How to Review & Approve AI Agent Actions

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, with 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.

Who Gets the Approval (Approval Routing)

When an action needs approval, TrustPager works out who should review it using the Approval Routing card at https://app.trustpager.com/settings/auto?section=approvals. The card lists a set of rules from top to bottom. The first rule that finds a person gets the approval, and the rest are skipped. The rules are:

• By Pipeline
• Opportunity Assignee
• Self Review
• Specific User
• Mailbox Owner (optional, see below)

Drag the rules into the order you want.

Mailbox Owner

Use Mailbox Owner when auto-responder drafts are sent in someone's name and you want that person to review them. For an auto-responder email reply (and any action the auto-responder queues for approval), the approval goes to the team member whose mailbox the reply is sent from, meaning the person who connected that Gmail account, as long as they are still a member of your workspace.

It only applies to auto-responder email approvals. For everything else (API keys, teammates, SMS auto-responders and TrustPager Mail addresses) it is skipped, so the next rule in your list runs instead.

It is not in the default chain. To use it, click + Mailbox Owner on the Approval Routing card, then drag it to the top so it runs first. For how auto-responders work, see https://trustpager.com/help-center/ai-auto-responders.

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.