The Legal eDiscovery Market at a Turning Point

Every eDiscovery team is currently weighing the same question: how far can Generative AI (Gen AI) really take us without compromising defensibility? The hype around Generative AI promises speed, scale, and the avoidance of 1L document review, but with those claims come critical questions:

• Can we avoid 1L review altogether?
• Will Gen AI deliver efficiency beyond what we already achieve with Continuous Active Learning® (CAL®)?
• Will Gen AI obviate the need for CAL®?

Is 2L review the new 1L review?

Many of the current Gen AI claims suggest that we will be able to bypass 1L review entirely. The marketing distinction between 1L and 2L review for corporate or law firm customers drifts towards an implication that it is outsourced review (1L) versus firm or client review (2L). The claims suggest that firms and corporate customers will no longer need to outsource reviews, which is an appealing message for customers.

Unfortunately, the true driver of outsourced review has always been the volume of documents and short timelines involved. Whenever production sets grow beyond 10,000 documents, outsourced review becomes indispensable for cost savings and rapid scaling. This is an area where Gen AI can potentially create advantages by getting to the initial cut of the production set more quickly, assuming, of course, that the rounds of prompting do not take longer than the review/training in CAL®.

Most of the claims about avoiding 1L review seem to be comparisons against a full linear review of all documents that survive culling. Gen AI is indeed much better than linear review. CAL® is likewise far better than linear review, and we have been using that very effectively for the past five years. In this sense, a claim about 1L avoidance is like telling someone they can avoid walking if they buy a particular type of car. CAL® and Gen AI are two types of cars; they’re both superior to walking.

Modern 1L review has evolved to mean only the documents we need to review in a CAL® process: all production documents and associated family members, along with a significant sample of the non-production documents to ensure coverage. It does not really matter whether we call it 1L or 2L review, because the volumes we review in a Gen AI or CAL® will be very similar. There will still be a first review pass on a significant portion of the documents, and if the volumes are high enough, this work will be handled through outsourced review.

Will Gen AI Beat CAL®’s Proven Efficiency

CAL® has long been the backbone of modern Technology-Assisted Review (TAR). By iteratively training models based on reviewer feedback, CAL® dramatically reduces review populations by over 76% on average, as measured across projects over several years. These efficiency gains often translate into significant cost savings and faster turnaround times.

For Gen AI to outperform CAL®, it would need to reduce review volumes beyond what CAL® already eliminates. For the sake of signing off on Substantial Compliance, customers will likely continue to want to review all production documents and associated family members, along with a significant sample of the non-production documents to ensure coverage. This means there will be little or no additional savings from using Gen AI instead of CAL®.

There are also situations where neither CAL® nor Gen AI will produce any savings. When the responsive rate approaches 100%, we end up reviewing all the documents as with a linear review. These factors can and do create situations where there are no cost savings at all. This will be effectively true for Gen AI-based processes for the same reasons.

Richness Drives Savings

Where Gen AI Fits, and Why CAL® Still Matters

Gen AI introduces new possibilities for accelerating document review, particularly for early triage and prioritisation. However, it doesn’t eliminate the need for CAL®. Even if Gen AI is deployed earlier in the workflow, CAL® can be helpful in improving confidence in the overall outcome of the process. Courts and regulators demand transparency, and CAL® provides the structured, defensible, and trusted framework that Gen AI alone has not established in the legal marketplace.

By running CAL® on the Gen AI results, we can learn to use both tools better. Determining strengths and weaknesses of each will allow all of us to better understand the appropriate use cases for each.

Many of the gains now attributed to Gen AI may, in practice, already be realised through CAL®. If cost is the main driver for decision-making, CAL® may still be the best option. If speed and large volumes are the main concerns, we could easily recommend deploying Gen AI. While Gen AI can summarise and prioritise, its outputs still require human oversight until it can be fully trusted to produce high precision and recall repeatedly. CAL® offers that validation layer, making it the keystone for any Gen AI-based review strategy.

Conclusion

The future of eDiscovery is undeniably Gen AI-driven, but not Gen AI-only. CAL®, particularly when integrated through platforms like RelOne’s Review Center, will continue to define how we deliver verified, defensible results in an era of generative technology. Legal teams will embrace Gen AI, but will adopt more rapidly with CAL® at the core. It’s both a best practice and the key to confidence in every review.

Want to explore how Gen AI and CAL® can transform your review process? Contact Elevate via the form below for a consultation.