Inflation is a concept that everyone has direct experience with. The same 100 yuan that could buy more things in the past can now buy fewer things. The money hasn't disappeared, and the goods haven't disappeared either - what has changed is the yardstick used to measure value. When the entire price system shifts upward, the money that was enough yesterday is no longer enough today.
So, what inflation truly changes is not just prices, but the standard of what is **"enough"**. In the past, 5,000 yuan could support a certain lifestyle, but today it may require 8,000 yuan; a project that could be undertaken with 100,000 yuan in the past may now require 150,000 yuan. Although the numbers are increasing, people do not naturally become more affluent as a result.
And as AI enters enterprises on a large scale today, we may be experiencing another phenomenon that is highly similar to inflation, yet rarely discussed seriously. I would rather call it efficiency inflation.
It is not a strictly defined economic term, nor is it a traditional price increase, but rather an increasingly evident competitive phenomenon:
When technology universally improves everyone's efficiency, the efficiency that was once a competitive advantage will gradually be absorbed by the entire system and eventually become the new minimum standard.
As people work faster, companies produce more, and organizations respond more promptly, it is strange that people have not become noticeably more relaxed and companies have not necessarily made more money. Instead, we have collectively entered a new, faster competitive paradigm.
The reason why the 650 points we had when we were young may become 700 points today
There's a very intuitive question worth considering first: why is it that in many people's memories, a score of 650 was once enough to determine one's fate, while today in some highly competitive regions, the threshold for top universities seems to be getting higher and higher?
Strictly speaking, it's not possible to make simple horizontal comparisons between different years, provinces, and exam papers. However, if we temporarily set aside these technical differences and look at the underlying competitive logic, we'll see a thought-provoking fact: students have become stronger, learning methods have improved, and educational resources have become more abundant, but Tsinghua remains Tsinghua, and Peking University remains Peking University. The scarcity of top-tier universities has not expanded with the improvement in efficiency of the entire education system.
In the past, there was no mature question bank, no systematic training industry, no rich online courses, no highly summarized methodology, and no complete information dissemination system. Later, everyone became more skilled at taking exams. The result was not that it became easier for everyone to get into top universities, but rather the competition standards were raised together - the excellence represented by 650 points yesterday may just be one of many competitive positions today.
What's truly interesting about this is:
When the rewards themselves are still scarce, the overall improvement in participants' abilities will eventually become a universal increase in the threshold.
Your learning efficiency has improved, and so has everyone else's; you're completing exercises faster, and so is everyone else. Everyone's absolute abilities are rising, but what ultimately determines the outcome is still relative position. As a result, everyone is getting stronger, but also more exhausted. This is the most counterintuitive aspect of the competitive system: technology and methods can enhance a person's absolute abilities, but they cannot automatically improve a person's relative position.
It's worth noting that this logic is not unique to the exam scene, it just happens to be particularly pronounced in the context of the gaokao.
AI Repeats the Same Story in the Corporate World
Today, almost all enterprises are discussing AI-driven efficiency gains, and these gains are indeed real: programmers can write code faster, marketing teams can generate content more quickly, sales teams can manage more clients simultaneously, customer service can respond 24 hours a day, and analysts can process information in minutes that previously took days to complete. Looking at these numbers alone, AI is undoubtedly a massive productivity revolution.
The issue is that what companies ultimately care about is not "can we do things faster today than yesterday," but rather "can we make more money as a result." And there is no inherent equivalence between these two things.
If only one company possesses AI, the increased efficiency can certainly be converted into excess profits: a business that originally required 100 people can now be completed with 60, and customers still pay the original price, so costs decrease and profits naturally rise. This is the most beautiful stage of technological dividend, but it usually does not last too long - because AI will not belong to just one company, competitors will also use it, and the entire industry will start to improve efficiency.
When all companies can accomplish with sixty people what previously required a hundred, having sixty people do the job is no longer an advantage, but merely the normal standard. Then the market begins to reprice: as costs decrease, customers naturally demand lower prices; as delivery speeds up, customers naturally demand shorter cycles; as one person can serve more customers, companies naturally redefine the amount of work one person should handle. The surplus originally created by technology is gradually absorbed by the competitive system, ultimately resulting in an absurd yet real situation:
Every company is more efficient than it was five years ago, yet none feels that doing business has gotten any easier.
Everyone got faster, so speed lost its value. Everyone got cheaper, so low cost became the default expectation. Everyone could mass-produce content, so content itself grew increasingly worthless. Everyone could respond around the clock, so customers began to treat 24/7 responsiveness as a given. This logic is almost identical to that of the college entrance exam: once, 650 points was an advantage; later, 650 points was merely the ticket to entry. Today, knowing how to use AI may still count as a capability. A few years from now, not knowing how to use AI may simply mean you are not qualified to compete.
