Speaker: Dmitry Kharchenko

Every team wants AI agents: sales in the CRM, finance on invoices, support on tickets. The question for security and business leaders is simple: can we say yes without losing control of what these agents actually do?

Most companies pick one of two bad options. Say no, and employees use AI tools on their own, connected to company data with nobody watching. Say yes, and agents inherit the access of the people who set them up, often far more than they need. Either way, the first time anyone looks closely is after something has gone wrong.

This session shows a third way: make every agent action enforceable. Every agent gets an owner, a purpose and a clear limit on what it may do, and every action is checked against policy at the mo ...