If you have used either, you already know how to use this. The conversation, the uploads, the model picker and the history all behave the way you expect.
| In ChatGPT / Claude | Here |
|---|---|
| Start a new chat | Start a new chat |
| Upload a file | Attach a file to the message |
| Choose GPT-5 / Claude | Choose a model from the picker |
| Custom instructions | Instructions inside a reusable assistant |
| Custom GPT / Project instructions | Reusable assistant |
| Chat history | Conversation history, searchable |
| Canvas / Artifacts | Artifacts panel |
It runs on Uni-edit's own Microsoft Azure account. Your conversations are not training data for anyone, are not visible to other companies, and are not visible to other staff. They are stored on our own server in Hong Kong and processed in the Asia-Pacific region.
Daniel can see usage totals — how many questions, which model, what it cost — as part of running the trial. He does not read your conversations.
If you have used ChatGPT Projects, that specific feature does not exist here. Conversations are organised individually.
What does the same job is a reusable assistant: a saved set of instructions you can start a conversation from, so you do not re-explain yourself every time.
This is the difference that will matter most, and it is the one thing this tool does less than you might expect, not more.
It has no access to the customer database, job records or expert information. "Show me this customer's job history" will not work.
That connection has been built and tested. It is not switched on because who may see which records has to be correct before it is turned on rather than after — Louise seeing Taiwan customers, Yuki seeing Japan customers, and nobody seeing fields their job does not need. You will be told when it changes.
It can be confidently wrong here exactly as it is anywhere else. Fluent, well-structured, plausible and incorrect. Being private to Uni-edit makes it safer with your data; it does not make it more accurate.