How to Understand AI’s Economic Future with Anthropic’s Scenarios

In September 2026, Anthropic’s Economics team published Scenarios for Our Economic Future, an interactive model of how AI might affect jobs, growth, and unemployment in the US through 2030.

Unlike a normal forecast, it does not give one answer. It lets you:

• Read three distinct scenarios for the economy
• See how each one changes GDP, jobs, wages, and who receives the gains
• Enter your own predictions about AI
• See the 2030 economy your predictions imply
• Compare your view with other people’s

It is based on the technical report Economic Scenarios for Transformative AI (Korinek et al., 2026). Source: anthropic.com/institute/econ-scenarios

It is a tool for thinking about possible futures, not a prediction of the future.

Anton Korinek, Chad Jones, Szymon Sacher, Tess Cotter, and Peter McCrory developed the economic model and co-authored the technical report.

Eighteen outside economists, including Daron Acemoglu, David Autor, Pete Klenow, Ben Moll, David Romer, and Jón Steinsson, gave detailed comments on an early draft. Anthropic notes that reviewers were not asked to endorse its conclusions.

The model represents every job in the economy as a bundle of tasks, based on the US Department of Labor’s O*NET task lists. AI can affect each task in one of four ways:

Unchanged

AI does not affect the task.

Augmented

AI helps a person do the task better or faster.

Automated

AI does the task itself.

New

AI creates tasks that did not exist before.

Jobs do not simply disappear or survive. Their bundles of tasks change.

Anthropic uses a nurse to show how one job splits across all four outcomes:

Only humans

Bathing a patient

Augmented

Drafting discharge instructions, monitoring patients remotely, planning the shift’s care schedule

May be automated

Charting a patient’s vitals, ordering the ward’s supplies

New tasks

Checking how well an AI triages patients, reviewing an AI-proposed care plan

The result: the nurse oversees and accomplishes more, and can spend more time talking with patients. Productivity rises.

Add up every task performed in the US, by people and by the machines and software they use, and you get the US economy: over $30 trillion of value created in the past year.

How AI changes that economy depends on:

• Which tasks AI can do
• Whether it augments or automates them
• How much more productive it makes people
• How quickly workers and companies adopt it
• Which new tasks it creates

Modest

Impact similar to the internet. Hard to see in macroeconomic data.

Substantial

A bigger impact than the internet or the railroad.

Extreme

A completely transformed, unprecedented economy, likely driven by recursively self-improving AI and faster adoption.

AI has roughly the same kind of impact as the internet did. It drives real gains, but they stay within the historical norm for new technologies and arrive gradually.

GDP in 2030

$34.1 trillion, 1.6% higher

Labor share

59.4% to workers, 40.6% to capital

AI can do half of all knowledge work by 2030, the majority of it autonomously. But it is not adopted for all of that work, so most knowledge work tasks are still done without AI.

GDP in 2030

$36.3 trillion, 8.3% higher

Growth

The economy grows at twice its normal rate

Wages

Knowledge workers’ wages stay essentially flat. Other workers see gains.

Labor share

56.1% to workers, 43.9% to capital

AI is more productive than humans at the vast majority of knowledge-work tasks, does nearly all of them autonomously, and creates essentially no new knowledge tasks for people.

GDP in 2030

$44.4 trillion, 32.4% higher

Growth

Annual GDP growth reaches 15%, so the economy doubles in size every 4.5 years

Jobs

Many fewer people work in knowledge work, and unemployment rises beyond typical recessionary levels

Wages

Knowledge workers’ wages fall by more than 10% by 2030

Labor share

45.2% to workers, 54.8% to capital

Anthropic says this scenario would likely require recursively self-improving AI, adopted quickly for knowledge work.

GDP rises in all three scenarios, but the scale varies enormously, from 1.6% to 32.4% higher by 2030. All GDP figures are at 2025 price levels.

Growth is not the question. How the growth is shared is.

There is always churn in the job market. In the substantial and extreme scenarios, knowledge workers face much more automation and displacement.

Anthropic’s example: coders and call center agents may have to switch to jobs like electrician and nurse, which are less exposed to AI.

Switching occupations is hard because people:

• May not want to change occupations
• May need to learn new skills
• Still need time to land a new job

In most scenarios, job reallocation and unemployment stay within ranges history has seen. In the extreme scenario, unemployment could spike to historic levels.

