As a lawyer, I should acknowledge my bias against my own obsolescence. That said, I find some of the more extreme predictions premature at best, and more likely fanciful, at least for the foreseeable future.
As you quite rightly point out, every previous advance in technology has created more legal work, not less. For every gain in efficiency, client and regulatory expectations rise accordingly, and the net is neutral or even costlier than the prior situation. But there’s more to it than that.
A lot of what contemporary AI tools can do well are tasks that were already largely automated. Document review tools have used intelligent algorithms to cull results for more than a decade. Generating contracts sounds impressive, but template contracts that only require the lawyer to add specifics and tweak key provisions have existed since before even the advent of personal computers. An AI-generated contract that requires only minor revisions is at best an incremental improvement.
None of this is to say AI doesn’t have many applications for lawyers - it obviously does - or that it won’t eliminate some categories of legal work. I wouldn’t be at all surprised to see things like the drafting of wills or non-contested divorces be entirely automated in the next 5 years. But that is already commoditized work that is largely driven by existing technology. At the end of the day, the value provided by a $2300/hour law firm partner comes from doing things that are almost entirely judgmental and relationship-driven, based upon their years of experience. It is hard for me to envision even a much more sophisticated AI than what we presently have being able to replicate that. (To that point, I would add that I have used all the leading AI legal tools, and while they have their applications, they all currently suffer from flaws that make them reliable only for simple and preliminary tasks.)
Finally, I would be remiss if I didn’t point out that the bar is insular and protectionist. If we start to
see *any* technology force large numbers of lawyers out of work, I wouldn’t be surprised to see the adoption of new rules that create significant disincentives to relying on it. For example, a rule could be adopted that the attorney client privilege is waived just by virtue of using an AI assistant to process confidential information (frankly, there is a decent argument for this reading under even existing rules).
AI, like word processors (which eliminated a great deal of legal secretarial work), the internet (which turbocharged legal research), and big data tools (which made massive document reviews possible), will have a significant impact on the legal profession. My guess, however, is that the impact will be evolutionary, rather than revolutionary.
AI is already making some legal services cheaper for some consumers. A friend reports that he needed to consult with attorneys on two separate occasions pertaining to similar types of matters. The first time, he burned a fair number of legal hours discussing his situation with the attorney before he was about to obtain actionable outputs. The second time, he discussed his situation with an AI Chatbot and had it prepare a legal document. Then he had an attorney review the document. The attorney confirmed it was fine and the review took a fraction of the time (and cost a fraction of the fees) than it would have as a full consultation.
I think this is a great use of AI (and a good opportunity to clarify that we don’t mean AI will never make legal services cheaper). For example, with simpler tasks without adversarial dynamics like drafting a will, we think AI will likely bring down costs. This does also illustrate the potential negative consequences of the first bottleneck. Your friend can currently access AI chatbots with impressive legal reasoning abilities. But if bar associations or private individuals sued OpenAI or Anthropic for Unauthorized Practice of Law (and they might win since AI chatbots providing specific legal recommendations starts to look like practicing law), it may be a lot harder for your friend to use AI in this way. This is essentially what happened with LegalZoom, which tried to leverage the benefits of software to bring down the costs of legal services. It was mired in UPL lawsuits, forced to pay large settlement fees, and had to change its business model multiple times. We don’t want to see the samething happen for AI as well.
I'm having trouble understanding the point of this article. Is there someone out there saying that AI is going to transform the legal industry independent of its adoption by lawyers? And don't all these proposed reforms pre-date ai's introduction to lawyers (~2023)? Also, I wonder if you are misunderstanding the reasons why legal services are so expensive. What about how impoverished legal education is? Looking back on it, a curriculum full of survey courses did very little to prepare me for the realities of litigation practice. It was more about signaling than training future lawyers. Do law firms fill this gap? Probably not. Mine did not. Clients of large and mid-size firms subsidize the education and training of junior lawyers. That's part of the expense (and it's expensive). What about the lack of systemization when it comes to doing the actual legal work? As far as I can tell, most lawyers practice intuitively and transfer knowledge to junior lawyers informally. Not to mention what Casey Flaherty taught us 10 years ago when he was trying to get lawyers to improve their efficiency by learning how to use their software (e.g., Excel, Microsoft Word, Adobe Acrobat). What about the lack of knowledge management? I'll bet clients pay their law firms millions to research and write briefs and memorandums that have all been done before by other attorneys at the same firm. All that is also very expensive for clients.
