Four futures for AI and work—and how to prepare for more than one

What changed

The Conference Board’s September 15 release outlines gradual augmentation, concentrated gains, massive displacement, and uneven disruption. It also notes that broad employment and wage effects remain difficult to measure. These scenarios frame different possibilities; they are not four measured states of the economy.

Sources: The Conference Board

How the frameworks differ

Anthropic’s September scenario explorer models a US economy of tasks, varying augmentation, automation, productivity, and adjustment assumptions. Some scenarios combine a larger economy with worse outcomes for knowledge workers. The model omits forces including policy responses and business cycles, and is not an individual career forecast.

Our comparison: the Conference Board provides a vocabulary for discussing alternative paths; Anthropic provides an explicit model for exploring selected mechanisms. A label from one framework should not be assigned a numerical probability or GDP estimate from the other. Their categories are not interchangeable.

Sources: The Conference Board; Anthropic

Our interpretation: build a plan with branches

A career plan can survive uncertainty if it names the observations that would change it. Suppose an analyst is considering an expensive qualification. Before committing, they could define a small set of target roles and check whether employers repeatedly request it. They could also compare the qualification with a shorter project that demonstrates relevant analytical judgment. The decision should respond to that evidence, not the most dramatic scenario headline.

For a manager, a useful exercise is to write two versions of the next hiring plan. In one, demand grows as routine work becomes faster. In the other, demand stays flat and review work absorbs some of the saved time. Identify which responsibilities exist in both versions, and which investment becomes difficult to reverse. This is a planning exercise, not an assertion that either outcome is likely.

Workers and employers also need different decision thresholds. A company may be able to run several experiments at once; an individual has a limited training budget and living expenses. Advice to retrain should therefore name prerequisites, time, local hiring evidence, and the cost of being wrong.

What to watch next

Start a monthly record for the role or team you care about: openings, required skills, offered pay where available, workload, and time spent correcting AI-assisted work. Keep the sample consistent. Record changes in responsibilities alongside changes in headcount. Decide in advance which signal would justify another training project, a wider job search, or a revised staffing plan.

Use our occupation pages to frame those questions, then compare them with your own situation. An exposure score can start an investigation; it cannot settle a personal decision about leaving a job or changing professions.