Managing Shadow AI

Section AI Webinar | September 2026

A useful reframing of shadow AI: employees using unapproved tools aren’t always a security problem. Their behavior can also reveal where approved tools, policies or processes aren’t keeping up with how people actually want to work.

My key takeaways:

• Treat shadow AI as a signal. Before asking “How do we stop this?” ask why employees are going outside the approved system.

• There are two costs. There’s the obvious security and compliance risk, but also an opportunity cost. When AI work is scattered across personal tools and accounts, organizations lose shared knowledge and the ability to build on what employees are learning.

• Use a tiered tool strategy. Provide a strong default AI platform for everyone, specialized tools where there’s a business case, and a controlled environment where power users can experiment.

• Make the approved path the easiest path. Slow tool approvals can actually encourage shadow AI. Section recommends a clear intake process and roughly a two-week decision cycle.

• Investigate before you punish. Unapproved usage may expose real risk, but it may also reveal a tool or capability employees need that the organization hasn’t provided.

The idea I’m keeping:
The goal isn’t to eliminate AI experimentation. It’s to create an environment where useful experimentation can happen safely and where what employees learn can benefit the larger organization.