The Trust Score is the mechanism that determines how far the Marketing Agent can act on its own judgment. The more you use it, the wider the range of work you can hand over to the agent.
What is the Trust Score?
The Trust Score is a metric that sets the agent's level of autonomy for each operation category, so the higher the score, the more the agent can proceed without waiting for your approval.
Each time you choose "Execute" in response to the agent's proposal, the score goes up and more operations can be carried out automatically. When you choose "Not now", the score becomes harder to raise. This lets you start by reviewing each step and then widen the range you delegate once you feel you can trust the agent.
In the user-facing screen, the Trust Score is shown as "Trust in AI Agent".
The three levels of autonomy
As the Trust Score rises, the way the agent behaves changes across the following three levels.
| Autonomy | Agent's behavior | What you do |
|---|---|---|
| "Requires Approval" | Pauses before executing and asks for your approval | An approval request arrives in Chat. Nothing runs until you choose "Execute" |
| "Notification Only" | Executes automatically and notifies you once finished | No prior approval is needed. The result is reported to you in Chat |
| "Automatic" | Executes automatically, with no notification | Runs in the background, and you can review the result in the history |
Low-risk operations (such as data analysis or drafting content) proceed automatically regardless of the Trust Score.
Operation categories and how they progress
The Trust Score is managed independently for each operation category, so even if your "Draft Creation" score rises, it doesn't affect the autonomy level for "Segment Creation".
The operations the agent can perform are divided into the following categories.
| Category | Description |
|---|---|
| Draft Creation | Create a draft of a Message Broadcast |
| Segment Creation | Create a new Segment of recipients |
| Segment Edit | Change the conditions of an existing Segment |
| A/B Test Creation | Set up an A/B Test |
| Broadcast Schedule | Schedule the sending of a Message Broadcast |
| Broadcast Send | Deliver a message to your customers |
| Workflow Creation | Build a Workflow |
| Workflow Activation | Put a Workflow into a running state |
| Workflow Deactivation | Stop a running Workflow |
The more of a track record you build, the more each category's autonomy level rises step by step, from "Requires Approval" to "Notification Only", and then to "Automatic". The greater an operation's impact on your customers, the more of a track record it requires.
Operations that always require approval
The following three operations always require your approval, no matter how high the Trust Score gets.
| Operation | Reason |
|---|---|
| "Broadcast Schedule" | Because this operation schedules a message to your customers, it prevents unintended scheduling |
| "Broadcast Send" | Because this operation delivers a message to your customers, it prevents unintended sends |
| "Workflow Activation" | Because this operation puts automated processing into a running state, it prevents unintended starts |
These are shown on screen as "Always requires approval". No matter how much the agent is trusted, the final decision is always yours to make.
Checking the Trust Score
You can check each category's autonomy level in the "Trust in AI Agent" widget on the Dashboard.
The widget shows a list of operation categories and the current autonomy level for each one (Requires Approval / Notification Only / Automatic). Scheduling and sending broadcasts and activating Workflows always require approval.
Changing the autonomy level
The autonomy level rises automatically with your track record of approvals (as a guide: about 5 approvals for Notification Only, 20 for Automatic). The three always-approval operations aren't covered by this guide. There's currently no screen for changing it manually.
Each time you choose "Execute" for one of the agent's proposals, the Trust Score for that category rises. Start by reviewing each step; as your track record builds, the range the agent can handle automatically widens. If you want an operation you've delegated back on a confirmation basis, contact support.
When things don't work
Here are common situations around the Trust Score and how to handle them.
| Situation | How to handle it |
|---|---|
| The autonomy level isn't rising | You need to build up a track record of choosing "Execute" for the agent's proposals (as a guide: about 5 approvals for Notification Only, 20 for Automatic) |
| You want to put a delegated operation back to requiring confirmation | There's currently no screen for changing this manually. Contact support |
| You want to start the settings over | Resetting the Trust Score isn't currently available from the user screens. Contact support |