Trust and security
Look up where CAB processes your data and what you control.
Core CAB: processing in your org
Section titled “Core CAB: processing in your org”Audiences, runs, results, snapshots, outcomes, the Filter Library and the run history are records in your Salesforce org. CAB builds and runs audiences inside the org. Its audience queries run with the object and field access of the user who runs them. A scheduled run uses the sharing context your administrator sets under Which sharing context should be used during scheduled audience runs?. See CAB Settings.
Who may see an audience follows Salesforce sharing. See Set sharing for audiences. Who may use which part of CAB follows its permission set groups. See Permission sets.
What leaves the org
Section titled “What leaves the org”CAB calls these services only. Each one runs only when you turn on the feature that needs it.
| Destination | Used for | What CAB sends |
|---|---|---|
Cabby service, https://api.campaign-audience-builder.com | Cabby | See What Cabby receives. |
Account Engagement, https://pi.pardot.com | Adding audiences to Account Engagement lists | For each person: email address, first and last name, Salesforce ID, and the campaign. |
Salesforce sign-in, https://login.salesforce.com | Connecting a user to Account Engagement | The OAuth sign-in. |
Your Marketing Cloud tenant, <subdomain>.auth, .rest and .soap.marketingcloudapis.com | Data extension adds | The fields you map to the data extension’s columns. |
Account Engagement and Marketing Cloud are your own Salesforce tenants. See Integrations.
What Cabby receives
Section titled “What Cabby receives”Cabby is off when CAB is installed. An administrator activates it and chooses where AI requests are processed. See Cabby settings.
With every request, Cabby receives:
- the conversation: the prompts, earlier messages in it, and the user’s Salesforce language;
- your organization’s custom instructions;
- the audience it works on and recent audiences as examples;
- the Campaign context when someone starts from a Campaign record. See Cabby settings.
Schema statistics
Section titled “Schema statistics”When Enable Metadata Statistics is on, a nightly job sends object and field names, record counts and how often fields are filled. It sends no field values.
Field values Cabby looks up
Section titled “Field values Cabby looks up”To write real values into a filter instead of guessing, Cabby can look up values while it works:
- the active values of a picklist;
- names and IDs of records that match a search term the user gave, for example to find one Account;
- the existing values of a short text or picklist field with at most 50 different values. Email, phone, name and ID fields are never sampled.
These lookups run with the user’s own access, so Cabby gets only what that user may see.
Data you choose to share
Section titled “Data you choose to share”Outcome data, Filter Library entries and place names for geolocation are shared only when an administrator turns each on. What each includes is in Cabby settings.
Deleting Cabby’s data
Section titled “Deleting Cabby’s data”Deactivate & Erase Data deletes what the Cabby service stores for your org. It doesn’t touch records in Salesforce. See Cabby settings.
Support access
Section titled “Support access”To look into your org, the CAB support team needs login access that a user grants for a limited time. See Grant login access to support.

