Skip to content
Assistant & search

Ask the assistant

What it knows, where it reads from, and why it searches on every message.

Last updated 1 min read

Three sources, one answer

The assistant reads from three places: indexed documents, feeds and notes, and the structure of the records themselves — which client has which engagement at which stage.

Retrieval runs again on every message. So an answer is given against today's data, not against something read once at the start of the conversation.

Your conversations are personal. A colleague sees neither your questions nor the answers.

A document question has two phases

When the question is about document content, the assistant first looks at the document's structure — its headings and page spans — and then opens only the pages it needs.

That is what makes it possible to ask about a two-hundred-page document without reading all of it every time.

If the assistant ends up opening no page at all, its answer is not used. An answer that did not come out of the document has no citation and nothing to stand on.

The assistant sees with your eyes

Answers are built from what you are allowed to read — not from everything in the workspace.

A restricted document, or a meeting whose access list you are not on, does not exist for the assistant either.

Still stuck?

If the answer wasn't here, write to us. It is a real inbox and it gets answered.

Talk to us