What can eDiscovery practitioners learn from the Chinese AI bubble? My concern is not predicting when frontier AI valuations will collapse. It is whether our technology workflows and licenses make economic sense if today’s AI pricing changes. The linked analysis cites more than $27 trillion in AI market capitalization over the last 3 years, while frontier AI labs are generating <$140 billion in revenue. I am not an economist, and market capitalization is not investment cost; those figures alone do not establish that the business model is unsustainable. But they do make me question the expectations behind the valuations. In our tiny legal services segment—cited here as 1.5% of the U.S. market—we are burning frontier-class tokens to read our email, summarize documents and otherwise cannibalize our billable hours with ever-expanding AI agents (thanks Vipin). Our economy is betting that massive AI infrastructure and research investments will generate investor wealth while reshaping a service sector representing 76.4% of U.S. GDP. That leaves a practical question: what can eDiscovery practitioners do to protect their roles and technology investments?
Understand the AI embedded and dependent functions
For the last 3 years every NY Legal Week exhibitor booth has been plastered with “AI Powered!” proclamations. It is our responsibility to ask hard questions and be educated consumers such as:
- What functions actual use an AI LLM or other model?
- Many agentic workflows have multiple AI steps/cycles. You must understand what is behind the output to be an educated consumer.
- Is that AI model an external frontier model or an internally deployed open-source AI model?
- Open-source AI models may not be as powerful or accurate, but they can be isolated and much more efficient for dedicated skills.
- What is sent to that external AI model and what does it return?
- This is basic data/privilege security. It is also important to manage your overall token burn.
- What are the token or frequency limits on that AI powered function?
- Responsible AI utilization means only burning tokens for high value outputs and working within base Annual Contract Value (ACV) subscription limits. Just look at how Microsoft had to ban Claude use by developers when the bill came in.
- How does the provider or service make a reasonable margin from AI functions?
- If it is a loss leader or buried in a larger $/GB hosting fee it is not sustainable.
- What agentic AI functions have the potential to accelerate cyclical token burn and how to you protect yourself against surprise bills?
- AI agents are the rage and one savvy citizen developer/vibe coder can do amazing things for an astounding token cost if you do not have guard rails in place.
- What are the matter/document/token costs for a given function and who should approve those estimates?
- This really requires a formal workflow to document costs and justify the value. Map out AI use and give teams approval thresholds before they dump entire custodial mailboxes into your hosting platform for AI reasoned analysis on a low exposure investigation.
Invest in your AI skills and sharpen your critical thinking
AI is an educational accelerator if you leverage it properly. Take CLE’s or watch videos on basics to get up to speed. Practitioners need to be comfortable using these tools without abandoning your critical oversight of the process and output. Run parallel chats, projects or development tasks with multiple AI engines. Use them to check the other’s homework and explain ANYTHING that you do not comfortably understand. Treat those comparisons as a way to surface questions, not as independent verification: agreement between models is not a substitute for checking the underlying sources and actual documents. While the models are rapidly evolving, project markdown (.md) files and repeatable AI skills can help you organize instructions and workflows; do not assume they behave identically across tools.
My ‘trust but verify’ eDiscovery mantra is a survival skill in this AI age. I am not saying to become a coder. Rather that we cannot abdicate our judgment to an AI driven discovery process without understanding its basic workflow, scope, limitations and associated costs. If you are going to add an AI line item to an invoice you should be able to explain it to the client and demonstrate how it added value to your services.
No one really knows how AI will reshape legal services. We all must be flexible and evolve our services if we want to thrive in this new age. Go back to legal function basics of dispute resolution, risk management and business process enablement to stay relevant. Make AI work for you. Don’t be a Reverse Centaur. Protect your technology toolkit from an AI bubble pop by being an educated consumer of products and licenses that make economic sense.
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.