Are You Becoming an eDiscovery Reverse Centaur?

Doctorow’s writing is great, so give it a read.

The heart of eDiscovery has always been extracting the evidentiary story from mountains of raw ESI. If AI can ‘summarize’ that mountain via a RAG-LLM model, does that make eDiscovery professionals into Cory Doctorow’s ‘Reverse Centaurs’, i.e a squishy meat appendage for an uncaring machine? I think my recent return to coordinating matters for clients has shown me our value proposition in the eDiscovery lifecycle.

First a bit of context (regular readers can skim/skip):

In 2018 the leaders in the analyst market (Gartner/Forrester) seemed to decide that Microsoft Purview would convert eDiscovery from an independent technology and service market into an IG feature. That killed eDJ Group’s plan to be acquired, so the team disbanded and I returned to my strategic eDiscovery consulting practice under the same name. During the 2020 Pandemic pause, my AI mentor Skip Walter recruited me into a fun start-up Personal Knowledge Management (PKM) application that applied eDiscovery analytics and actions to connect all your various data sources. The early versions of ChatGPT were terrible and once again I was too early to market with a great idea.  That left me delivering strategy consulting in the face of rapidly evolving AI capabilities. Late in 2024 I looked at those capabilities and let my long-term clients know that I was now available to coordinate and support their ugly, hot matters. 2025 was an exhausting and satisfying ride. The new AI tools are powerful and will indeed kill a lot of the 1^st pass relevance/privilege market (IMHO). I also encountered a lot of ‘AI resistance’ from conservative counsel. The context should explain how my return to using and testing cutting edge AI has shown me what it does well and why I think that is a better tool than agentic eDiscovery replacement.

Back to who is driving the eDiscovery bus:

The Redgrave LLP aiR vs. RAL study proved the importance of the input content and decisions. The sheer hours required to craft and validate the aiR prompt by a SME supports AI’s need for quality, expert input. The recall and precision of Active Learning models is dependent on accurate, consistent reviewer decisions. I think that the input issues will eventually be resolved using diverse, parallel models/methods with dynamic feedback.

While aiR for Review gives a relevance score and reasoning/citations to support that score, it struggles with deceptive or inferred context. ESI based investigations were relatively easy back in my Enron matter days. Most of our bad actors were not technically sophisticated. Sure, broker-dealers would hide notes in white text, wingdings font or try to double delete evidence. My recent investigations have shown a high level of target awareness regarding corporate monitoring, logs and digital fingerprints.  They take their real conversations off-line, IRL or just exercise good conversational discipline.

Productions are based on identifying everything reasonably relevant to the discovery request. Productions should be a rare, unfortunate consequence of the failure to resolve the matter upstream. Counsel needs the real story along with the underlying facts and evidence. Even excellent tools like aiR for Case Strategy struggle to fill in Facts that a human can synthesize from indirect patterns, absent data, tone and experience with the human relationships. That is why I think the future is bright for practitioners who want to use AI output and their own review to elevate the truth.

The intelligence community has long used sophisticated technology to process and analyze raw data from diverse sources (All-Source Fusion), connect the dots and produce finished intelligence. eDiscovery is still in the broader adoption phase with this new generation of tools. Crafting recent investigation reports has highlighted just how much my 30+ years added to what AI surfaced or missed. No one should mourn the loss of repetitive, mind-numbing discovery processing, review and raw analysis tasks.  An AI-driven reverse centaur workflow will never generate quality hypothesis generation or make intuitive leaps. And no human will be better at basic pattern recognition from raw data. We can see the unsaid and surface latent connections. Maybe AGI will make me eat my words and kill all our roles. So what. We live in an age of innovation and adaptation. Embrace it and have confidence in your own ability to thrive. Trust yourself and be that centaur.

Greg Buckles wants your feedback, questions or project inquiries at Greg@eDJGroupInc.com.  Reach out for a free 15 minute ‘Good Karma’ call if he has availability. He solves problems and creates eDiscovery solutions for enterprise and law firm clients.

Greg’s blog perspectives are personal opinions and should not be interpreted as a professional judgment or advice. Greg is no longer an investigative journalist and all perspectives are based on best public information. Blog content is neither approved nor reviewed by any providers prior to being published. Do you want to share your own perspective? Greg is looking for practical, professional informative perspectives free of marketing fluff, hidden agendas or personal/product bias. Outside blogs will clearly indicate the author, company and any relevant affiliations. 

Greg’s latest nature, art and diving photographs on Instagram.