Ai_gtm

How do we safely connect an AI assistant to our CRM?

Give the model its own scoped service identity — a licensed integration user with explicit object and field permissions — rather than letting it inherit a person's access. That way permissions are auditable and revocable in one place, which is the specific control the mature implementations have.

About 40% of the panel now has a model connected directly to a CRM or warehouse through a tool-calling protocol, and the constraint that bites is governance, not capability: 30% show shadow assistants bought by individual reps or projects blocked for weeks on admin rights. Decide the identity model before the pilot, not after.

Evidence

What this is based on

Every claim above comes from work LeanScale published. These are the sources.

More

More on ai_gtm

All questions in this cluster →