Back to the trust center

Decisions, AI and audit

Review and approve AI actions with a complete change history

See who requested each AI action, who approves it, which permissions apply, and what AgentticCRM records afterward.

Important AI actions require approval from the responsible person

Every AI proposal shows who requested it and its current status. When it is approved, permissions are checked again. Connected tools can perform only authorized actions, and every important change is recorded.

Trust topic

Review each AI action from request through history

Who requested each AI proposal, which permissions apply, what connected tools can do, and what an administrator can review afterward.

How do you know somebody answers for this?

What the Assistant does shows up as a proposal, with its status and its buttons. If it has consequences, it waits for a person to look at it and decide.

How it protects your team

On screen you see the proposal, what state it is in, and the review and approve buttons, before anything happens.

What to confirm

The reviewer remains responsible for evaluating and approving the action correctly.

Are permissions checked again before execution?

When it runs, who you are and what you can do today gets checked again. A proposal approved yesterday is not a permanent permission if someone’s role has changed since.

How it protects your team

Running it checks the identity and the permissions of right now, not of when it was approved.

What to confirm

Every new Assistant action has to go through the same path before it becomes available.

What can a connected tool do?

External agents connect through MCP and operate through explicitly allowed actions rather than receiving unrestricted application or database access.

How it protects your team

Every connection gets validated and limited to specific actions, instead of being handed open access.

What to confirm

Every action from a connected tool checks permissions again and validates the request.

What stays written down after somebody does something that matters?

Who did it, what they did, to what, and when. Anyone with permission can look it up in your company’s history.

How it protects your team

The history shows who acted, what changed, on which record, and when.

What to confirm

History coverage varies by action. During evaluation, confirm which events your organization needs to record.

Questions to ask in a demo

Questions for reviewing important AI decisions

Take an action that genuinely matters in your company and look at: whose proposal it is, who approves it, what happens if you change a permission, and what stays written down.

  1. Which important actions require approval first?

    Find the exact moment something changes, gets sent, costs money or touches the client. Then decide who approves it.

  2. Which permissions are checked before applying the change?

    Try a proposal before and after changing the approver’s permission. Check that they are two different things.

  3. What can whoever administers review afterwards?

    Look at the proposal’s status, who it is, who approved it, what happened when it ran, and what ended up in the history.

Trust priorities

Review the controls your business needs

Explore data separation, access controls, AI approval, change history, recovery, and deletion. Each section includes questions to use in a demo.

See how a person controls important AI actions

We’ll show how an AI action is prepared, approved, checked against permissions, and recorded.

Prefer to explore on your own? work out your current costsSee how the Assistant worksuse the CRM evaluation guide