Generative AI in Live Operations: Evidence of Productivity Gains in Cybersecurity and Endpoint Management

Abstract

We measure the association between generative AI (GAI) tool adoption and four metrics spanning security operations, information protection, and endpoint management: 1) number of security alerts per incident, 2) probability of security incident reopenings, 3) time to classify a data loss prevention alert, and 4) time to resolve device policy conflicts. We find that GAI is associated with robust and statistically and practically significant improvements in the four metrics. Although unobserved confounders inhibit causal identification, these results are among the first to use observational data from live operations to investigate the relationship between GAI adoption and security operations, data loss prevention, and device policy management.

Tasks

Reproductions