If you haven’t been paying attention (and bless you if that’s the case), eDiscovery has been overrun with claims about how AI is transforming everything and eDiscovery will never be the same. There are truly transformative technologies hitting the market (ahem, Exterro ARMOUR). Between autonomous agents, prompt engineering, and large language models (LLMs) that churn out complete legal memos in seconds, it’s easy to feel as though the rulebook has been rewritten overnight.

But the fact is, the foundation of defensible eDiscovery hasn't vanished—it has simply found a new application.

In a recent webinar titled AI in the Courtroom: How Generative AI is Changing the Way Legal Teams Go to Court,  hosted by Doug Austin, editor of eDiscovery Today, a panel of legal technology veterans—including Martin Tully, Kelly Twigger, and Bryant Bell—tackled the noise surrounding Generative AI. Their overarching message was clear and reassuring: you don't need a brand-new legal framework to handle AI. You just need to apply the time-tested principles of eDiscovery to modern tools.

Grounding "Jetsons Tech" in "Flintstone Rules"

When faced with emerging technologies, legal teams often panic, assuming they need entirely new rules of civil procedure or evidence. But as Martin Tully pointed out during the discussion, the core duties governing discovery remain unchanged: relevance, proportionality, privilege, and technical competence.

"We need a Flintstone approach to Jetsons technology," Tully noted, reminding practitioners that legal common sense still dictates how we handle new data types. 

Stripping away the sci-fi branding reveals that LLMs are ultimately probabilistic software. Just as courts expected legal teams to understand the mechanics and limitations of Predictive Coding and Technology-Assisted Review (TAR) a decade ago, they expect attorneys to understand the inputs and outputs of Generative AI today.

These core principles apply to Large Language Models (LLMs) because, despite their complexity, they remain tools for information processing. Because they operate as such, they must function within established legal boundaries where the obligations of relevance, proportionality, and rigorous verification remain constant, regardless of the sophistication of the underlying technology.

Treat the AI Like a First-Year Associate

One of the biggest anxieties around Gen AI in litigation is the risk of hallucinations—where a model confidently fabricates facts or legal citations. But rather than viewing AI as an infallible oracle, legal teams should treat it like a brilliant, eager, but inexperienced team member.

Kelly Twigger emphasized the necessity of maintaining a "human in the loop" when incorporating AI into eDiscovery workflows. She noted that outputs from Generative AI tools should be viewed as a baseline or Minimum Viable Product (MVP)—a starting point that requires rigorous human review, validation, and oversight before it ever sees the inside of a courtroom.

“It's almost the equivalent of using someone to help you draft something, but then it's your responsibility to check everything that's in it and make sure it's valid, that it's well-reasoned, that the cases that are cited are actually referenced.”

• Kelly Twigger

Just as a partner wouldn't sign off on a motion written by a first-year associate without checking the citations, a legal professional cannot rely on an LLM output without verifying the underlying source data. This human oversight isn't just best practice; it's an ethical requirement.

Prompts Are Just the Latest Form of ESI

In the early days of eDiscovery, the struggle was converting paper to TIFFs. Then came email, followed by mobile text messages, and eventually ephemeral messaging apps like Slack and Teams. Generative AI is simply the next iteration of Electronically Stored Information (ESI).

During the webinar, Bryant Bell highlighted how the focus is shifting toward the inputs of these systems—specifically, user prompts.

Prompts and chat logs are increasingly being sought in discovery to establish a user’s intent, state of mind, or pre-existing knowledge. If an employee asks an internal corporate AI chatbot, "How do I work around this compliance rule?" or "What is our liability on this contract?", that prompt becomes a piece of discoverable evidence.

Understanding this reality allows legal operations and litigation teams to update their custodian interview scripts, ESI protocols, and legal hold processes to account for AI chat histories today—using the exact same mechanisms they used for Slack threads yesterday.

The Path Forward: Evolution, Not Revolution

Generative AI isn't a lawless frontier. It is a powerful tool operating within a well-established ecosystem of rules, court precedents, and ethical guidelines. By applying core eDiscovery fundamentals—proportionality, verification, and proactive information governance—legal teams can confidently adopt AI while building a defensible posture for litigation.

Want to dive deeper into the conversation?

This blog post only scratches the surface of what was discussed. To hear the full debate—including practical advice on AI governance, authentication of machine-generated evidence, and real-world case law examples, access the on-demand webinar here.

‍