Skip to main content
Connect AI controls access with two kinds of rules. Some govern what a user can administer, and others govern what data a user can reach. This page explains how roles and policies fit together and how to move from access roles to policies.

Administrative Access and Data Access

Connect AI separates the following two concepts:
  • Administrative access–Determines which actions a user can take, such as managing users, workspaces, or connections. System roles control administrative access.
  • Data access–Determines which data a user can query or change, and which stored procedures a user can run. Access roles and policies control data access.
The following diagram shows how the pieces fit together.

System Roles

Each user has one system role per workspace. System roles are fixed and cannot be customized. The available system roles are:
  • Administrator
  • Workspace Administrator
  • Connection Administrator
  • User Administrator
  • Query
The Query role can access only the data it has been granted. See Roles for a description of each system role and the System Roles Matrix for what each one can do.

Access Roles

Access roles are being replaced by policies, which are now the recommended way to grant permissions to data.
Access roles grant data permissions inside a workspace. An Administrator creates an access role and grants any of the following, per connection or virtual schema:
  • Select, Insert, Update, and Delete on tables and views.
  • Execute on stored procedures.
Access roles are additive, so a user receives the combined permissions of every access role assigned to them. You can assign access roles directly to users or through SCIM groups. An access role is a fixed bundle of permissions that you cannot tailor to individual objects as precisely as a policy.

Policies

A policy is a named set of data permissions that you can turn on or off. Policies work as follows:
  • Grant-only–A policy grants access. Anything a policy does not grant is denied. There is no explicit deny.
  • Granular–You can scope a policy to individual tables, views, and stored procedures in a connection or virtual schema, and choose the specific access each one allows.
  • One workspace per policy–You create a policy in a workspace, and it applies only to that workspace.
  • Active or inactive–Setting a policy to Inactive removes the access it grants. Policy changes take effect within about a minute.
  • Audited–Every policy change is recorded in the Audit Log.
  • Administrator managed–Only Administrators can create and modify policies.
You can assign a policy to individual users, SCIM groups (if SCIM groups are provisioned), or system roles other than Administrator. Administrator already has full access. See Policies to create and manage policies.

How Roles and Policies Work Together

Roles and policies coexist. When a user belongs to both, Connect AI allows an action if any role or policy permits it. A policy cannot remove access that an access role or an administrator role already grants.
Data security rules, such as PII rules, apply to all connectors regardless of who runs the query. They are separate from roles and policies. See Data Security for more information.

Move from Access Roles to Policies

Connect AI does not migrate access roles to policies automatically. To move to policies:
1
Review the access roles in your account and note the permissions each one grants.
2
Create a policy that grants the same access, and assign it to the same users, SCIM groups, or system roles.
3
Run the access role and the policy in parallel and confirm that users can still reach the data they need.
4
Remove the permissions from the access role.
Because permissions from roles and policies combine, running both in parallel does not interrupt access.
Last modified on November 17, 2026