Lawyer Lin Shanglun's article: When words become an assembly line: AI's industrial revolution in the legal industry

Lawyer Lin Shanglun believes that AI has fully cracked text generation, and the legal industry is experiencing a "Ford assembly line moment." Top American law firms have already integrated AI into their litigation file organization processes. Tasks that previously took hundreds of hours for ten lawyers can now be completed by a senior lawyer with AI in two to three hours for an initial draft. In the short term, junior lawyers face the greatest impact, but senior lawyers are about to enter a golden age.
(Background summary: "Silicon Valley Light Talk" Kenji announced his resignation from Phantom Wallet! Planning to rest for at least 5-10 years, feeling numb when seeing salary deposits)
(Additional background: Over 1,500 Meta employees signed a petition in anger! Demanding to narrow the scope of "AI monitoring keyboard and mouse," allowing half an hour of pause daily)

Table of Contents

Toggle

• The "Ford assembly line moment" in text work
• What top U.S. law firms have already experienced
• Not "replacement," but "amplification"
• Long-term impacts of AI from a Ford assembly line perspective
• Polarization of impact: senior vs. junior lawyers
• The golden age of senior lawyers
• The real situation for junior lawyers
• Will law firms reduce staffing needs? My answer: "Now"
• The "dragging behind" effect of systems and policies
• The most core and hardest-to-replace value of lawyers at this stage
• How should lawyers respond to the AI era?
• Advice for young law students
• The arrival of the Ford moment in the legal profession

Over the past two years, AI has disrupted the tech industry, but what it is truly beginning to reshape are those "text-centric" professional service fields. Among these, law is one of the most representative industries, and also one of the most worthy of in-depth observation.

If I had to summarize the biggest change AI has brought to the legal industry in one sentence, I would say: This is an industrial revolution exclusive to legal services. In my view, AI has completely cracked "text generation." Deep understanding of text, reading text, writing text—these have never been able to be mass-produced like they are now. As text generation enters the era of assembly lines, any process in a lawyer’s workflow that is text-based is being thoroughly redefined.

The "Ford assembly line moment" in text work

When you start thinking—which parts of a lawyer’s work can be "mass-produced"—you must consider what impact this kind of mass production will have on the entire industry. The most immediate effects are: reading and extracting complex data, organizing tedious tables, dissecting judgment reasons, causal reasoning, capturing facts. These are tasks that previously might have required years of experience for a lawyer to write, but now, like an assembly line, they can be accurately and massively duplicated.

This means: in legal workflows, as long as the part is related to text and centered on text—whether it’s reading files or drafting documents—the hours that used to be spent can now be greatly compressed.

What top U.S. law firms have already experienced

Honestly, Taiwan’s legal work has not yet been significantly impacted by AI. Taiwan’s legal industry is still very primitive; even courts and many practitioners still resist AI entering legal work. But I want to share from the actual situation in the U.S. to show how terrifyingly AI has already infiltrated lawyer’s work—not just infiltrated, but has actually replaced many tasks. I believe this is also what will happen in Taiwan in the future.

In the U.S., the most top-tier law firms have already integrated AI into their workflows. For example, suppose there is a patent litigation case involving a large number of patent descriptions, or two companies suing each other for patent infringement. These cases involve massive files: "intricate patent descriptions, expert reports, evidence of infringement." These used to require lawyers to read through each document.

Now, the approach is: they let AI read everything first. After AI has processed all the data, what happens? Every key point is marked by AI. For example, if you ask "When did the infringement occur in this case," AI will directly tell you: what specific actions the accused party took, which emails revealed relevant content, all presented at once.

These initial drafts are then reviewed by lawyers. Every question and follow-up from the lawyer results in a new version of the整理稿—reorganizing the raw data according to the lawyer’s needs. Then, the lawyer further writes the legal claims (e.g., "The other party infringed") based on these organized contents (infringement timing, proof, related evidence), ultimately generating the legal pleadings directly.

What was the process like before? Just finding the data might take dozens of hours; after gathering data, drafting facts and capturing facts also took dozens of hours; after that, searching for relevant precedents and supporting opinions could take dozens more hours. Now, with AI, the entire process can be completed in two to three hours for a draft.

