AI’s business value comes down to two questions about each task: whether the work is getting cheaper to complete, and what happens to the time people save. Counting AI users and assistants shows how much AI is being used, and it can leave both unanswered.
A McKinsey report takes up the first question. It asks whether AI can perform or support a given task at acceptable quality and a sustainable total cost. Its AI value curve compares what a task costs when it is done the current way with the full cost of doing it with AI. That full cost includes the fees for the AI model and the human time spent reviewing, checking and redoing the work.
The curve then combines the saving per hour with the hours of work AI can take on. A small saving on one task can add up across many hours, while heavy review or rework can shrink the saving or erase it. On that basis, the report estimates the potential value in work people already do at $2.52 trillion a year worldwide.
This task-by-task view helps explain why the biggest gains can show up in well-paid professional work. Legal professionals, healthcare practitioners, managers, and computer and math workers are among the groups where AI adds the most value per hour. AI can support routine preparation and analysis in these jobs, which leaves more of a professional’s time for expertise and judgment. A little time freed in a highly paid role can be worth a lot. Even so, the curve counts only the tasks AI can support more cheaply, and never treats a whole job as work AI can take over.
McKinsey calls its estimates uncertain because they rest on assumptions. They leave out one-time spending on platforms, training and organizational change, and results vary from company to company with local wages and the mix of tasks.
The estimates also move with the cost of using AI, because the curve counts that cost. A higher cost can make tasks that were barely worth handing to AI more expensive than doing them the current way. In one illustration, that cost rises 30 percent, and the worldwide value estimate falls to $1.67 trillion a year.
