Bold presumption that businesses have any useful way to evaluate agent fitness. Hell, they struggle to evaluate human fitness and do basic things like plan and execute OKRs. What makes you think they’d be any good at continuous quality improvement on entities that can’t correctly explain their own reasoning?