Reading a Wall Street Journal article about the AI-driven always-on economy, I paused for a long time. Not because the article was particularly well-written, but because it made me realize something: we’ve been using the wrong framework to understand this transformation.

The article’s narrative goes like this—AI enables economies to operate 24/7 without interruption, ushering in a brand new era.

But is it really “brand new”?

The Always-On Economy Was Never New

Convenience stores in Taipei have been operating 24/7 for twenty years. Global financial markets trade in relay across different time zones, from Tokyo to London to New York—the sun never sets on trading floors. Semiconductor fab production lines run 365 days a year without stopping, because restarting a furnace costs more than keeping it running.

We’ve been living in an always-on economy all along.

The difference is that in the past, this always-on operation relied on human lives.

I remember when I first started my company, we had a period where we were running a cross-timezone project. Taiwan handled the Asian side during the day, and at night we dealt with US client demands. It sounded very international, but in reality it was just me and a few team members taking turns staying up all night, never daring to put our phones on silent. Once, I was woken up at 3 AM to handle a system alert. After investigating for two hours, it turned out to be a false alarm. But you couldn’t afford not to check, because “what if it was real?”

The cost of those days wasn’t the overtime pay on paper. It was the increasingly weary look in my team’s eyes, the growing silence in creative meetings, everyone conserving energy, only dealing with the most urgent matters at hand.

This is the truth of the old version of the always-on economy: using human physiological limits to forcibly sustain a system that exceeds physiological design.

AI Changes Not Time, But Cost

When AI enters the scene, what changes isn’t the “24-hour operation” itself. What changes is the cost structure of maintaining it.

Here’s a concrete example. My team now uses AI agents to handle first-line customer demand routing and system monitoring. Previously, a 3 AM false alarm required someone to get up and confirm it. Now AI first assesses severity levels, only triggering notifications when human intervention is truly needed. The result? Response to real emergencies is faster because the person on duty has full energy, not depleted by five previous false alarms.

This isn’t a story about “30% efficiency improvement.” This is a story of qualitative change.

When you compress the marginal cost of maintaining always-on operations to near zero, many previously unreasonable things suddenly become reasonable. Real-time monitoring of every supply chain status? Previously required an entire department, now one AI agent can handle it. Real-time risk assessment for every transaction? Previously only affordable for large financial institutions, now accessible to SMEs.

But there’s a trap here: many people think AI is just about having machines do what humans used to do. This is the most superficial understanding, and the most dangerous strategy.

Three-Layer Reconstruction: Not Upgrade, But Renovation

Observing different enterprises implementing AI, I’ve discovered a pattern: the successful ones aren’t those that “replace human labor with AI,” but those willing to rethink their entire operational logic.

This rethinking has three layers.

The first layer is process reconstruction. Traditional business processes are designed around “human working hours.” Monday meetings, Wednesday reports, Friday reviews. This rhythm made sense in the human-driven era because humans need time to digest information, form judgments, and coordinate actions. But when AI can process and analyze in real-time, this rhythm becomes an artificial bottleneck. True process reconstruction isn’t “changing weekly meeting frequency to daily,” but asking: “Do we really still need fixed-frequency meetings? Or can we shift to event-driven—gather when there’s an issue, push forward individually when there isn’t?”

The second layer is human-machine collaboration reconstruction. This layer is most easily misunderstood. Many people draw a chart: AI handles A, B, C, humans handle X, Y, Z, with handoff points here and there. This static division chart simply doesn’t work because AI’s capability boundaries shift every three months. My approach in my own company is designing a “dynamic authorization framework”—AI has basic autonomous authority, but decisions exceeding certain complexity or risk levels automatically escalate to humans. This threshold isn’t fixed but continuously adjusted based on AI performance and team trust levels. Like training a new employee: you check everything at first, six months later you only check key decisions, a year later you only check results.

The third layer is value reconstruction. This is the deepest and least discussed layer. When AI pushes execution efficiency to the limit, competition between organizations shifts to areas AI cannot easily replicate—understanding cultural context, building trust with people, making responsible judgments in ambiguous situations. I discussed a similar perspective in “Thinking in the Post-Code Era: When Taste Becomes Humanity’s Key Competitive Advantage”: when execution costs approach zero, judgment becomes the only differentiating factor. At the organizational level, this “judgment” is corporate culture, decision quality, and resilience when facing uncertainty.

From Socket to Sanctuary: The MCP Metaphor

There’s a technical development worth special mention: MCP (Model Context Protocol).

MCP is a standard protocol that allows different AI models and tools to communicate with each other. It sounds technical, but its significance extends far beyond technology.

Imagine before MCP, every AI tool was an island. Your customer service AI didn’t know what your inventory AI was thinking, your analytics AI couldn’t see your marketing AI’s data. To make them collaborate, you had to write countless custom integration programs with extremely high maintenance costs.

What MCP does is like when the computer industry standardized USB interfaces. Before USB, every manufacturer had their own connector—buying a printer meant praying the connector would fit your computer. After USB standardization, you no longer had to think about connectors, just focus on what you wanted to print.

MCP’s impact on the AI ecosystem is similar. When AI agents can communicate seamlessly, the always-on economy is no longer “a bunch of siloed AIs working in different corners,” but an organic, real-time coordinating intelligent network.

I call this a “sanctuary”—an operational space with AI as foundational infrastructure, where information and decisions can flow without being constrained by human schedules. Once this sanctuary forms, efficiency inside and outside operate on completely different scales.

Taiwan Enterprises’ Reality Challenge

Having discussed all this, let’s return to Taiwan’s reality.

Taiwan enterprises face a structural difficulty in transforming to the always-on economy: our organizational culture relies too heavily on “people.”

This isn’t a bad thing—Taiwan SMEs’ flexibility, trust networks, and culture of bosses jumping in to work alongside employees are reasons we maintain a position in the global supply chain. But this also means many of our processes “follow people” rather than “follow systems.” Bosses remember every client’s preferences, senior sales staff rely on intuition to judge order authenticity, factory supervisors use experience to adjust production line parameters.

This “tacit knowledge” becomes a double-edged sword in the AI era. On one hand, it’s a valuable asset that AI cannot easily replace. On the other hand, it’s also resistance to transformation because this knowledge hasn’t been systematized and AI cannot take over.

My own experience is that the first step of transformation isn’t implementing any AI tools, but spending time “translating” the team’s tacit knowledge into forms systems can understand. This process is painful, slow, and often meets resistance—“This kind of thing can’t be written into rules” is what I’ve heard most often. But once completed, AI can truly function rather than just being an expensive toy.

Are We Ready? The Question Itself Is Wrong

Every discussion about AI-driven transformation eventually leads to someone asking: “Are we ready?”

I think this question itself is problematic. It assumes a “ready” state, as if you could perfect your swimming technique on shore before entering the water. But reality is, the water is already at our ankles.

A more practical question than “Are we ready?” is: Do you have a framework that can continuously adjust amid change? Is your organization flexible enough to continuously adapt as AI capabilities upgrade every three months? Does your team have people who understand both technology and business, who can translate between the two?

The always-on economy, powered by AI, is transforming from “business as usual sustained by human endurance” to “instinct driven by intelligence.” But instinct isn’t innate—for organizations, it’s designed.

Those who start redesigning now won’t necessarily win. But those still waiting to be “ready” are probably already too late.