JobBench: AI Assistants That Work With You, Not Against You
Researchers created JobBench, a new way to test AI assistants on tasks that actually help people. It focuses on real-world work needs instead of just replacing jobs. This could lead to AI tools that empower workers rather than replace them.
A team of researchers released JobBench, a new benchmark for testing AI assistants. Unlike current tests that focus on replacing human jobs, JobBench evaluates AI on tasks that experts say would be most helpful to delegate. It covers 130 tasks across 35 different occupations, using real-world work environments to test how well AI can assist.
Most AI benchmarks today measure how well AI can do a job instead of a human. JobBench flips this by testing how well AI can help humans with their work. Think of it like a personal assistant that helps you organize your files, draft emails, or analyze data—tasks that make your job easier, not replace it. This could lead to AI tools that feel more like helpful coworkers than competitors.
To see this in action, try using an AI assistant like Microsoft Copilot or Google Workspace's AI features. Ask it to help you organize a messy folder of files or draft a professional email. These tools are already using some of the principles behind JobBench to make AI more helpful in everyday work.