Artificial intelligence could add significant momentum to the US economy by 2030. But the strongest growth scenario could also leave almost one in five knowledge workers unemployed, according to a new economic model published by Anthropic.

The company behind Claude has released a technical paper and an interactive tool examining how AI adoption might affect productivity, employment, wages and the distribution of income.

The researchers modelled the economy as a collection of tasks that AI could leave unchanged, support, automate completely or create from scratch. They then explored how different combinations of technological capability and adoption could influence the US economy.

Anthropic emphasises that the scenarios are not forecasts and that no probabilities have been assigned to them. Instead, they illustrate a wide range of possible outcomes, from limited economic impact to a period of growth without a clear historical precedent.

Three Possible Economic Scenarios

In Anthropic’s modest scenario, AI remains a relatively limited technology from an economic perspective.

By 2030, GDP would be approximately 1.6% higher than it would have been without AI, while annual economic growth would reach 2.4%. Employment in cognitive occupations, including management, professional services, sales and office roles, would decline by only 0.5%, with little effect on overall unemployment.

The substantial scenario assumes AI becomes capable of performing approximately half of all knowledge-work tasks, although most would still be completed without AI assistance.

Under these conditions, annual growth would reach 5.4%, roughly double the economy’s normal rate. GDP would be 8.3% higher by 2030, while cognitive employment would fall by 3.9%.

Unemployment among knowledge workers would rise to 4.5%. Wages would also begin to diverge: compensation for cognitive work would decrease slightly, while wages in other areas of the economy would rise by almost 6%.

Unprecedented Growth, Accompanied by Severe Job Losses

Anthropic’s extreme scenario describes a much more dramatic transformation.

Annual economic growth would reach 15.4%, placing GDP 32.4% above the trajectory expected without AI. At that pace, the economy could double in size approximately every four and a half years.

The labour-market consequences, however, would be substantial.

Employment in cognitive occupations would decline by 21.5%, while unemployment among those workers would reach 17.9%. Across the entire workforce, unemployment would rise to 11.9%, exceeding the levels generally associated with a typical recession.

Wages would also move in opposite directions. Knowledge workers could see earnings fall by 11.5%, while wages for other categories of workers could increase by 33.6%.

The model suggests that AI-driven growth could therefore create a considerably wealthier economy without distributing the benefits evenly.

Labour’s share of national income would fall from 60% to 45.2%, while income generated by capital would increase by more than 80%. In practical terms, a larger portion of the economic gains would flow towards companies, investors and asset owners rather than employees.

Public Expectations Are Closer to the Middle Scenario

Anthropic accompanied its economic model with a Morning Consult survey of US adults conducted in August.

The expectations of the median respondent were closest to the substantial scenario. This would translate into an economy approximately 8% larger by 2030, accompanied by a decline of around 4% in cognitive employment.

Anthropic CEO Dario Amodei has previously presented a more disruptive outlook. In May 2025, he warned that as many as half of entry-level office jobs could disappear within five years, potentially pushing unemployment to between 10% and 20%.

Those estimates align more closely with Anthropic’s extreme scenario than with the expectations expressed by the public.

Adoption Could Matter More Than Capability

The economic impact of AI will depend not only on what the technology can do, but also on how quickly organisations integrate it into real workflows.

Anton Korinek, who leads Anthropic’s research into the economics of transformative AI, noted that even highly capable systems will have limited economic influence if businesses and individuals do not use them.

Anthropic co-founder Jack Clark has expressed a similar view. While he expects AI capabilities to advance rapidly, he believes the diffusion of the technology across organisations could be slower and more difficult than many anticipate.

Implementation involves more than access to powerful models. Companies must redesign processes, connect AI systems to reliable data, establish governance frameworks, train employees and determine where automation produces measurable value.

For business and technology leaders, the rate of organisational adoption may therefore become just as important as the pace of model development.

The Risks Outside the Economic Model

Anthropic’s scenarios assume that the economy continues to function normally throughout the AI transition. The model does not examine outcomes in which advanced systems create broader safety, security or societal failures.

That omission has attracted attention as concerns about the direction of AI development continue to grow within the industry.

Jacob Coxon, a former researcher at OpenAI and Anthropic, recently resigned and publicly criticised both companies’ approaches to advanced AI. He argued that the industry is moving towards increasingly autonomous and self-improving systems without sufficient accountability or public oversight.

These concerns are reinforced by incidents in which AI models have reportedly gained unauthorised access to real organisational systems. Although such events do not fit easily into a conventional economic model, they remain relevant when evaluating the complete impact of large-scale AI adoption.

What This Means for Technology Leaders

Anthropic’s research highlights a tension that organisations can no longer treat as theoretical: the same technology that increases productivity and accelerates economic growth can also disrupt jobs, wages and the distribution of value.

CTOs and business leaders should prepare for multiple outcomes rather than building strategies around a single prediction. This means identifying tasks suitable for augmentation or automation, investing in workforce development, introducing clear governance and measuring how AI affects both operational performance and people.

The central question is no longer simply how capable AI will become. It is how organisations will deploy it, how quickly they will adapt and how the resulting benefits will be shared.

At Control F5 Software, we help companies move from AI experimentation to secure, practical implementation by connecting technology with real business processes, reliable data and measurable objectives. The organisations best positioned for the next phase will be those that treat AI adoption as a strategic transformation, not merely a software upgrade.

Source