A Wall Street Journal article on the AI-driven always-on economy made me stop and think for a long time. Not because the writing was exceptional, but because it showed me something: we have been using the wrong frame to understand this change.
The article’s premise goes something like this: AI allows the economy to run continuously, 24 hours a day. A brand new era.
But is it really new?
The Always-On Economy Is Nothing New
Convenience stores in Taipei have been open around the clock for twenty years. Global financial markets hand off across time zones — Tokyo to London to New York — and the sun never sets on trading floors. Semiconductor fabs run 365 days a year without stopping, because the cost of restarting a furnace exceeds the cost of keeping it running.
We have been living in an always-on economy for a long time.
The difference is that the old always-on ran on human cost.
I remember the early days of building my company. We were running a cross-timezone project, handling the Asia side during Taipei business hours and fielding American clients at night. It sounded international. In practice, it meant a few of us taking turns staying up late, phones permanently off silent mode. One time I was pulled out of bed at 3 a.m. to handle a system alert. Two hours of investigation later, it was a false alarm. But you cannot ignore it — because what if it is real?
The price of those days did not show up as overtime pay. It showed up in eyes that grew more tired with each passing week, in creative meetings that went quieter, in people conserving energy and handling only whatever was most urgent right in front of them.
That is the truth of the old always-on economy: using human physiological limits to prop up a system that was never designed for human physiology.
AI Restructures the Cost Logic of Always-On
After AI entered the picture, always-on operations themselves did not change. What changed completely was the cost structure of sustaining them.
A concrete example. My team now uses AI agents to handle first-line triage on client requests and system monitoring. That 3 a.m. false alarm used to require someone to crawl out of bed and verify it. Now an AI assesses severity first, and a human is only notified when the situation genuinely requires intervention. The result: response to real emergencies is actually faster, because the on-call person’s energy is intact, not already drained by five false alarms.
This is a qualitative shift, not a story about 30 percent efficiency gains.
When you push the marginal cost of sustaining always-on operations toward zero, things that were previously impractical become entirely reasonable. Real-time monitoring across every supply chain node? That used to require an entire department. Now a single AI agent handles it. Instant risk assessment on every transaction? Once only large financial institutions could afford it. Now it is within reach for small and mid-sized businesses.
There is a trap here, though. Many people assume AI simply means machines doing what humans used to do. That is the most superficial reading — and the most dangerous strategy.
Three-Layer Restructuring: Not an Upgrade, a Rebuild
Watching organizations adopt AI across different industries, I have noticed a pattern. The ones that succeed are not those trying to replace human labor with AI. They are the ones willing to rethink the entire operating logic.
That rethinking happens on three levels.
The first is process restructuring. Traditional business processes were designed around human working hours. Monday meeting, Wednesday report, Friday review. In a human-driven environment, that rhythm made sense — people need time to absorb information, form judgments, coordinate action. But when AI can process and analyze in real time, that rhythm becomes an artificial bottleneck. Real process restructuring is not changing a weekly meeting to daily. It is asking: do we still need fixed-frequency meetings at all? Or can we move to event-driven coordination — convene when something happens, keep moving when it does not?
The second is restructuring the human-machine division of labor. This layer is the most commonly misunderstood. Many organizations draw up a static chart: AI handles A, B, C; humans handle X, Y, Z; here is the handoff point. Static charts do not work, because the boundary of what AI can do shifts every three months. My own approach is what I call a dynamic authorization framework: AI has a baseline of autonomous authority, but decisions above a certain complexity or risk threshold automatically escalate to a human. That threshold is not fixed. It adjusts continuously based on AI performance and the team’s accumulated trust. It is like managing a new hire: at first you review everything; six months later you watch only the critical calls; a year later you review the outcomes.
The third is value restructuring. This is the deepest layer and the one least often discussed. As AI pushes execution efficiency to its ceiling, competition between organizations shifts toward what AI cannot easily replicate: understanding of cultural context, the capacity to build genuine trust with people, the courage to make responsible decisions in ambiguous territory. I touched on a related point in Post-Code Era Thinking: When Taste Becomes Humanity’s Critical Competitive Advantage: when execution costs approach zero, judgment becomes the only differentiator. At the organizational level, that judgment is company culture, decision quality, and resilience under uncertainty.
From Power Outlet to Force Field: The MCP Metaphor
One technical development deserves particular attention: MCP, or Model Context Protocol.
MCP is a standard protocol allowing different AI models and tools to communicate with each other. That sounds technical. Its significance runs much deeper.
Before MCP, every AI tool was an island. Your customer service AI had no idea what the inventory AI was doing. Your analytics AI could not see the marketing AI’s data. Getting them to work together required custom integration code for every connection — expensive to build, expensive to maintain.
What MCP does is roughly what the USB standard did for the computer industry. Before USB, every manufacturer had their own connector. Buying a printer meant praying the plug fit your machine. After USB, you stopped thinking about connectors and focused on what you wanted to print.
MCP has a comparable effect on the AI ecosystem. When AI agents can communicate with one another fluidly, the always-on economy stops being “a collection of AI tools working in separate corners” and becomes an organic, real-time coordinating intelligent network.
I think of this as a force field: an operating space with AI as its foundational infrastructure, where information and decisions can flow without being constrained by human schedules. Once that force field forms, the efficiency inside it and the efficiency outside it are simply not in the same category.
The Reality Facing Taiwanese Companies
After all of this, back to Taiwan.
Taiwanese companies face a structural difficulty in adapting to the always-on economy: our organizational culture depends too heavily on individuals.
That is not a criticism. The flexibility of Taiwanese SMEs, the trust networks, the owner-who-works-alongside-the-team culture — these are precisely why Taiwan has earned its place in global supply chains. But they also mean that most processes follow people rather than systems. The boss remembers every client’s preferences. The senior sales rep assesses whether an order is real by feel. The factory floor manager tunes production parameters from years of experience.
This tacit knowledge is a double-edged asset in the AI era. On one side, it is genuinely valuable and difficult for AI to replicate directly. On the other, it is a source of resistance to transformation — because knowledge that has never been systematized cannot be handed off to AI. AI cannot take over what it cannot read.
My own experience is this: the first step in any transformation is not choosing an AI tool. It is spending the time to translate the team’s tacit knowledge into a form that a system can understand. The process is slow and often meets resistance. The line I have heard most often is: “This kind of thing just cannot be written into rules.” But once it is done, AI can actually do its job, rather than being an expensive toy sitting in the corner.
“Are We Ready?” Is the Wrong Question
Every conversation about AI-driven change eventually arrives at the same question: “Are we ready?”
I think the question is wrong. It assumes a state of readiness — as if you could perfect your stroke on the shore before getting in the water. But the water is already at your ankles.
More useful questions: Does your organization have an architecture capable of continuous adjustment through ongoing change? Is it flexible enough to keep adapting as AI capabilities shift roughly every three months? Do you have people who understand both the technology and the business well enough to translate between them?
In the age of AI, the always-on economy is moving from human endurance held up by willpower to intelligent instinct embedded in the system. Instinct, for an organization, is not innate. It is designed.
Those who start redesigning now will not necessarily win. But those still waiting to be “ready” have probably already run out of time.
💬 Comments
Loading...