Few questions in economics get more attention right now than whether AI will make workers — and whole economies — more productive. The evidence so far points in two directions at once.

Evidence from individual tasks

Customer support. In a widely cited study of more than 5,000 support agents, Erik Brynjolfsson, Danielle Li and Lindsey Raymond found that access to an AI assistant raised the number of issues resolved per hour by about 14% on average, with the largest gains for newer, less experienced agents.

Writing tasks. An experiment by Shakked Noy and Whitney Zhang, published in Science in 2023, found that professionals using ChatGPT completed writing tasks faster and produced higher-rated output.

A pattern emerges: AI tends to help less experienced workers the most, narrowing skill gaps on the tasks it is good at.

Evidence from the economy as a whole

Macroeconomic estimates are more cautious. Economist Daron Acemoglu estimated in 2024 that AI might raise total factor productivity by only around 0.7% over ten years, because the share of tasks that AI can profitably automate in the near term is limited. Other forecasters are considerably more optimistic.

Why the numbers differ

  • Task vs. economy. A big gain on one task matters little if that task is a small share of total work.
  • Adoption takes time. Previous technologies, from electricity to computers, took decades to show up in productivity statistics.
  • Measurement. Quality improvements and time saved are hard to capture in GDP.

What to watch

New field studies across professions, company-level productivity data, and whether gains concentrate in some firms or spread across the economy.

Three key studies

  • Customer support: a study of more than 5,000 agents by Erik Brynjolfsson, Danielle Li and Lindsey Raymond found an AI assistant increased issues resolved per hour by about 14% on average, and by about 34% for novice and less-skilled workers.
  • Writing tasks: an experiment by Shakked Noy and Whitney Zhang, published in Science in 2023, found that access to ChatGPT reduced the time professionals took on writing tasks and improved quality as rated by evaluators.
  • Consulting: a field experiment with Boston Consulting Group consultants found large gains on tasks within AI’s capabilities, but worse performance on a task just outside them — a “jagged frontier” where AI helps on some tasks and hurts on others.

Why gains may take time to show in the economy

Economists often point to the productivity paradox: computers spread through workplaces in the 1970s and 1980s, but productivity statistics only clearly reflected their impact in the late 1990s. General-purpose technologies require complementary investments — new processes, skills, organizational changes and software — before their full benefits appear. AI may follow a similar pattern, with measurable economy-wide effects lagging behind individual task-level gains.

Who benefits

Early evidence suggests AI often helps less experienced workers most, narrowing skill gaps within jobs. But effects on wages, employment and inequality across the economy depend on how companies use the gains: to augment workers, cut costs, or create new products and services.

What it means for workers and managers

For individuals, the evidence suggests learning to use AI on suitable tasks is worthwhile — and so is recognizing where it fails. For managers, measuring results on real tasks, redesigning workflows rather than simply adding tools, and investing in training are more likely to deliver gains than mandates to “use AI” without guidance.

See also which jobs are most exposed to AI and skills to stay relevant in the age of AI.

Key takeaways

  • Task-level studies show significant productivity gains from AI in support, writing and consulting.
  • Gains are often largest for less experienced workers.
  • AI can hurt performance on tasks outside its capabilities — the “jagged frontier”.
  • Economy-wide productivity effects may lag due to the need for complementary investments.
  • Distributional effects depend on how organizations use AI.

Frequently asked questions

Has AI increased productivity in the whole economy?

Clear economy-wide effects are not yet established; most strong evidence comes from specific tasks and companies.

Why do productivity forecasts differ so much?

Economists make different assumptions about which tasks AI affects, how quickly it is adopted and how much it improves each task.

What is the productivity paradox?

The observation that new technologies can spread widely before their effects appear in productivity statistics.