Skip to content

Scenarios

Upload the PDF, put it in a collection, give it a tag. Six months later, “what did it say about deadlines” is a question for the agent — it finds the fragment and shows you where it took it from.

Upload them all into one collection: the collection and tags are chosen once in the upload form rather than per document afterwards.

Money operations from mail and receipts are created by a separate path, in Finance. The documents here remain the primary source you can pull up and re-read.

Gather the documents into a collection with a name that means something. From there both search and the agent find them by content, not by whatever the sender called the file.

Images go through text recognition, so they’re found by content and not just by filename. A photo of the whiteboard after a session stops being dead weight in your gallery.

Finding something when you forgot the name

Section titled “Finding something when you forgot the name”

The list searches by title — enough when the filename is meaningful. When it isn’t, ask the agent: it searches by meaning. “There was a document about return conditions somewhere” is found even if the word “return” never appears.