Benchmarks for AI agents doing real work.
Yardstick tests agent teams on real tasks from real industries, with the approvals, controls and local rules that the work demands.
200tasks
20industries
3levels of agent autonomy
How it works
- 01
Specify
Describe a task and get a process map you can edit, step by step.
- 02
Run
Agents work in a sandbox of mock systems. People make the approval calls.
- 03
Measure
Every run is traced and scored on the outcome and on the controls it kept.
Industries
Bring a task from your industry.
We are working with practitioners, industry bodies and development partners to set the standard for agentic work.