By Russell Brandom
Publication Date: 2026-01-22 21:42:00
It’s been nearly two years since Microsoft CEO Satya Nadella predicted that AI would replace knowledge work — the white-collar jobs of lawyers, investment bankers, librarians, accountants, IT workers and others.
But despite the enormous progress that foundation models have made, change in knowledge work is progressing slowly. Models are adept at in-depth research and agent planning, but for whatever reason, most office work has remained relatively unaffected.
It’s one of the biggest mysteries in AI – and thanks to new research from training data giant Mercor, we’re finally getting some answers.
The new study examines how leading AI models perform when performing actual office tasks in consulting, investment banking and law. The result is a new benchmark called APEX-Agents – and so far every AI laboratory has received a “fail” grade. Faced with requests from real professionals, even the best models had difficulty answering more than a quarter of the questions…
