The short answer
Every write action, such as pausing a campaign, changing a budget or publishing a tag, is proposed in chat and held until you approve it. The hold is enforced by the system, not just by the AI following instructions, so it applies even if someone asks it to skip confirmation.
In more detail
When you ask for a change, Markitech shows what it intends to do: which account, which campaign or tag, and the values before and after. You can ask questions, narrow the change or cancel it before anything runs.
Once you approve, Markitech executes exactly the change you reviewed and nothing else. You can then ask a read-only follow-up in the same conversation to confirm the result.
This split between reading and writing is what makes it practical to let an AI work in live accounts. You get the speed of an assistant that does the work, and the account stays under your control.
Approvals are specific. Approving a budget change on one campaign does not approve anything else, and a new request produces a new change list. That keeps every action traceable to a decision a person made.
Learn more: Approval-gated AI actions.
Related questions
- What can Markitech read without asking?
- How does Markitech store my credentials?
- Is it safe to let AI manage ad accounts?
- How much does Markitech cost?
- What are Markitech credits?
Key terms
Every change, approved by you
Markitech proposes each write action in chat, such as a budget change, a paused campaign or a Tag Manager publish, and holds it until you approve it. Reporting and lookups run without approval, so you can ask as many questions as you need.