What agents probably can do faster than us is to detect, combine and sequence vulnerabilities identified into a more effective or damaging attack. Ethical questions around how they are trained, powered and cooled notwithstanding, there may be opportunities to use them similarly in defence.
