Anthropic posted another article: What will the economy look like in the AI era?

Will AI bring unprecedented economic growth, or will it trigger a massive wave of unemployment? Will it be neither, or will it lead to a future we have yet to foresee?

The Anthropic economics team recently released an economic scenario model that attempts to quantify the potential impact of AI on employment, growth, and wages. The model is made available to the public in the form of an interactive tool, allowing everyone to preview the corresponding economic scenario based on their own assessment of AI's capabilities.

The model revolves around three scenarios : from a moderate and gradual "moderate scenario," to a "substantial change scenario" where economic growth doubles, and then to an "extreme scenario" where average annual GDP growth reaches 15%, but a large number of knowledge jobs disappear.

Research has found that in most scenarios, unemployment rates and wage fluctuations remain within the historically normal range; however, in extreme scenarios, knowledge workers face declining wages or even long-term unemployment, and the share of capital in economic outcomes will also increase significantly.

Anthropic stated that this model, along with broader research, will guide its funded labor market studies and provide a basis for relevant policy recommendations, with the goal of ensuring that the economic benefits of AI can be widely shared throughout society.

The current version of the model is v1.0, which will be released in September 2026. The researchers acknowledge that the model has some limitations and say they will continue to iterate and update it.

Redefining the Economy: All Work is a "Task Package"

To understand how AI impacts the economy, we must first break down our preconceived notions about "work." In Anthropic's model, all jobs in the economy are not isolated entities, but rather comprised of a series of **"task packages"**.

Take a nurse's day as an example: Her work includes making rounds, drawing blood, triaging patients, recording vital signs, ordering ward supplies, and so on. With the intervention of AI, how will this "task package" change?

Some tasks can only be done by humans: for example, bathing a patient is beyond the capabilities of AI.Some tasks will be enhanced by AI: AI can help nurses draft discharge guidelines, remotely monitor patients, and plan shifts faster and better.Some tasks will be completely automated: such as recording vital signs or ordering consumables, which AI can do entirely for them.New tasks will emerge: History shows that new technologies always create new jobs. For example, nurses may need to review AI triage results or evaluate AI-proposed care plans in the future.

What are the results? As AI is integrated into the workplace, nurses' productivity increases dramatically, allowing them to spend more time communicating with patients. When this "task-level reshaping" occurs millions of times daily across all industries nationwide, it constitutes a massive economy worth over $30 trillion in the future.

GDP, the labor market, and workers' wage share will all depend on the speed and breadth of AI-enhanced or automated tasks.

Previewing the Future: Three Possible Scenarios

The future economic trajectory depends on the speed at which AI capabilities evolve and their adoption across industries. Anthropic's report highlights three distinct macroeconomic scenarios.

The first type is a mild scenario with gradual benefits.

In this scenario, AI's economic impact is analogous to the widespread adoption of the internet. It does drive real economic growth, but the growth rate remains within the historically normal range for new technologies; everything happens gradually. In macroeconomic data, you might not even immediately perceive the dramatic impact of AI.

The second major scenario is the revolution in mental labor.

By 2030, AI will be able to perform half of all knowledge-based jobs (most of which will be autonomous), but it will not be fully adopted. At this point, the economic growth rate will reach twice the normal level. It's worth noting that knowledge workers' wages will stagnate, while other types of workers will benefit.

A survey of over 10,000 Americans conducted by Anthropic in August showed that the expectations of most ordinary people are most consistent with this scenario: by 2030, GDP will be 10% higher than it would be without AI, and the overall unemployment rate will rise to around 5%.

The third scenario is that in extreme situations, profound economic reshaping will occur.

In this scenario, AI surpasses human productivity in the vast majority of knowledge-based tasks, completing almost all work autonomously and creating virtually no new knowledge-based tasks for humans. This typically requires AI to possess the ability for "recursive self-improvement" and to be rapidly adopted.

The results are staggering: annual GDP growth will reach 15%, and the economy will double in size every 4.5 years. As a society, we will be richer than ever before; but at the cost of a sharp decline in the number of knowledge workers and unemployment soaring to levels exceeding those of a typical economic recession.

Four harsh truths about 2030

Based on the above model, Anthropic derived four key findings that every professional should reflect upon.

Finding 1: The economic pie is getting bigger, but this may be accompanied by historic unemployment.

In all scenarios, AI will drive GDP growth and significantly increase overall societal wealth. However, in extreme scenarios, due to AI's recursive self-improvement and rapid proliferation, unemployment could soar to record highs.

Finding 2: A painful career reshuffle.

In major and extreme scenarios, knowledge workers will face severe automation replacement. On an individual level, programmers and call center customer service representatives may have to switch careers to become electricians or nurses, who are less affected by AI.

However, switching industries is extremely difficult: workers may be unwilling to change, need to learn new skills, and new jobs are not easy to find. The more this reshuffling requires, the more people will fall into prolonged periods of unemployment.

Finding 3: Wages are becoming increasingly polarized, with knowledge workers facing pay cuts.

The model shows that while average wages across society will rise, this will primarily be concentrated in non-knowledge-based jobs. Due to reduced demand for human mental labor, knowledge workers' wages will face downward pressure.

Conversely, AI improves the efficiency of mental labor, such as producing the design and permitting of physical infrastructure more quickly, which leads to a surge in demand for physical labor such as construction, driving up blue-collar wages.

In major scenarios, knowledge workers' wages will remain largely unchanged; however, in extreme scenarios, their wages will drop by more than 10% by 2030.

Finding 4: Capital is saturated, while the distribution of labor is shrinking.

Today, for every dollar of economic output, roughly 60 cents go to labor and 40 cents to capital. But as AI takes over more tasks, the technologies and resources that capital uses to create wealth will become more valuable and in higher demand, thus driving up their prices.

The study found that in major and extreme scenarios, the labor share will decline significantly, while the capital share will rise. In the extreme scenario, despite a rapid expansion of the economy, total income for the workforce will remain almost unchanged by 2030.

The vast majority of knowledge workers will face pay cuts or unemployment, while the working class will only get a smaller slice of a larger pie.

In conclusion: The future is not predetermined.

Faced with such a somewhat grim prediction, Anthropic stated that the economic landscape of 2030 is not an immutable fate. It depends on what AI can actually do, how businesses and workers apply it, and most importantly, how the financial benefits of this technology will be distributed.

The Anthropic team also admitted that, as a version 1.0 economic model, it greatly simplified complex realities, such as not taking into account policy responses, economic cycles, or the emergence of super robots.

But the core value of this explorer lies in its warning: in the face of extreme AI prosperity, our main challenge is no longer how to achieve economic growth, but how to ensure that the benefits are widely shared, rather than allowing the costs to be unfairly imposed on certain groups, especially knowledge workers.

The train of the AI era has started running. We need to pay attention not only to how fast it is going, but also to whether everyone on the train can have their own seat.

Risk warning and disclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.