The Truly Scary Part of Efficiency: It Redefines 'Normal'
If the efficiency dividend were merely absorbed by prices, the situation would remain relatively straightforward. The deeper shift lies in expectations.
Technological revolutions are always exhilarating at first, because what people see first is the time saved: where one task took a full day, now five can be completed. The immediate reaction is often — we can finally take it easy.
But competitive systems rarely operate that way. Once a person can complete five tasks in a day, the organization quickly asks: if five can be done in a day, why assign only one? The client asks: if technology can deliver in real time, why should I wait three days? Capital asks the same: if unit costs have already fallen, why hasn't growth accelerated? New productivity is soon written into new expectations—yesterday's "excess capacity" becomes today's "normal capacity."
This is the deepest layer of efficiency inflation:
Efficiency is not just about making people work faster, it will ultimately redefine what constitutes a reasonable workload.
So historically, many technological advances have not directly translated into leisure but have instead first shown up as greater work density. Computers did not make offices disappear, email did not reduce communication, instant messaging did not eliminate meetings, and mobile internet did not truly let people leave work behind—they simply enabled more tasks to fit into each unit of time.
AI is likely to push this trend forward by another order of magnitude. Where a person could previously handle ten problems in a day, in the future they may handle a hundred. The real question, then, is whether the additional ninety units of capacity ultimately become ninety units of free time or ninety new tasks. The answer is not determined by technology, but by competitive structure.
Why Productivity Growth Doesn't Equal Profit Growth
Once it is acknowledged that the competitive structure is the distributor, it is necessary to distinguish between two concepts that are often confused.
Productivity measures how much you can accomplish with fewer resources, while profit measures how much value you can retain once all competitors are in the game. The former is an absolute capability; the latter is a relative outcome. If one company doubles its efficiency while competitors stand still, that is a massive advantage. But if the entire industry doubles its efficiency, the likely result is simply that industry pricing, delivery speeds, and customer expectations all shift together, and competition finds a new equilibrium.
The efficiency dividend that emerges is not lost—it is merely redistributed: a portion flows into corporate profits, a portion becomes lower prices for consumers, a portion becomes better products, a portion becomes faster service, and a large share is converted directly into new competitive requirements. So the question entrepreneurs should truly ask themselves is not simply "How much efficiency has AI improved for us?" but rather:
Where do the efficiency gains from AI ultimately end up on whose balance sheet?
This is a strategic question of an entirely different order. If all efficiency gains are ultimately repriced by the market in short order, then a company can become extremely efficient and still only manage to avoid being eliminated. It hasn't actually captured a new profit pool.
The Most Dangerous Move: Using New Productivity to Play the Old Game
This is the biggest misconception many companies have when using AI today: they gain a new productivity, but use almost all of it to complete things that already existed in the past. Previously, ten people wrote code, now five people do; previously, one week was needed to write ten articles, now ten articles are written in one day; previously, one salesperson served fifty clients, now they serve three hundred; previously, a product was released four times a year, now it is released forty times a year.
All of this has value, of course, but it remains optimization within the same competitive framework. You've simply swapped in a new engine to keep running the same track. And once every competitor installs the same engine, the outcome is all too familiar: everyone accelerates together, the average speed rises, yet the race never ends—it only grows more brutal.
Since accelerating on the old track cannot solve the problem, the only direction worth serious discussion is how to leave it.
Six, the real breakthrough is not about doing things faster, but about doing things that would not have happened in the past.
If productivity gains materialize, the truly important question emerges: where should the additional productivity be applied?
My answer is this: don't use all of it to do yesterday's work faster. Instead, take the most important portion and devote it to things that simply weren't viable under the old productivity conditions. By "not viable," I don't necessarily mean humanity never imagined them. More often, it means: technically possible in the past, but not economically worthwhile; there was demand, but the cost of serving it was too high; it could be customized, but not at scale; it could be done manually, but marginal costs couldn't be brought down; only a very few could afford it. Truly massive productivity revolutions rarely consist of making old products a little faster. They consist of turning these "uneconomic" things into economic ones for the first time.
We've always known one-on-one instruction is more personalized than large-class teaching. The problem: no single excellent teacher can provide continuous one-on-one service to 100,000 students at once. The demand has always existed—the cost structure just never allowed it. If AI gives every student a private tutor that understands their learning state in real time, retains their weaknesses in long-term memory, and continuously adjusts teaching content, the result isn't merely faster lesson preparation. It's a service that previously couldn't exist at scale.
