All terms
Evaluation
Knowledge Work
A capability category measuring whether AI can do the professional office work people are paid for, like drafting and analysis.
Definition
Knowledge work, as a capability category on AI model comparisons, refers to the professional, information-based tasks people are paid to do — writing reports and legal briefs, analyzing data, building spreadsheets, preparing plans, and similar office work. Measuring it asks not just whether a model can hold a conversation, but whether it can produce the actual deliverables of real jobs at a useful standard. Benchmarks like OpenAI's GDPval score models on exactly these tasks, drawn from many occupations, which is why 'knowledge work' now shows up as a headline axis when labs compare their models.