Signal Agentic AI already raising operational risk before its work share arrives

Summary

By 2028, AI agents acting on their own may perform up to 15% of daily work, in tasks ranging from cybersecurity to real-time systems monitoring. Amazon Web Services has started to map the risk of agentic AI according to how much agency a system has. Its categories are none, prescribed, supervised and full. Risks are already materialising, as one survey found that 80% of organisations have seen operational risk rise because of AI agents. The problems reported range from unauthorised access to exposure of data.