give key stakeholders a good outcome. We recommend you begin using it on relatively low-risk but medium-impact business problems, or perhaps a series of smaller eDiscovery projects that are similar in nature and likely to benefit from the reuse of proven TAR models.

Whatever your potential use case, it’s important to have a goal or destination in mind before deploying AI tools — and you might need to lean on a partner for help. Look for a complete solution, not a pre-selected technology. The partner should guide you to a technology based on their analysis of your particular business problem.


If you’re not leveraging advanced technologies to help solve at least some of your eDiscovery challenges, you’re probably wasting time and spending more than you need to on some projects. Many law departments and law firms are already using AI because it gives them more agility and new ways to control costs. Find out for yourself by talking to a reputable provider who can help you determine which kinds of use cases common to your organization might benefit from the application of AI technologies. With a trusted partner at your side, you can begin exploring how AI can help you do eDiscovery more efficiently.

Oliver Silva on Email
Oliver Silva
Director of Enterprise Accounts at Casepoint
Oliver Silva is director of enterprise accounts at Casepoint, an AI-powered technology platform that helps leading corporations, government agencies, and law firms meet complex eDiscovery, investigations, data preservation, and compliance needs. To learn more about Casepoint please visit
Director of Enterprise Accounts at Casepoint

Share this article