And it’s important to emphasize: these AI tools are used by the top American law firms. Because AI is actually very expensive, only the most elite firms can afford the best tools, and their outputs are recognized by top lawyers. They are already doing this.

Not "replacement," but "amplification"

If you ask which tasks AI has replaced in lawyers’ work, I think framing it as "replacement" isn’t accurate. Instead, you should observe how AI enhances a lawyer’s capabilities.

A large patent case that once required ten lawyers and countless hours can now be done by one lawyer in a tenth or even less of the time. But do you know who is replacing whom? Actually, it’s not a simple replacement. Without a senior lawyer asking the right questions, integrating the data correctly, analyzing the judgment’s intent, and finalizing the content, this document simply cannot be produced. Without lawyers, this product cannot be made.

But conversely, when AI amplifies and accelerates a lawyer’s work, how many job opportunities are squeezed out? That’s the core issue. The point isn’t "which tasks are replaced," but that AI, by thoroughly cracking text understanding and generation, allows the work that used to take a long time to leap forward in efficiency.

Long-term impacts of AI from a Ford assembly line perspective

The most immediate market impact might be—lawyers temporarily need fewer people. But will this be the norm? I don’t think so.

Every major industrial revolution, with more than tenfold or even hundredfold increases in productivity, often does not lead to large-scale unemployment in the long run. What we are experiencing now can be called a "legal text worker’s industrial revolution."

Has such a revolution happened before? Of course. After the Ford assembly line, the speed of car production increased hundreds or thousands of times, with output soaring. Did this cause mass unemployment? Not really. Instead, it created more demand: more people needed to drive cars. As car prices dropped, everyone could afford a car and a smartphone, and this explosion of demand propelled human civilization forward.

This surge in productivity, rather than causing unemployment, actually increases new demand and creates more jobs. Previously, there might have been only a few car dealerships per day; now, with thousands or tens of thousands of cars produced daily, how many people need cars? How many dealerships? The production of cars also spurred related services—screens, satellite navigation, autonomous driving, in-car computers... all new jobs that were created.

So, in my view, in the short term, legal work will definitely be impacted by AI, and the market may be temporarily saturated. But once everyone is familiar with AI tools, and a lawyer’s case load can increase from 1x to 100x, the entire legal environment will change dramatically. Lawyer fees will decrease, more people will enjoy quality legal services, just like the widespread adoption of cars, leading to more related industries.

Senior vs. Junior Lawyers: Polarized Impact

So, which is more affected by AI—senior or junior lawyers? Honestly, both are hugely impacted. But I believe that, at this early stage of AI, it’s a very, very good time for senior lawyers, while the impact on junior lawyers is enormous.

Why? Returning to the workflow of top U.S. law firms, how important are the skills that support lawyers? Extremely important. Fact filtering, content adjustment, reasoning logic, correct question awareness—these all depend on years of experience. Even if AI can handle some tasks, senior lawyers can still intervene to improve the AI’s initial drafts.

Therefore, for senior lawyers, AI is a powerful amplifier; but for junior lawyers, the quality of AI-produced work will definitely be inferior to that of senior lawyers, and the speed may be slower, with more errors.

In the past, junior lawyers gained experience through practice, and firms were willing to invest in training. But now, if the entire process can be handled by a senior lawyer working with AI, do law firms still need to hire so many juniors? Tasks that once required ten juniors can now be easily done by one senior lawyer with AI. This will indeed create a gap in the training pipeline for junior lawyers.

The golden age of senior lawyers

Currently, less than 1% of the world’s population uses AI. Among the 99% who don’t, a senior lawyer proficient in AI can unleash astonishing productivity.

Industry data shows that a lawyer working independently can generate about 4 million in revenue annually. To surpass this, firms might have to hire one or two more lawyers, with each earning around 800,000 to 1 million per year.

But what about now? When AI tools are available, do senior lawyers prefer to hire more lawyers or to combine AI tools? I believe the latter. So, senior lawyers who are already using AI tend to hire more assistants or secretaries instead of more lawyers.

