Benedict Evans Says 3 Tests Show AI Job Forecasts Are Mostly Impossible
Updated
Updated · ben-evans.com · May 24
Benedict Evans Says 3 Tests Show AI Job Forecasts Are Mostly Impossible
2 articles · Updated · ben-evans.com · May 24
Benedict Evans argues job-by-job models of AI exposure are largely unreliable, saying specific forecasts at this early stage of the technology will be right only by luck.
3 historical checks underpin his case: the CPA test, where decades of automation did not shrink accountants; the newspaper test, where the internet broke business models more than tasks; and the Uber test, where smartphones reshaped taxi work unexpectedly.
He says automation can expand demand rather than cut labor through Jevons paradox, while jobs also mutate faster than official categories can capture—keeping titles stable even as underlying work changes.
That makes tools such as O*NET too incomplete to quantify AI risk, he argues, because they miss tacit skills, business-model dependencies and second-order effects that determine which roles grow, shrink or disappear.
As AI automates entry-level tasks, how can the next generation even begin a career?
Companies are spending billions on AI but struggling to profit. What is the secret to making AI pay?
Why does history show that making work cheaper with technology often creates more jobs, not fewer?
Beyond Job Loss: How AI Transforms Work, Skills, and Policy in the 2026 Economy
Overview
This report highlights Benedict Evans's skepticism about current AI job forecasts, arguing that predictions about job loss or creation due to AI are often flawed and misleading. Evans points out that many models wrongly equate AI exposure with actual job displacement, leading to exaggerated conclusions. He emphasizes that the real impact of general-purpose technologies like AI is unpredictable and cannot be understood by simply looking at current tasks. Instead, these technologies drive complex changes, with people adapting, innovating, and creating new job categories. The report urges a more nuanced view, focusing on human adaptability and the unpredictable nature of technological change.