AI Actions & Approvals
AI Actions are tasks the AI suggests or performs — sending follow-ups, updating contacts, creating deals, routing intake, etc. Medium/high-risk actions queue up "pending approval"; low-risk actions auto-approve and run immediately. Only OWNER and ADMIN can view and act on the audit log (the page exposes proprietary pipeline signals like model tiers and prompt shapes — it's intentionally restricted).
Steps
- Go to AI Actions under Tools > AI Intelligence (OWNER/ADMIN only — the sidebar item is hidden for TEAM and below).
- Review pending actions — each row shows action type, risk level, model tier ("Tier 1/2/3"), and a friendly summary.
- Approve or reject each action. Rejected actions require a reason; approved actions execute automatically.
- Watch the bell icon in the header — when a team member triggers an AI action that needs approval, every OWNER/ADMIN gets a notification with a link to this page.
Good to know
- High-risk actions (sending external messages, making bookings, changing pricing) always require approval. Medium-risk surfaces here too. Low-risk (tag changes, internal notes) auto-approve.
- The bell badge reflects pending-approval AI actions for your role. If you're OWNER/ADMIN and see a number, someone's waiting on you. Regular team members never see the badge because they don't have approval authority.
- Expanding any row shows a plain-English summary instead of the raw payload: "What AI did" (action label + contact) and "Result" (status reframed: awaiting approval, auto-approved (low risk), completed, rejected with reason). The previous behavior — JSON dumps of inputContext / outputContent — was meaningless to non-Super-Admins because the payload always came back as { redacted: true }.
- CSV export is gated to OWNER/ADMIN as well and the payloads stay redacted in the export.
Example: After 3 days of activity the founder (OWNER) opens AI Actions: bell showed "4 pending". Her sales rep triggered 3 follow-up drafts on warm leads (high-risk → review one by one) and a customer-tag change (auto-approved earlier, visible under "Auto-approved" tab). She approves each draft, adds feedback stars, and moves on. The rep got a "completed" notification once each approved action ran.
Troubleshooting
AI Actions shows "failed"
The AI attempted an action it couldn't complete (e.g., moving a contact to a non-existent pipeline stage).
Fix: AI Actions > click the failed action — the error message explains what went wrong. Common fix: confirm the entity (contact, pipeline, stage) the AI referenced actually exists.
I don't see AI Actions in my sidebar
The AI Actions audit log is gated to OWNER and ADMIN roles. TEAM / DIRECTOR / EXECUTIVE / BOARD / VIEWER don't see the sidebar entry.
Fix: If you need approval authority, ask an OWNER to change your role under Settings > Account > Organization Members. If you just want to see your own AI activity, that surfaces in the contact activity feed and the Notifications bell.
Bell shows 0 pending but I know someone triggered an action
Either the action was low-risk and auto-approved (check the Auto-approved tab), or you're not OWNER/ADMIN and the notification went to someone else.
Fix: Filter to "All" tab and sort by date. If the action auto-approved, risk level was "low". Approval-pending notifications fan out to OWNER + ADMIN only.