The application scenarios of AI in Taiwan have not yet connected to the path of “efficiency multiplication” through professional self-awareness. The author of this article is Lawyer Lin Shanglun, founder of M-Ross.
(Previous context: Lawyer Lin Shanglun’s article: The “Vibe Coding” craze brought by Gemini 3.0 is a misunderstood celebration)
(Additional background: Lawyer Lin Shanglun’s article: Universal Basic Income (UBI) and Blockchain—Are they social safety nets under the AI wave?)

Table of Contents

• A Hardcore AI Practical Course
• When the “Professional Fortress” is Breached by AI
• Common Misconceptions of AI Application in Taiwan: Treating Tools as Toys
• The Harsh Reality in Silicon Valley: Without AI Collaboration, You Don’t Even Qualify to Submit
• The Silent Reorganization of Professions

A Hardcore AI Practical Course

I was fortunate to be invited to Huainan Bank to give a three-hour, hardcore practical course on AI application and governance for a group of finance professionals. We started with the draft of the AI Basic Law, rigorously defining what AI is to exclude those automation programs misunderstood by the public; we discussed privacy and why many feel overwhelmed when using GPT or Gemini.

I also explained why some customer service bots are often “weak” to the point of being laughable. It’s not that AI is incapable, but because they choose low-end models, lack RAG (Retrieval-Augmented Generation) to supplement databases, or even fail to differentiate agents for different application scenarios.

When the “Professional Fortress” is Breached by AI

But the climax of the entire course happened when I switched back to my “lawyer” identity and demonstrated live AI operations.

I directly showed how to use AI to draft pleadings, prepare notarization letters, modify contracts, and perform other lawyer tasks.

The professionals in the audience from the finance sector looked from polite focus to extreme shock. Why shock? Because in the past, these were highly professional tasks that only intern lawyers or employed lawyers could complete with hands-on guidance.

But when they saw firsthand that these processes, regarded as “professional fortresses,” could be so smoothly automated by AI, the impact was enormous.

Common Misconceptions of AI Application in Taiwan: Treating Tools as Toys

Taiwan’s hardware technology always stands at the forefront of the world, but our software thinking and applications often lag five to ten years behind the international mainstream. When you open Facebook, what AI information does the algorithm push to you?

“Gemini made my photos look so beautiful!”
“Used Vibe Coding to write a data classification robot!”
“Built a customer service chatroom that responds more like a real person!”

Sounds impressive, right? But honestly, these were already old news when Cursor was released two years ago. If you are still training AI on “how to reply more naturally” or “not to reveal it’s a robot,” then it’s really a pity, because that is not the current AI development trend.

The most powerful strength of AI is not in making amateurs produce small tools, but in enabling domain experts with specialized knowledge to unleash tenfold or even dozens of times more energy.

In this financial speech, I did not demonstrate how AI completes financial industry tasks because I lack domain knowledge in finance or accounting, so I am not qualified to showcase AI applications in that field.

But what I want to emphasize is: “When a professional lawyer, accountant, or doctor knows how to combine their expertise with AI, transforming tedious, hard-to-scale tasks into efficient automated processes, that is the real blow to traditional methods.”

The Harsh Reality in Silicon Valley: Without AI Collaboration, You Don’t Even Qualify to Submit

To make everyone understand how huge this efficiency gap is, I brutally share a current situation in Silicon Valley. Many top tech companies there already have two shocking rules:

• First, all initial drafts of work must never be created from scratch by humans. No matter how complex the task, AI can generate at least 40-50% of the foundation. Its direction is usually correct, and employees are expected to iterate on this AI-generated draft rather than waste time reinventing the wheel.
• Second, when you process or iterate on the initial draft, backend software monitors in real-time how much AI assistance you are using. If you do not meet that ratio, your work cannot be sent for review or moved to the next stage.

The logic behind this is simple: bosses value cost and efficiency. In coding or document processing, if you do not achieve a certain proportion of AI-assisted work, you are not even qualified to submit your work. That’s the reality.

When the smartest minds worldwide have already mandated “human-AI collaboration” and implemented data-driven monitoring, if we still blindly believe in “human uniqueness” and that “professional work cannot be replaced,” then we are truly being arrogant.

The Silent Reorganization of Professions

I personally believe that what most people should do now is not learn how to code (unless you want to switch careers), but clearly clarify: “Where are my professional advantages?” and “Which parts of my workflow can be handed over to AI?”

If you don’t do this, future opportunities for respite will be scarce because under new workflows, the required manpower may be only one-tenth of before, and “you, I” could very well be excluded from future workplaces.