Similarly, companies would naturally love for every client to have a dedicated research team, but it's impossible to staff 100,000 analysis teams for 100,000 customers. If AI enables every client to receive continuous information services comparable to having their own independent research team, this isn't just "researchers writing reports faster" — it's an entirely new tier of service that never existed before. The software industry is even more illustrative: every user has inherently different needs, yet software must be standardized, otherwise customization costs become unbearable. But if future software can generate interactions, logic, and workflows in real time based on the user's current task, then software might no longer be the fixed, static product we understand it to be today.
This is where the productivity revolution truly unlocks its potential.
Tech Revolutions Create Wealth Beyond Old Boundaries
The value of what lies "beyond the boundaries" is easily overlooked because every time humanity confronts a new technology, it initially tries to understand it through the language of the old world. The automobile was first understood as a "horseless carriage," computers were initially used largely to simulate paper and spreadsheets, and in the early days of the internet, many websites simply moved offline information online.
AI is no different. Today we ask: Can AI help programmers write code, can it help marketers produce copy, can it help customer service teams respond to inquiries, and can it reduce headcount by how much? These questions are certainly important. But if we look back a decade from now, what truly reshapes the economic structure may well go beyond all of this—because these applications all operate within the old production frontier. The genuinely enormous value tends to emerge somewhere else entirely:
New productive forces are making possible, for the first time, economic activities that did not previously exist.
The most interesting part of the productivity revolution isn't shifting the existing curve upward—it's pushing outward the entire boundary of what's worth doing. A service that once cost a hundred units to provide can now be delivered for one. In that case, the real question isn't necessarily whether to cut the price of a 150-unit service to 120. It's this: what things were previously so expensive—at a hundred units—that nobody would buy them at all, and now, at one unit, can suddenly become a massive market?
This is where new productive forces are most worth seeking—and it delineates two starkly divergent paths for enterprises.
Efficiency Competition vs. Boundary Innovation: Two Fundamentally Different Corporate Strategies
In the future, companies using AI may become increasingly divided into two distinct categories.
The first type of company keeps asking: How much more can we cut costs? How many more people can we let go? How much faster can we deliver? This is efficiency competition. It's necessary, of course—if everyone else improves and you don't, you'll be eliminated. But efficiency competition is fundamentally defensive: it keeps a company at the table, yet doesn't necessarily create a new table to play at.
The second category of companies asks a different set of questions: With new productivity, what products that never existed before can now exist? What services that only a few could access are now ready to be democratized? What things that could never scale are now finally scalable? What demand that never had viable unit economics now works for the first time? This is boundary innovation.
The former is about running faster within the old world, while the latter is about searching for a new one. If efficiency inflation truly becomes a defining phenomenon of the AI era, then genuine excess profits will likely concentrate increasingly in the second category of companies. Efficiency dividends in the old world will be rapidly diluted by competition; only when a company carries new productive forces into a space where stable competitive standards have yet to form can it truly retain the value created by technology.
Don't Just Use AI to Cut Costs
For today's companies, then, there is a question that may be more important than "how AI improves efficiency":
What can AI enable us to do for the first time that we were fundamentally incapable of doing before?
This is the real strategic question. If the answer is merely hiring fewer people, writing more code, generating more content, and completing projects faster, then AI will likely end up as just another piece of office software. It will be extremely important, but the efficiency it brings will eventually be written into the cost structure of every company—becoming infrastructure that everyone has and therefore no one can leverage for advantage.
Just as no company today gains a lasting competitive edge from using Excel—it's essential, but everyone has it—AI will likely follow the same path. The businesses that truly generate new profits won't be those that simply "own AI," but those that harness it to redefine a product, service, organizational model, or market structure that previously had no economic viability.
Efficiency Inflation Reminds Us: Don't Waste Productivity on Yesterday
Humans have always hoped that technology would make life easier, but history has repeatedly offered the opposite answer: rising productivity does not automatically bring freedom. It may instead bring more production, more consumption, more competition, more expectations, and higher barriers to entry.
When everyone gains access to the same efficiency, that efficiency depreciates like currency. Yesterday, one unit of efficiency could buy a competitive advantage; today, it takes two. A score of 650 once meant excellence; today, it may take 700. Tsinghua remains Tsinghua, and Peking University remains Peking University—what has changed is simply the cost of capability required to enter the competition.
Companies face the same predicament. AI can make every business smarter, faster, and more efficient. But if all enterprises simply channel that new productivity into repeating what already existed yesterday, what we end up with may not be a more relaxed business world — but a faster, more demanding, and more intensely competitive version of the old one.
So what is truly worth seeking in the AI era has never been more efficiency — it's the next step that comes after it.
The real dividend of every productivity revolution lies not in doing yesterday's work faster, but in making what wasn't worth doing yesterday worth doing for the first time today.
Perhaps the real way for companies to break through efficiency inflation is not to keep pushing scores from 650 to 700, but to seek out an exam room that never existed before.