The Shadow AI Trail in Your Proxy Logs

Yesterday I shared how to spot unofficial AI usage by following the money trail.

Here's another technique that's equally revealing: your web proxy logs.

Search your proxy logs for POST requests to common AI endpoints like api.openai.com and claude.ai.

You might be surprised at what you find - each POST request represents someone exploring AI's potential.

The interesting part isn't just who, but when and how much. Are specific teams showing consistent usage patterns? Does traffic spike during certain projects? This kind of intelligence helps paint a picture of organic AI adoption across your company.

Here's the thing: Whether you discover AI usage through finance records or proxy logs, the goal is the same - understanding how your colleagues are innovating, and finding ways to support them safely.

Pro tip: Consider tracking AI endpoint traffic over time. When you can show leadership clear evidence of adoption, conversations about formal AI policies become much more practical.

What's working better for you - following the money trail or analyzing web traffic? I'm curious to hear about your findings.

Cheers, Craig