I believe one of your conclusions is questionable. You suggest that consumers won't benefit because those with more money will also benefit from AI. This will merely escalate the arms race, cancelling out any benefits to consumers.
This might be true if both sides were beginning from an equal starting point. This is not the case. Right now, most consumers, especially those of modest means, have zero legal representation. If AI provides any benefit, it will tend to reduce the gap.
This question is slightly different to the one that most people focus on: what will happen to the employment of lawyers, and the career entry points?
This is a convincing case that the labour cost of law will rise sharply. It will become so labor-intensive to do anything law-related that only AI is cheap enough. Except of course, for the figurehead human lawyer: a figleaf over a lot of burned electricity. There is no career pathway here, and the same figleaf can cover many, many cases, just as one building in Delaware can be the registered office of thousands of corporations.
10 Feb 2026 : Judge Jed Rakoff ruled that 31 documents a defendant generated using an AI tool and later shared with his defense attorneys are not protected by attorney-client privilege or work product doctrine.
The logic is simple: an AI tool is not an attorney. It has no law license, owes no duty of loyalty, and its terms of service explicitly disclaim any attorney-client relationship. Sharing case details with an AI platform is legally no different from talking through your legal situation with a friend (which is not privileged).
You can't fix it after the fact, either. Sending unprivileged documents to your lawyer doesn't retroactively make them privileged. That's been settled law for years. It just hadn't been tested with AI until now.
As a lawyer, I should acknowledge my bias against my own obsolescence. That said, I find some of the more extreme predictions premature at best, and more likely fanciful, at least for the foreseeable future.
As you quite rightly point out, every previous advance in technology has created more legal work, not less. For every gain in efficiency, client and regulatory expectations rise accordingly, and the net is neutral or even costlier than the prior situation. But there’s more to it than that.
A lot of what contemporary AI tools can do well are tasks that were already largely automated. Document review tools have used intelligent algorithms to cull results for more than a decade. Generating contracts sounds impressive, but template contracts that only require the lawyer to add specifics and tweak key provisions have existed since before even the advent of personal computers. An AI-generated contract that requires only minor revisions is at best an incremental improvement.
None of this is to say AI doesn’t have many applications for lawyers - it obviously does - or that it won’t eliminate some categories of legal work. I wouldn’t be at all surprised to see things like the drafting of wills or non-contested divorces be entirely automated in the next 5 years. But that is already commoditized work that is largely driven by existing technology. At the end of the day, the value provided by a $2300/hour law firm partner comes from doing things that are almost entirely judgmental and relationship-driven, based upon their years of experience. It is hard for me to envision even a much more sophisticated AI than what we presently have being able to replicate that. (To that point, I would add that I have used all the leading AI legal tools, and while they have their applications, they all currently suffer from flaws that make them reliable only for simple and preliminary tasks.)
Finally, I would be remiss if I didn’t point out that the bar is insular and protectionist. If we start to
see *any* technology force large numbers of lawyers out of work, I wouldn’t be surprised to see the adoption of new rules that create significant disincentives to relying on it. For example, a rule could be adopted that the attorney client privilege is waived just by virtue of using an AI assistant to process confidential information (frankly, there is a decent argument for this reading under even existing rules).
AI, like word processors (which eliminated a great deal of legal secretarial work), the internet (which turbocharged legal research), and big data tools (which made massive document reviews possible), will have a significant impact on the legal profession. My guess, however, is that the impact will be evolutionary, rather than revolutionary.
And right on cue!
https://mashable.com/article/ai-attorney-client-privilege-court-evidence
AI is already making some legal services cheaper for some consumers. A friend reports that he needed to consult with attorneys on two separate occasions pertaining to similar types of matters. The first time, he burned a fair number of legal hours discussing his situation with the attorney before he was about to obtain actionable outputs. The second time, he discussed his situation with an AI Chatbot and had it prepare a legal document. Then he had an attorney review the document. The attorney confirmed it was fine and the review took a fraction of the time (and cost a fraction of the fees) than it would have as a full consultation.
