AI actions: what the assistant does itself
Learn which actions the AI assistant carries out itself, why it writes emails but never sends them, and how to recognize tools that are running.
The AI assistant in Univents does more than answer questions, it also works inside your workspace: it creates tasks, edits events, drafts documents. One thing it deliberately does not do — send freely written emails. It writes them for you; you send them.

Email: the assistant drafts, you send
For freely written emails the assistant deliberately has no tool. This covers:
- replying to an email in your mailbox,
- a new email to any address,
- the reply in an event's partner chat, which reaches the partner as an email,
- shift reminders and replacement requests to your staff.
If you ask for one anyway, you don't get an error — you get the finished draft: subject and body right in the chat, plus one sentence telling you where to send it from. You read it, change whatever you want to change, and send it yourself — from your mailbox, from the event's partner chat, or from the shifts tab.
The reason is simple: an email that has gone out cannot be pulled back. Whoever sends it should have read it first.
Your documents are not affected by this. Sending a quote or an invoice to a customer is a fixed action with a fixed recipient and your own template — the assistant starts that on your instruction. What it does not send is freely written prose.
What the assistant carries out directly
Read-only tasks, such as "show me the open quotes", the assistant handles right away. But it also changes data: creating tasks, editing events, adding line items, recording an incoming payment, drafting a quote. What it does is driven entirely by your instruction in the chat.
That makes two habits worthwhile:
- Be specific. "Create a task order drinks for the Meier wedding on 12 Sept" is unambiguous; "take care of the wedding" is not.
- Check the result briefly. The assistant only reports back what an action actually returned. After several changes in a row it pays to glance at the list or the document — that is where the real state is.
Whatever the assistant created or changed in your workspace can be edited or deleted in the same place as everything else.
How to recognize running tools
While the assistant is working, it shows small status pills in the chat so you can read off the progress of each step:
- A pill with a spinning symbol and a tool icon means the action is currently running.
- A green checkmark shows that the step finished successfully.
- A red cross signals an error. Hover over the pill to read the error message, and click Try again to repeat the step.
This way you can always see whether the assistant is still working, has finished, or whether something went wrong.
Tips
- Do ask the assistant for email copy — a rejection, a follow-up, a date change. That is exactly what it is for; only the sending stays with you.
- Tell it what tone the email should have and which key facts belong in it. That saves you a second round.
- For bigger changes, work in two steps: have it create things, check briefly, then carry on.
- If an answer looks odd, ask in the chat what exactly was carried out. The assistant only names what really happened.