1 articles with this tag
Michele Grant told AI Engineer trust must be engineered before agents touch core workflows, with baselines, owners and auditable proof.