In the past, one lawyer could easily manage a secretary, because the workload was manageable. But now, if a lawyer uses AI to increase productivity tenfold, they might need two secretaries to handle related tasks. Secretaries earn around 400,000 to 500k, much cheaper than lawyers at 1 million, and they can handle a lot of the procedural work for senior lawyers. Is this highly productive? Absolutely, much more than before.

The real situation for junior lawyers

But in practice, I don’t think junior lawyers will find it so hard to find a way out. The labor market is still very tight; many law firms are hiring, and junior lawyers can still find internships. Lawyers of my age or older are still looking for young lawyers.

However, if AI becomes truly widespread in the legal industry, it will indeed prevent junior lawyers from accumulating skills through practical experience. But similar things have already happened in the U.S. Before, summer associates at big firms would attend meetings, do some chores, and get a preliminary look at data; now, big firms assign very clear tasks to summer associates.

In the past, lawyers’ work was not industrially integrated; junior lawyers might only handle 1% of what senior lawyers do, with the difference mainly in volume, not quality. But now, senior lawyers love AI-generated content that is fast and high-quality. In AI engineering, "data processing" is crucial.

Today, what you feed AI to read is exactly what many summer associates do in big firms: "organizing and classifying data." To clarify: they are not training large language models; they are preparing data in formats that AI can read optimally.

For example, cleaning up low-resolution data, consolidating similar data, or conversing with clients to interpret handwritten, messy drafts—these are new skills in the era where AI can replace many manual tasks. How to handle raw data has become a new key skill, and this is currently happening inside top U.S. law firms.

Will law firms reduce staffing needs? My answer: "Now"

If you ask whether law firms will reduce staffing due to AI in the future, honestly, I have no idea. Because right now, the "AI systems" we see are the weakest, worst versions we will have in the next five years. I cannot imagine how AI will evolve.

So I have to reframe the question: "Will current AI development cause law firms to reduce staffing? What are the most core, hardest-to-replace abilities of lawyers today?" I can only answer "now," not the future.

Based on current conditions, AI’s strength lies in batch text production and reading. So I believe law firms will reduce staffing. Currently, there’s less need for hired lawyers. This is also common in the legal industry: if AI tools are useful, firms tend to hire fewer associates.

But I think the reduction mainly affects junior lawyers’ staffing needs. Law firms will develop in a healthier direction—requiring more MIS engineers, possibly AI engineers, and more secretaries. These are all possible and are indeed happening.

The "dragging behind" effect of systems and policies

But if you ask whether, in the future, there will be a "one-stop AI lawyer"—AI capable of doing everything—I think it depends more on whether regulations and government policies keep pace with technological development. If policies keep up, the future will be very different.

Currently in Taiwan, lawyers must physically appear in court; you cannot just submit an AI-made video. For court filings, can you directly submit digital files created by AI? No, you still need to deliver paper documents. So, the system has not changed much, and I don’t think AI will make the legal industry completely disappear or unprofitable anytime soon.

AI technology is ahead, but policies and supporting measures are not necessarily keeping pace.

The most core and difficult-to-replace value of lawyers at this stage

So, what is the most core and hardest-to-replace value of lawyers at this stage?

Honestly, the experience accumulated from past cases is still very important. Currently, every step of AI’s work still requires human intervention, organization, and questioning—these are all based on experience. The differences in text generation by humans will gradually diminish as AI develops; the variability in legal writing will become smaller and smaller, I believe.

What will become the most valuable trait for lawyers? Clearly, it will return to the era of "sales and business development."

Today, what matters most is: does the lawyer possess strong persuasive power and human warmth? Can they communicate and consult with clients in a way that provides endless warmth and confidence, making clients more willing to trust you?

This value is not only relevant in the "future" but also in the "present" and even the "past." Top international law firms’ senior partners all have a team of very strong associates—deep legal knowledge and excellent drafting skills. But why do clients still need to see the most senior person? Because they bring stability and persuasive power.

A more intuitive example: if you hire a top senior lawyer with ten years of experience in a firm, and compare their ability to draft pleadings or analyze issues with the managing partner (the biggest boss) of that firm, who would win? Honestly, not necessarily the managing partner. But why can that senior lawyer reach the top? Because they have the most top-tier, most persuasive, most skilled business acumen. Of course, they also have strategic strengths, but more importantly, they have accumulated "client sensitivity"—knowing how to reassure and persuade clients.

