In the 19th century, English economist William Stanley Jevons found that tech-driven efficiency improvements in coal use led to increased demand for coal across a range of industries. The paradox of course being that if you assume demand remains constant, then the volume of the underlying resource should fall if you make it more efficient. Instead, making it more efficient leads to massive growth, because there are more use-cases for the resources than previously contemplated. The paradox has proven itself repeatedly as we've made various aspects of the industrial world more productive or cheaper, and especially in technology itself.
For instance, In the early years of the mainframe, units were measured in the hundreds, and only the world's largest companies could afford them. In the early years of the minicomputer (a smaller, cheaper version of the mainframe), units were in the tens of thousands. And in the early days of the PC, units were in the millions. That's a 100-fold increase for each new era of computing in just three decades.
While you would have to have been a Fortune 500 company to access powerful software to do your accounting in the 1970s, by the 2000s with the cloud, it was available to every barbershop in the world. This happened for CRM systems, communication technology, marketing automation, document management software, and nearly every enterprise software application. The advantages that a large enterprise had in procurement, installation, maintenance, computing capacity, and more, simply evaporated overnight because of the cloud.
As a result, efficiencies in computing led to the democratization of automation of deterministic work (through software) for decades in almost every field. But this has never been possible before for the non-deterministic work that represents the vast majority of things we do every day in an enterprise: reviewing contracts, writing code, generating an advertising campaign, doing advanced market research, handling 24/7 customer support, and thousands of other categories of tasks.
AI agents bring democratization to every form of non-deterministic knowledge work. And this will change most things about business. For most large companies today, they can effortlessly move resources around between projects, afford to experiment on new ideas, hire the top lawyers or marketers for any new project they need, contract out or hire engineers to build whatever new initiative they're working on. This has always been an advantage of the world's largest companies, but is a benefit that is only achieved after decades (or in some cases, centuries) of business success and survival. That means for the vast majority of companies and entrepreneurs in the world, you're at an extremely stark disadvantage on day one no matter what you do.
AI agents fundamentally change the calculus here. Now, we can dramatically lower the cost of investment for almost any given task in an organization. The mistake that people make when thinking about ROI is making the "R" the core variable, when the real point of leverage is bringing down the cost of "I". Every entrepreneur, business owner, or anyone involved in a budget planning process before knows how scarce resources are when running a business. When you're a small team, you're making decisions between having a good marketing webpage, building a new product experience, handling customer support inquiries, taking care of something important in finance, finding new distribution, and so on. Every one of these areas of investment and time are trading off from one another, all of which hold you back from growth.
Now, we have the ability to blow up the core constraint driving many of these tradeoffs: the cost of doing these activities. Roon, on X, pointed out that any consumer now has better access to education and tutoring than an aristocrat would have had, due to AI. And now, every business in the world has access to the talent and resources of a Fortune 500 company 10 years ago.
Demand will go up 10X or 100X for many areas of work because we've lowered the other various barriers to entry of doing many more types of work that most companies wouldn't have even experimented with before. Imagine the 10 person services firm that didn't have any custom software before for their business. From a standing start, it may have taken multiple people to develop a full app, keep it running, keep customer requests incorporated, ensure the software stays secure and robust, and so on. The project just doesn't even get started because of this. Now, someone on the team builds a prototype in a few days, proves out the value proposition in a matter of days. You can analogize this to any other type of work or task in an organization.
Of course, many are wondering what happens to all the jobs in this new world? The reality is that despite all the tasks that AI lets us automate, it still requires people to pull together the full workflow to produce real value. AI agents require management, oversight, and substantial context to get the full gains. All of the increases in AI model performance over the past couple of years have resulted in higher quality output from AI, but we're still seeing nothing close to fully autonomous AI that will perfectly implement and maintain what you're looking for.
It's clear that AI agents are successfully taking over various tasks that we do today (like researching a market, writing code for a new feature, creating digital media for a campaign), but incorporating those tasks into a broader workflow to produce value still requires human judgment and a ton of effort. Even as AI progresses to accomplish more of an entire workflow, we will simply expect more from the work that we're doing. Ultimately ensuring that today's jobs are tomorrow's tasks.
Historically, this actually happens all the time. If you told someone about Figma or Google Adwords in the 1970s they'd have expected marketing jobs to plummet since we could do many different jobs inside of a single role in the future; well, the opposite has happened. Back of the envelope math (from AI of course) suggests that there were a few hundred thousand people employed across marketing related job categories in the 1970's (PR, graphics, advertising, type jobs) in the US; today, it's in the low millions.
How did we experience a 5X+ increase in these jobs in 50 years at the exact same time that technology made this work far more efficient? Actually precisely because of those efficiencies. We went from advertising being the domain only of the largest companies -your CPG or car companies- to something that almost any small business could participate in. The marketing technology, CRM systems, analytics, graphic design software, targeting platforms, new distribution channels and many other tech-enabled trends allowed more companies to justify the ROI of doing more sophisticated marketing. This will similarly happen in many fields because of AI.
Jevons paradox is coming to knowledge work. By making it far cheaper to take on any type of task that we can possibly imagine, we’re ultimately going to be doing far more. The vast majority of AI tokens in the future will be used on things we don’t even do today as workers: they will be used on the software projects that wouldn’t have been started, the contracts that wouldn’t have been reviewed, the medical research that wouldn’t have been discovered, and the marketing campaign that wouldn’t have been launched otherwise.
编译摘要
1. 浓缩
- 核心结论1: AI 智能体将 Jevons 悖论引入知识工作——效率提升不会减少总需求，反而增加
- 关键证据: 历史上计算设备从大型机→小型机→PC，每个时代销量增长100倍；云让小企业获得企业级软件
- 核心结论2: 降低投资成本(I)比提高投资回报率(ROI)更能创造价值
- 关键证据: 小团队面临资源稀缺，需在多个投资领域权衡；AI降低每项任务的投入成本
- 核心结论3: AI 不会消灭工作，而是创造前所未有的新任务类型
- 关键证据: 1970年代美国营销相关岗位几十万人，如今数百万；技术效率提升反而扩大了市场
2. 质疑
- 关于"效率提升增加需求"的质疑: 历史上硬件成本下降确实带来销量增长，但AI可能存在本质差异——AI可能达到某种饱和点，能力提升不再带来边际需求增加
- 关于"工作不消失"的质疑: 如果AI最终达到完全自主，现有结论可能不成立；当前AI仍需人类管理、监督和大量上下文
- 关于数据可靠性的质疑: 声称"50年增长5倍"——需验证数据来源和统计口径；"10X-100X需求增长"是预测而非事实
3. 对标
- 跨域关联1: 此现象类似软件开发中的"工具改进悖论"——更好的IDE和框架导致更多代码被编写，而非减少程序员需求
- 跨域关联2: 类似教育培训领域——在线教育降低了获取知识的门槛，但教师需求反而增加（市场扩大）
- 可迁移场景: 任何AI降低准入门槛的领域——医疗研究、法律服务、软件开发、内容创作