,

Five Generations under AI|The 90後: meet AI head-on, and shape the future with the machine

Introduction|The critical point for Taiwan’s 90後: no more “adapting slowly”
On the spectrum of Taiwan’s workplace, people born in 1990–1999 (the 90後) are being pushed onto a new critical line. They grew up in the wave of broadband, smartphones, and social platforms; from PTT, Wretch (無名), and Yahoo! Kimo to Facebook, YouTube, and short video, they are at home with the grammar of a platform and the ecology of content. They entered university in the years of expansion and multiple admission channels, and entered society just as they ran into the 2008 financial crisis, the sharing economy, and the rise of e-commerce platforms. Now generative AI and multi-agent systems are soaking quickly into the organization. They are not AI natives, and they also do not have the privilege of “adapting slowly” — because the reconstruction of the enterprise has already begun. It is less that the 90後 are being changed by AI, and more that they are the first main generation that has to shape the future with AI: on the path of advancement, the recasting of capability, and the identification of value, they are at once the people who push the transformation, and the people who bear the pressure.

NO° 01|The spiritual ground of the 90後: native to platform logic × conscious of individual value

The technical socialization of Taiwan’s 90後 began in internet forums and instant messaging, and was completed in e-commerce, social, and content platforms. For them, technology is not only a tool. It is also the context of socialization, and a container for self-expression: recommendation mechanisms, algorithmic distribution, and topic heat decide how attention is gathered and converted. This “nativeness to platform logic” lets the 90後 create momentum, amplify volume, and iterate quickly. It also easily confines them to a short-chain mindset — fluent at the application end, and relatively thin in their grasp of the underlying data structure, governance rules, and system permissions.

At the same time, the 90後’s individual values are highly self-aware. Faced with a generational ceiling and a culture of long hours, they are not strangers to the vocabulary of involution (內卷), lying flat (躺平), and “the hired hand” (打工人). They know an institution is not easily shaken, and so they put more weight on self-protection and on the right to choose. In the age of AI, this self-awareness turns into a posture of action: do not wait for approval; try first. In many teams, the first person to pull AI into a workflow is often a 90後 — because they understand that the cost of waiting for an instruction is missing the dividend.

Nativeness to a platform, however, also brings a structural shortcoming. When a task is upgraded from “content—distribution” to “goal—path—governance,” knowing only the grammar of a platform is no longer enough to carry change at the level of an organization. If the 90後 want the initiative in an AI transformation, they have to take their sharpness about platforms one step further, and turn it into a sharpness about system design and data responsibility.

NO° 02|AI’s double effect: freeing the individual × compressing the organization’s boundary

The most immediate positive effect of AI on the 90後 is the amplification of an individual multiplier. A marketing plan can, in a few hours, finish the insight, the script, the distribution plan, and several versions of the assets. An analyst can query a metric in natural language and produce a narrated conclusion. An operations partner can feed customer-service knowledge, user comments, and process rules into a RAG system, and automate replies and tickets. A product colleague can connect user research, path simulation, and A/B iteration into an agent pipeline. For the first time, the 90後 can steadily move from “executor” to “collaborative operator at the controls,” and build a personal tool stack, a prompt framework, and a multi-agent work net of their own, so that intelligence is amplified continuously, and does not only flash from time to time.

At the same time, AI is also compressing the organization’s boundary and the need for layers. Conversation is not the end; action is. When a task can be broken down semantically and dispatched automatically, the value of intermediary mid-level roles is diluted, and the flattening of layers becomes the normal state. For the 90後 this is an opportunity and a test: people who understand both AI and the business begin to be given the jobs of “translator, structural orchestrator, operator of the enterprise brain” — turning a need into a chain of tasks, turning the chain of tasks into permissions and governance rules, and turning the governance rules into a dashboard that can be observed. Whoever can persuade upstream and downstream in the language of architecture can grow quickly in the new order.

A fall in the density of the organization, though, also means that competition is more direct. AI enlarges the gap between individuals: people who can use it, dare to try, and can tune it pull far ahead of people who only “do as they are told”; people who can write a result into the system pull further ahead of people who only “use it once.” The double edge of AI shows most clearly on the 90後: the more you can drive yourself, the more you are freed; the more you depend on instruction, the more you are pushed to the margin.

