What an AI employee can do on its own
An AI employee’s actions, from replying to a client to tasks and data records, follow one rule: propose, approve, perform, log.
Updated October 9, 2026
An AI employee doesn’t just reply. Connected to tools, it works with data and handles tasks — but only within the permissions you have given it.
What an AI employee can do
- Lead conversations and reply in messengers, the website chat and email.
- Work with attachments and documents, and prepare documents.
- Reply in different languages.
- Record client data and add contacts.
- Quote prices and availability from the catalog and add a sale to a conversation.
- Create tasks and hand work over to a person.
- Use call transcripts.
- Read analytics to answer questions about the figures.
How an action is performed
- The AI employee proposes an action.
- A person approves it — or the system does, if the action is safe and the employee is allowed to perform such actions on its own.
- The action is performed.
- The activity log records what it was before and after, who approved it and how it ended.
Sensitive actions
Deletion, bulk changes and anything you’ve marked With approval are performed only after a person decides. Anything marked Not allowed, the employee never does.