Anthropic Models Three AI Scenarios for US Jobs, Wages and Growth by 2030

Imagine you’re a project manager in Maidstone, a solicitor in Canterbury, or an administrator at an NHS trust in Ashford. You spend your days doing exactly the kind of knowledge work — reading, writing, analysing, advising — that artificial intelligence is increasingly being trained to do. So when one of the world’s leading AI companies publishes a detailed economic model suggesting that, by 2030, AI could either gently nudge wages upward or send unemployment to historic highs, it’s the sort of thing worth paying attention to.

Anthropic, the AI safety and research company behind the Claude family of AI assistants, released its *Economic Scenarios for Radical AI* report on 9 September 2026, alongside a publicly accessible interactive tool that lets anyone plug in their own assumptions about AI and see what the numbers spit out. The report was produced by Anthropic’s economics team, part of the Anthropic Institute, and draws on a survey of 10,980 Americans conducted in August 2026.

The tool and the technical report are explicitly described as hypothetical projections, not forecasts. The aim is to give policymakers, businesses and ordinary people a clearer sense of what different paths of AI adoption could mean for jobs, wages, GDP and the share of income going to workers rather than capital.

Three Scenarios, Very Different Worlds

The model divides the economy using what economists call a task-based framework. Rather than treating a job as a single thing that either survives or disappears, it breaks roles down into individual tasks — some of which AI can already handle, some it might handle soon, and some it may never touch. The framework focuses especially on cognitive and knowledge work: management, professional services, sales and office roles. But it also tracks what happens to everyone else.

From there, Anthropic’s team maps out three scenarios to 2030.

The Modest scenario treats AI as roughly comparable in impact to the internet. US GDP in 2030 is projected at around $34.1 trillion (about £26.5 trillion at current rates), roughly 1.6% above what you’d expect without AI. Unemployment stays within historical ranges, and average wages nudge upward. Disruptive, yes — but manageable.

Meanwhile, the Major scenario is where things get more complicated. Here, AI is assumed to handle around half of all knowledge-work tasks by 2030. US GDP climbs to about $36.3 trillion (around £28.2 trillion), some 8.3% above the no-AI baseline. Unemployment edges up to between 4.6% and 5%, and wages for knowledge workers go flat — while workers in other occupations see modest gains.

Then there’s the Extreme scenario. In this version, AI outperforms humans at most knowledge tasks and begins improving its own capabilities in a self-reinforcing loop. GDP growth hits around 15% per year, with 2030 US GDP projected at about $44.4 trillion (around £34.5 trillion) — roughly 32% above the no-AI path, and the economy doubling around every four and a half years. But the cost is steep: historically high unemployment, and knowledge-worker wages falling by more than 10%.

The Trade-Off at the Heart of the Model

The pattern running through all three scenarios is hard to miss. Faster AI-driven growth comes hand in hand with greater disruption to jobs and wages — especially for the people currently doing knowledge work.

That’s a genuinely uncomfortable finding. The assumption that economic growth automatically lifts all boats gets tested here. In the extreme case, GDP could be surging while significant numbers of office workers, professionals and managers face falling pay or unemployment.

Some figures circulating on social media — including estimates of job displacement rates of 0.3% in the modest scenario, 2.5% in the Major, and 13.5% in the extreme — have not been fully verified against the primary technical report, and should be treated with some caution until confirmed.

Jack Clark, a prominent figure in AI policy and research with ties to Anthropic, has previously written about the difficulty of anticipating how quickly AI capabilities will develop and how institutions will respond. The honest answer, built into Anthropic’s own model, is that nobody knows which scenario we’re heading for.

What the Survey Reveals

One of the more unusual features of the tool is that it incorporates the views of real people. The 10,980 Americans surveyed in August 2026 were asked about their expectations for AI capabilities, how quickly they think adoption will happen, and what effects they anticipate on the labour market. Users of the interactive tool can compare their own assumptions directly with those survey responses — which gives a sense of where public expectations sit relative to the scenarios Anthropic has modelled.

It’s a clever design choice. Rather than presenting the model as a top-down expert pronouncement, it invites people to interrogate their own assumptions and see how different beliefs about AI lead to very different economic outcomes.

Reactions and Limitations

Critics have raised fair questions. Some argue the extreme scenario overstates the risk of unemployment; others suggest even the extreme case may underestimate disruption if AI development accelerates beyond current projections. Labour economists have noted that forecasting ground-breaking technologies has a poor track record — the error margins are large, and real-world adoption is shaped by regulation, culture, infrastructure and political choices as much as by raw capability.

There’s also the question of who gets left out of the numbers. Focusing on US aggregates can mask very different outcomes across regions, income groups and sectors. A model that shows average wages rising in the modest scenario doesn’t tell you much about what happens to a specific community where a particular industry dominates.

The model does not include UK-specific projections. It doesn’t draw on data from the Office for National Statistics or the Bank of England, and it makes no claims about regional labour markets outside the United States.

What This Means for Kent Residents

Anthropic’s model is built around US data and makes no direct claims about the UK or about Kent specifically. But the underlying dynamics — AI automating knowledge-work tasks in sectors like professional services, public administration, education and healthcare — are just as relevant here. Kent residents working in those fields, whether at NHS Kent and Medway ICB, Kent County Council, or in private professional services, may find the scenario framework a useful lens for thinking about how their own roles could change over the next few years. For anyone curious, the interactive tool is publicly available and worth exploring.