NO° 03|Three risks in the shadow: shallow capability, weak structure, high anxiety

First, the boundary of skill blurs, and the anchor of value loosens. Being able to do a little copy, a little reporting, and a little operations used to be an advantage of the platform generation. In front of AI, these “shallow crafts” are copied quickly. Rather than being a jack of all trades, build a clear functional anchor: at which node of the value chain are you indispensable? Can your output be amplified by AI, and can it be traced?

Second, a lack of systems training makes it hard to move up. A short-chain way of thinking easily lets a person stop at “can cut, can edit, can apply.” But what is scarce in the age of AI is “can design.” When a task is upgraded from “output” to “the design of a path and the orchestration of governance,” if you lack the concepts of data responsibility, a permission boundary, and the handling of exceptions, it is hard to take a higher right to speak. Creativity and diligence without a systems base will, in the end, be worn down in endless rework.

Third, constant comparison brings high anxiety. Social feeds are full of posts that say “I used GPT to finish two weeks of work in three hours,” and it is easy to be pushed into a loop of “endless self-devaluation” — not fast enough, not cool enough, not intelligent enough. If the anxiety is not structured into a rhythm of learning and experiment, it will turn into aesthetic fatigue and a paralysis of action.

NO° 04|A path of breakthrough: from user of a tool to someone who understands the system

The first step of a breakthrough is to make the role precise. First anchor yourself at one or two key interfaces of the value chain: for example, “the person who turns a market question into a data query,” “the person who turns customer-service knowledge into an answer that can be retrieved,” or “the person who turns a product hypothesis into a trial that can be run.” Once the role is clear, you can build a model of cooperation of “what I can do × what AI can amplify” — your input defines the quality, AI’s output provides the throughput; your standard of judgment guards the boundary, AI’s automation guarantees the speed.

The second step is to move from using to designing. String everyday high-frequency tasks into a reusable workflow: use RAG to make a knowledge assistant that can truly be traced (not an all-purpose FAQ); use an agent to pack insight, creativity, distribution, and returning data into a closed loop; turn cross-department communication into an “AI collaboration dashboard,” so that decisions happen in the observable. You are not “a person who can use GPT.” You are “a person who can design a GPT workflow” — that is the moat.

The third step is to take part in the organization’s AI agenda, and raise your bargaining power. Propose a single-point pilot on your own: start from a scene with a clear metric and a controllable risk, first prove an improvement in the three indicators of output, quality, and risk, and then expand to an adjacent process. At the same time, push a “capability map” and “governance rules”: who needs which tools, how permissions are graded, how data is labeled, and how exceptions are escalated. When you turn change into language and into a rhythm, your standing moves from “replaceable execution” to “irreplaceable design.”

NO° 05|Closing: 90後, do not only use the tool — set the rules

This is not a gentle age. AI does not speak the language of favors. It speaks only speed, quality, and what can be observed. And it is precisely because of this that the 90後 have a critical set of genes: sensitivity to tools, the ability to express, platform logic, the habit of collaboration, and an affinity for technology. If the 80後 are still updating their system, and the 00後 are still accumulating a feel for the field, the 90後 happen to stand at the crossing of “understanding the structure × driving execution.” Your fate no longer depends on whether you can use AI. It depends on whether you can turn AI into a system you have designed: building a new logic of work, becoming a bridge between people and agents, and leading a recasting of capability — and a leap in the organization — that belongs to you. AI will not define the future for you. But you can entirely redesign your own place.

Topic | Five Generations under AI (AI洪流下的五代人)

The judgment, restated

My judgment is that the 90後’s advantage is not being better at using a tool, but connecting a personal tool stack into a system the organization can orchestrate.

Three things this week

  • Inventory your personal tool stack: which tools serve only you, and which can become a shared module for the team.
  • Externalize one high-frequency flow (input / steps / acceptance), so that an agent can take it over.
  • Find one cross-generational collaborator, and practice the translation of “platform language ↔ organization language.”

The series overview is in Five generations in concert. You are welcome to subscribe to Future Insights Lab, or to reply and discuss how you are making the connection.

訂閱 Future Insights Lab

不定期寄送 AI × Business 短篇筆記,協助你掌握實務重點。隨時可取消訂閱。