How to Review & Approve AI Agent Actions
When AI agents draft emails or take actions on your behalf, the approvals queue lets you review, edit, and approve before anything is sent. It now validates each action and flags sender or safety issues before you approve, plus a direct link to the conversation thread for context.
AI agents can draft emails, create tasks, and take actions on your behalf — but nothing goes out without your say-so. The Approvals page is where you review, edit, and approve or reject every AI-drafted action before it reaches your clients.
Open your approvals queue at https://app.trustpager.com/settings/api?tab=approvals.
The Approvals Queue
Each approval card shows what the AI wants to do (send an email, create a task, etc.), the drafted content, and which contact and opportunity it relates to.
Reviewing an Action
Click any pending approval to see the full details. If it's a drafted email, you can read the subject, body, and recipient. If it's a task or record update, you see exactly what will change. Everything is transparent — no black box.
Built-in Checks Before You Approve
As you review an action, the card validates it and flags anything worth a second look, so you are never approving blind.
- Who an email sends as: email approvals show the exact From address that will be used, with a confirmation of the sender. If a draft specifies a sender that is not connected to your workspace, the card flags it and disables Approve until it is resolved, instead of quietly sending from a different address.
- Safety and format checks: webhook destinations must be a secure public address; signing and form sends must have valid recipient details; and any action missing a required field is flagged inline.
A red note is a blocking issue: fix the flagged detail (or reject the action) before you can approve. Amber and green notes are advisory and do not block.
Viewing Conversation Context
When an AI-drafted email or SMS is a reply to an existing conversation, the approval card shows a direct link near the top so you can read the thread before deciding.
- Email replies — a "View thread: [subject]" link appears when the agent is replying within an existing email thread. It opens https://app.trustpager.com/inbox/email in a new tab showing that conversation.
- SMS replies — a "View SMS conversation with [number]" link appears whenever there's an existing conversation with that contact on that phone line. It opens https://app.trustpager.com/inbox/sms in a new tab.
The link only appears when there's an existing thread to view. First-ever messages to a new contact won't show one.
Editing Before Approving
You're not limited to approve-or-reject. If an AI-drafted email is 90% right but needs a tweak, edit the content directly in the approval view. Adjust the wording, fix a detail, or add a personal note — then approve the edited version.
Approving and Rejecting
Click Approve to let the action proceed — the email sends, the task is created, or the record is updated. Click Reject to cancel it. Either way, the action is logged in the audit trail so you can review patterns over time.
Tip: If you find yourself consistently editing the same type of AI output, update the agent's instructions or templates. The approval queue isn't just a safety net — it's a feedback loop for improving your AI agents.