I think this is a great use of AI (and a good opportunity to clarify that we don’t mean AI will never make legal services cheaper). For example, with simpler tasks without adversarial dynamics like drafting a will, we think AI will likely bring down costs. This does also illustrate the potential negative consequences of the first bottleneck. Your friend can currently access AI chatbots with impressive legal reasoning abilities. But if bar associations or private individuals sued OpenAI or Anthropic for Unauthorized Practice of Law (and they might win since AI chatbots providing specific legal recommendations starts to look like practicing law), it may be a lot harder for your friend to use AI in this way. This is essentially what happened with LegalZoom, which tried to leverage the benefits of software to bring down the costs of legal services. It was mired in UPL lawsuits, forced to pay large settlement fees, and had to change its business model multiple times. We don’t want to see the samething happen for AI as well.
I'm having trouble understanding the point of this article. Is there someone out there saying that AI is going to transform the legal industry independent of its adoption by lawyers? And don't all these proposed reforms pre-date ai's introduction to lawyers (~2023)? Also, I wonder if you are misunderstanding the reasons why legal services are so expensive. What about how impoverished legal education is? Looking back on it, a curriculum full of survey courses did very little to prepare me for the realities of litigation practice. It was more about signaling than training future lawyers. Do law firms fill this gap? Probably not. Mine did not. Clients of large and mid-size firms subsidize the education and training of junior lawyers. That's part of the expense (and it's expensive). What about the lack of systemization when it comes to doing the actual legal work? As far as I can tell, most lawyers practice intuitively and transfer knowledge to junior lawyers informally. Not to mention what Casey Flaherty taught us 10 years ago when he was trying to get lawyers to improve their efficiency by learning how to use their software (e.g., Excel, Microsoft Word, Adobe Acrobat). What about the lack of knowledge management? I'll bet clients pay their law firms millions to research and write briefs and memorandums that have all been done before by other attorneys at the same firm. All that is also very expensive for clients.
Thanks for this excellent paper.
I believe one of your conclusions is questionable. You suggest that consumers won't benefit because those with more money will also benefit from AI. This will merely escalate the arms race, cancelling out any benefits to consumers.
This might be true if both sides were beginning from an equal starting point. This is not the case. Right now, most consumers, especially those of modest means, have zero legal representation. If AI provides any benefit, it will tend to reduce the gap.
P.S. Always glad to see analysis from Narayan & Kapoor, which is why I gave their book AI Snake Oil the highest recommendation. https://www.llrx.com/2025/05/recognizing-and-dealing-with-ai-snake-oil/
I'll be adding Mr. Curl's name to my must-follow list.
This question is slightly different to the one that most people focus on: what will happen to the employment of lawyers, and the career entry points?
This is a convincing case that the labour cost of law will rise sharply. It will become so labor-intensive to do anything law-related that only AI is cheap enough. Except of course, for the figurehead human lawyer: a figleaf over a lot of burned electricity. There is no career pathway here, and the same figleaf can cover many, many cases, just as one building in Delaware can be the registered office of thousands of corporations.
10 Feb 2026 : Judge Jed Rakoff ruled that 31 documents a defendant generated using an AI tool and later shared with his defense attorneys are not protected by attorney-client privilege or work product doctrine.
The logic is simple: an AI tool is not an attorney. It has no law license, owes no duty of loyalty, and its terms of service explicitly disclaim any attorney-client relationship. Sharing case details with an AI platform is legally no different from talking through your legal situation with a friend (which is not privileged).
You can't fix it after the fact, either. Sending unprivileged documents to your lawyer doesn't retroactively make them privileged. That's been settled law for years. It just hadn't been tested with AI until now.
https://storage.courtlistener.com/recap/gov.uscourts.nysd.652138/gov.uscourts.nysd.652138.22.0.pdf
https://storage.courtlistener.com/recap/gov.uscourts.nysd.652138/gov.uscourts.nysd.652138.22.0.pdf