How-ToAI AgentsOctober 22, 2026

Air Canada's approach to testing AI agents with no benchmark

Read original source →youtube.com

Nathanael Weill, who leads an AI agent team at Air Canada, works backward from a known answer: generate the input from it, then check whether the agent recovers that answer. The method targets tasks too specific for public benchmarks, where domain experts have little time to grade outputs.

People · Nathanael Weill

1 source

More stories today

Open the live feed