We initially had to strike a balance between letting users work quickly and efficiently with AI tooling and making sure we could do it securely and correctly. Without the right controls, users would either connect anything and everything to their AI tools, or we would have to restrict and block everything that wasn’t vetted and approved with the app owners—and that approach felt very all-or-nothing. It also meant we’d be relying on user permissions being accurate and appropriate, and trusting that any actions the AI tools took would be acceptable. Otherwise, we’d be forced to either reduce general user permissions across the board or accept the risk of accidental writes by the AI tooling, for example.

Runlayer let us connect various tools while still setting fine-grained permissions specifically for AI functionality, without affecting users’ day-to-day use of those tools. Users can still read, write, etc. as their regular user, but for key systems, Runlayer allows us to limit what AI can do for that same user; for instance, restricting AI usage to read-only. Review collected by and hosted on G2.com.