AI Unemployment by 2030: Anthropic's Three Economic Scenarios Explained

Most conversations about AI unemployment jump straight to the dramatic version: machines take the jobs and unemployment soars. A new model from Anthropic makes a quieter outcome look more likely, and for people paid for what they know, arguably more important to plan for.

What Anthropic released

In September 2026, Anthropic published an interactive explorer based on its technical report on economic scenarios for transformative AI. It lets users set assumptions about how capable AI becomes and how widely it is used, then shows what the US economy would look like in 2030 under those assumptions.

Anthropic describes it as version 1.0 and is explicit that it "is not a prediction." It leaves out policy responses, business cycles and financial market disruptions, among other factors.

The three scenarios

Anthropic's three scenarios for the US economy in 2030

Scenario	GDP vs no-AI path	Unemployment	Knowledge-worker wages
Modest	About 1.6% higher	Limited disruption	Rise modestly
Substantial	About 8.3% higher	Around 5%	Do not rise
Extreme	32.4% higher	Beyond recession levels	Fall more than 10%

Which scenario do Americans expect?

Anthropic paired the model with a Morning Consult survey of more than 10,000 Americans, run in August 2026. Running the median respondent's answers through the model produces an outcome close to the substantial scenario.

In other words, the typical American's expectations imply a growing economy, unemployment around 5%, and knowledge workers whose pay stops rising while other workers gain.

Why the middle scenario deserves the most attention

The extreme scenario gets the headlines because it is frightening. The substantial scenario is easier to overlook precisely because it looks manageable.

• The economy grows. There is no recession to trigger a policy response.
• Unemployment stays moderate. Around 5% does not read as a crisis.
• The pain is a flat line, not a cliff. Knowledge workers keep their jobs while their wages stall.

A stalled wage does not arrive as an event. Each year looks roughly normal. The difference only becomes visible when you compare where you are with where the rest of the economy went.

Who gains and who stalls

In the substantial scenario, the workers who see gains are broadly those whose work depends on physical presence, hands-on skill or in-person service. Those who stall are those paid mainly for information, analysis and expertise.

For a consultancy, an agency, an accounting practice or a legal team, that is not a small slice of the workforce. It is almost all of it. The same pattern shows up at small scale in Gusto's payroll data, where AI-adopting small businesses mostly hired hands-on, customer-facing roles.

Is AI unemployment the right thing to worry about?

For most knowledge workers, probably not first. Outright replacement is the extreme scenario, and even Anthropic's model treats it as one of three paths rather than the expected one.

The more practical risk is rate pressure: keeping the job while the value of each hour slowly declines relative to what AI can nearly do. That calls for a different response than preparing for sudden job loss.

How to prepare for flat wages rather than lost jobs

1. Move pricing away from hours. If the value of an hour of expertise is under pressure, charging by the hour ties your income to the thing losing value. Price on outcomes where you can.
2. Own decisions, not just analysis. Analysis is what models do increasingly well. Being accountable for a recommendation, and for the result, is harder to substitute.
3. Use AI to raise output per person. In a flat-wage world, the people and firms that do better are the ones producing more with the same time. Integrating AI into existing workflows is how that happens in practice.
4. Revisit the plan every year. The model is explicitly a tool for comparing assumptions. As capability changes, so does which scenario looks closest.

The useful question is not whether AI will take your job by 2030. It is what you would need to change for your work to still be worth what you charge if your rate never rose again.