Speaker: Martin Kylian

Most organizations have one of two problems with enterprise AI. Either they have built an approved AI architecture — a gateway, a sanctioned model, a policy — and have no way of telling whether anyone actually uses it or quietly routes around it. Or they haven't built anything yet, and AI adoption is happening regardless, one browser tab at a time.

This session is a practical guide to answering both questions from network traffic alone — no agents, no browser extensions, no changes to endpoints. We look at the signals that reveal a bypassed AI architecture, how prompts and sensitive data can be inspected and redacted in real time before they leave the network, and how raw traffic becomes an AI inventory classified by business intent: summarization ...