Average wages rise in all three scenarios, but the increase is concentrated outside knowledge work.

• Less demand for human knowledge work puts downward pressure on those wages
• Workers take time to move to occupations where demand is rising
• Faster designs and permits can mean more construction projects
• More demand for construction workers pushes their wages higher

Knowledge workers’ wages are essentially flat in the substantial scenario and fall by more than 10% in the extreme one.

Today, about 60 cents of each dollar the economy produces goes to workers and about 40 cents to capital.

Modest

Capital share up 0.6 points

Substantial

Capital share up 3.9 points

Extreme

Capital share up 14.8 points

In the extreme scenario, total labor income is barely changed by 2030, even though the economy is far larger.

Anthropic: the main challenge is not achieving growth, but sharing its benefits broadly.

In August 2026, Anthropic surveyed 10,980 Americans, with Morning Consult, about AI’s capabilities, adoption, and how easy it would be to find new work.

• The typical answer implies outcomes close to the substantial scenario
• That means GDP about 10% higher by 2030 than it would be without AI
• And an overall unemployment rate of around 5%
• Around 10% of respondents hold views in line with the extreme scenario

Capabilities

What share of tasks can AI do?

Adoption

How much do people use AI?

Autonomy

How much does AI do by itself, from almost none to almost all?

Productivity

How much more productive does AI make people, from the same to 10 times or more?

Adjustment

How long does it take people to find a new job, from one to two months to three years or more?

Your answers produce your own version of the 2030 economy.

Anthropic calls the model a stark simplification. Version 1.0 leaves out:

• Hyper-capable robots
• Policy responses
• Business cycles
• Aggregate demand and financial market disruptions
• Catastrophic risks
• Demand effects from the data center buildout
• Individual workers’ paths, so displacement costs are only roughly captured

Reviewers also disagreed on the edges. Some read the extreme scenario as a thought experiment, others felt the modest one understates what is already visible, and some argued AI could speed up technological progress more than the model assumes.

Actual outcomes may differ materially. Use the scenarios to think, not to bet.

Apply the model to yourself. List what you actually do in a normal week:

• Recurring tasks
• Writing and documents
• Analysis and decisions
• Meetings and conversations
• Hands-on or physical work
• Relationships with clients or colleagues
• Tasks you supervise or review

Be specific. “Write the weekly sales report” is a task. “Marketing” is not.

For each task, ask whether AI is likely to leave it unchanged, augment it, automate it, or create new work around it.

Unchanged

Protect and deepen these strengths

Augmented

Learn to do these faster and better with AI

Automated

Plan for them to shrink

New

Move toward these early

Open the scenario explorer and answer the five questions honestly. Then:

• Note which scenario your answers are closest to
• Compare your view with the survey results
• Change one answer at a time to see what matters most
• Write down why you chose each answer

Revisit your answers every few months as the evidence changes.

In every scenario, people who use AI well on augmented tasks are better placed. Practice:

• Giving AI clear context and goals
• Reviewing and correcting AI output
• Checking facts and sources
• Supervising multi-step AI work
• Explaining AI-assisted results to others

The model shows that the time it takes to find new work drives unemployment. Reduce yours:

• Keep a current record of your skills and results
• Learn one adjacent skill each year
• Build relationships outside your current role
• Notice which nearby occupations are growing
• Keep a financial buffer where you can

Scenarios are only useful if you update them. Watch for:

• Anthropic’s Economic Index, which measures how AI is used across the economy now
• Job postings and wages in your field
• How quickly your own employer adopts AI
• How much of your work AI handles on its own
• Official unemployment and wage data

Act as my career analyst. My job is [role] in [industry]. Here are the tasks I do in a typical week: [list]. For each task, classify it as unchanged, augmented, automated, or new-task opportunity, based on what AI can do today. Explain your reasoning briefly and mark anything you are unsure about. Then suggest three skills to build and two adjacent roles to explore. Do not invent statistics.

• Treat any scenario as a forecast
• Quote the extreme scenario as what will happen
• Ignore the scenario you find least likely
• Assume your job is safe because AI cannot do all of it
• Assume your job is gone because AI can do some of it
• Share the numbers without saying they come from a model