This is what Jensen Huang calls "human warmth, human empathy." These are aspects AI cannot replicate—yet.

How should lawyers respond to the AI era?

My advice for facing the AI era is: you must first understand AI’s characteristics.

Many colleagues react to AI with: "I’ve seen my clients’ GPT outputs are terrible, awful, how can I use this? It’s just a scam!" Indeed, many clients use GPT or Gemini randomly, and the results are poor. But why? You need to understand why.

Why do you think GPT is so bad? But why are some AI tools widely used by top law firms? What’s the difference? From the past GPT single-core AI to later agentic collaboration architectures—where is AI suitable for human use? Why do consumer-grade GPTs handle legal issues so poorly? These are questions we must know.

Let me give a simple example. Why can’t ordinary people use consumer GPT to answer legal questions? Or why does GPT itself say, "We do not recommend using our model for legal issues"? The straightforward reason is: consumer GPT tools have input token limits. You can input a lot of data, but the amount they can read is limited—probably less than 10k tokens.

For ordinary users, "I won’t ask more than 10,000 words," so they think GPT is great. But for professionals, the content discussed with clients might be four or five hours’ worth, totaling 50,000 words; a case analysis might be 70,000 or 80,000 words. Can you analyze that by inputting it all? No way, because it doesn’t have enough resources to read your data. The output will be poor.

However, professional-grade AI services designed specifically for lawyers will optimize this—allowing input tokens of 100k or even 1 million, or combining multiple models to handle larger volumes of text.

The key point: don’t judge AI’s potential based on consumer GPT’s poor results. If you dismiss AI this way, you’re already falling far behind. You should ask: why does this happen? What are the underlying AI technology principles?

And now, acquiring this knowledge is very cheap—these questions can be fed into GPT, Gemini, Claude, and they will give you very good answers. When you can access AI-related information so quickly, why not learn? It would be a huge missed opportunity not to.

My advice: stay sensitive to AI developments and do research. Understand each AI model’s characteristics, why breakthroughs are happening in document reading, how agentic architectures work—these are essential knowledge.

Advice for young law students

What about young law students? What advice would I give?

In Taiwan, the national exam is still very difficult. If you truly want to be a lawyer, you must pass the exam. As I mentioned earlier, young lawyers need solid legal fundamentals, good question awareness, and case sensitivity—these all depend on personal experience accumulation. Only with strong personal ability can you better harness AI.

Let me explain the current AI capability logic: if a very skilled lawyer also has AI knowledge, their output will be millions of times better than that of a law novice or a young law student with AI knowledge but no legal expertise. Because AI use fundamentally amplifies and accelerates your existing abilities and results. If your baseline ability is zero, AI amplifying it tenfold is still zero; if your baseline is 100, then AI amplifying it tenfold becomes 1,000.

So, for law students, we still need to focus on passing the national exam. The exam itself does not allow AI use. I believe law students should prepare thoroughly for the exam first, then start thinking about AI and related issues afterward.

For young lawyers already in practice, it’s essential to grasp some basic AI concepts rather than resist. When you see poor GPT outputs and immediately think "AI can’t do legal work," that kind of intuitive thinking is already a big step behind. You should ask: why does this happen? Learn more about AI technology. Everything that exists has a reason; every phenomenon has its logic.

The Ford moment in the legal industry has arrived

Every technological revolution redraws industry boundaries. The railway age gave rise to the modern bond market, taking nearly half a century; the internet revolution reassembled global commerce over nearly 20 years. AI is now reshaping the "text-centric" professional service industry at a similar scale—but this time, the timeline seems to be only three to five years.

The legal industry is experiencing its own "Ford assembly line moment." In the short term, junior lawyers will face the most direct impact, and law firm staffing structures will be reshuffled; in the medium term, senior lawyers will enter their best-ever golden age because AI is an amplifier, and 99% of people haven’t even boarded yet; in the long run, the core value of lawyers will return to "human warmth, persuasion, and trust"—parts that AI cannot replicate.

And the "most powerful AI" we see now is actually the weakest version in the next five years. So instead of arguing whether AI will replace lawyers, it’s better to quickly learn how to become a lawyer who masters AI.