Hello! I'm your financial journalist and economic scholar friend. The article we're going to discuss today might seem quite complex at first glance, with terms like "the religization of efficiency," "friction," and "irreversibility," which sound like topics from a philosophy class. But don't let those words put you off; it actually addresses a very poignant and realistic issue:
Why do companies, even when they are aware of the risks, still obsessively pursue speed and cost-cutting? What kind of abyss are they pushing us towards?
The main argument of this article is quite sharp: Efficiency was once just a tool, but now it has become a kind of "faith." When efficiency is no longer questioned, and when slowness and redundancy are seen as sins, businesses lose a crucial ability—the ability to stop in critical moments.
Below, I'll break down this long article into five parts and explain it in plain language.
---
1. The order has been reversed: From "why we need to be fast" to "fast just for the sake of being fast"
In the past, the logic of management was clear:
1. Step 1 (goal): What do we want to achieve? (For example, making cars safe or food delicious.)
2. Step 2 (method): How can we do it with the least money, time, and people? (This is where efficiency comes into play.)
Now the order has quietly reversed.
In many large companies, people no longer ask why a process exists; instead, they ask why it can't be faster. They don't discuss how much safety margin is needed; instead, they wonder why resources aren't being fully utilized.
It's like going to a restaurant. Before, the chef would ask, "How spicy do you want it? How cooked?" (Focusing on the goal and quality.) Now, the chef might just say, "I have to serve it in 30 seconds, or I'll get paid less." (Only focusing on speed and efficiency.)
What are the consequences?
Quality, safety, and even people's time for thinking have become things that need to justify their existence in terms of efficiency.
- If you want an extra safety check, you have to prove it won't slow down the process.
- If you want to keep some inventory as a precaution, you have to prove it's not a waste.
- If you want manual verification, you have to explain why machines can't do it alone.
When everything has to bow to efficiency, efficiency becomes the judge, even a god.
---
2. Efficiency has become a "new religion": It's not the faith that's wrong; it's that you're not executing it well enough
There's a brilliant metaphor in the article: Efficiency is turning into a kind of "religion."
What are the characteristics of a religion?
1. Unquestionable goodness: In the eyes of efficiencyism, faster = progress, fewer people = optimization, higher utilization = management ability. These assumptions are taken for granted as truths.
2. A closed-loop of explanation: If an accident occurs due to the pursuit of efficiency (like too many layoffs leading to a service collapse or automation errors causing losses), the organization doesn't reflect on whether they went too far with efficiency. Instead, they say, "It's because our automation isn't perfect, or our processes aren't optimized enough."
This creates a vicious cycle: Whenever the results are poor, the problem is attributed to the faith not being followed thoroughly, not to the limitations of the faith itself.
This is dangerous because it means risks can accumulate without external triggers. They grow quietly with each iteration of "one more optimization, one more speed increase." It's like someone who thinks they're healthy and keeps staying up late and eating junk food until they suddenly fall ill, only then realizing that their "efficient lifestyle" was actually depleting their health.
---
3. The green lies on the dashboard: Good numbers don't mean the system is safe
Modern companies have many "rituals": KPIs, OKRs, weekly reports, dashboards (data displays).
These tools are supposed to help managers see the reality. But over time, they end up defining reality.
- If the dashboard shows green, everyone thinks everything is fine.
- As long as the growth rate is rising, internal structural problems can be ignored.
- If server utilization goes from 40% to 95%, everyone celebrates.
But there's a huge trap: In systems engineering, resilience means allowing some resources to seem "inefficient" during normal times.
- Airlines need redundant backup systems.
- Banks need capital buffers.
- Data centers need backup power.
- Supply chains need safety stockpiles.
If you raise server utilization to 95%, you almost have no buffer. A sudden surge in traffic or a local failure can paralyze the entire system, and the cost of recovery is extremely high.
The so-called "waste" is sometimes just the "insurance fee" the system pays for uncertainty.
When companies cut this insurance, they don't become stronger; they just shift the risk from today's balance sheet to an unknown future moment. That moment might be the day the system crashes.
The most dangerous times are often when all the numbers look perfect. The real world won't stop causing problems just because your dashboard is green.
---
4. The fatal flaw of the AI era: Mistakes are no longer isolated incidents; they become a flood
For the past hundred years, humans have been trying to eliminate friction.
- Taylorism eliminated wasted movements.
- Assembly lines eliminated waiting times.
- The internet eliminated transaction friction.
- Now, AI agents aim to eliminate cognitive and execution friction.
Previously, tasks required human understanding, operation, and confirmation. People might hesitate or notice anomalies—these inefficiencies were natural filters and buffers.
Now, AI agents are tireless, never complain, respond in milliseconds, and handle thousands of tasks simultaneously. They take efficiency to the extreme.
Here's a huge risk that's often overlooked:
We worry about AI making mistakes (illusions, misunderstandings). But what's more frightening is: What if AI is mostly correct?
- A person might make 10 mistakes a day.
- An AI agent can execute 100,000 actions a day.
In the past, a wrong decision would be filtered and diluted through multiple layers of people and processes. Now, mistakes can be amplified by automation systems with the same efficiency.
The error rate might remain the same, but the "volume" of mistakes increases dramatically.
It's like fish escaping the net one by one before; now, an entire river of fish is lost because no barriers were in place.
So, the biggest problem of the AI era is not whether AI is smart, but who has the authority to stop it when it's running at full speed?
---
5. Re-inventing "friction": Some difficulties must be preserved
The article ends with a profound suggestion: We need to re-understand friction.
In the past 20 years, the tech industry has favored a "frictionless experience." Frictionless payments, frictionless deployments. In that context, friction was indeed a cost.
But in the era of AI agents, some friction is a "safety valve" designed by society to prevent irreversible mistakes:
- Double confirmations for bank transfers.
- Circuit breakers in the stock market.
- Dual authorization for critical operations.
- Checklists for pilots.
- Emergency stop buttons on industrial equipment.
These "troubles" are there to prevent certain things from happening too easily.
Future corporate architectures need to distinguish between two types of actions:
1. Reversible actions: You can search again if you miss something, or correct a written mistake. These can be optimized to extreme efficiency, even fully automated.
2. Irreversible actions: Money has been transferred, a machine has been started, data has been deleted, a drug has been administered. These actions cannot be undone once they happen.
The conclusion is:
- Cognitive aspects: You can pursue unlimited efficiency.
- Real-world execution: There must be boundaries.
A company's security architecture should shift from "governing AI" to "governing execution." It's not about preventing AI from thinking or working, but about maintaining a question that's independent of the execution system before it actually changes reality:
"Should this really happen right now?"
This question must be guaranteed by structure, not by individual discretion. In a system that executes every second, personal caution cannot be extended as needed.
---
Summary: Efficiency is not the enemy; the problem is efficiency without boundaries
Finally, I want to clarify: This article is not against efficiency.
Most of modern civilization's progress comes from improved efficiency. Without it, we couldn't afford smartphones, the internet, or cheap food.
The real question is not whether we want efficiency, but whether there is an upper limit to it.
A mature, powerful organization's most important ability is not to always know how to be faster, but to know when to stop being fast.
In the past hundred years, we've taught companies how to eliminate waste. Next, AI might force us to learn something even harder: how to identify things that seem like waste but actually protect the system.
Because the most dangerous moment for a system is not when it's most chaotic, but when it's running smoothly—without waiting, without hesitation, without redundancy, and without anyone to say, "Stop for a moment."