Yes, you can. This page exists because an earlier version of this manual said the opposite, and some of you read it.
The first version of this system checked every request before it ran, and handed anything that looked like customer work to a project manager instead of answering. If you saw a note saying your request had been "routed for review", that was it.
That check is not in the portal you are using. It was part of a different application, and the portal at https://chat.uniedit.org does not go through it. Nothing you type is inspected, classified, blocked or handed to anyone else. You type, it answers.
The rule changed too, not just the mechanism. Working over customer material with AI assistance is now permitted, deliberately and on the record — including a customer's actual manuscript.
There is no software boundary any more. There are two human ones, and they matter more than the old gate did because nothing enforces them.
Anything you send out has your name on it, not the assistant's. There is no approval step between what the assistant writes and what reaches a customer — that step was removed on purpose, which means you are the last person who reads it before they do.
The assistant is fluent and sometimes wrong, and those two things look identical on the page. A drafted reply can be perfectly written and contain the wrong deadline, the wrong price, or the wrong customer's name.
Read it the way you would read a draft from a capable new colleague: worth having, not yet trusted.
Using the assistant to help you do your work is what it is for — understanding a manuscript, drafting a reply, checking a glossary, working out how to explain something to an author.
It is not a replacement for the editing and translation services that produce customer deliverables. Those have their own experts, quality controls and records, and a job that belongs there still goes there. The assistant helps you handle the job; it does not become the job.
Worth knowing, since you are now putting real customer material in — see What you can put into it for the full picture. In short: our own Azure account, processed in the Asia-Pacific region, stored on our own server, not sent to any public chatbot service and not used to train anything.
The old manual had a page telling you what to do when a colleague's request was routed to you. Nothing routes to you any more. No queue, no records arriving for review, no decisions waiting. If you were expecting them, that is why none came.
What is worth your attention instead:
If you think the boundary is in the wrong place for some class of work, say so — that is a change to the policy, and the policy is written down and can be changed. It is not something to work around quietly.