Place documents with AI

Placing files one by one is fine for a handful of documents. For a whole upload, let the app do the first pass: it reads each unplaced Workspace file, compares it with the folders and placeholders in your Dataroom and their placement instructions, and proposes a slot for every file that clearly fits. You then review the proposals in one place. Nothing is placed until you accept a suggestion, and a file the AI is unsure about is simply left in the Workspace for you.

Before you start

• The Dataroom has a structure to place into: folders and, ideally, placeholders with a short Placement instruction each. The instructions are what the AI matches against, so Signed board minutes, one file per meeting produces better suggestions than an empty field. See Build the Dataroom structure.
• The files are uploaded and show Ready in the Workspace.
• The data room is in Preparation, or you are adding files to a published room whose structure already contains the right slots.

Steps

1. Start the AI in one of two ways. If you are uploading now, choose Place with AI in the upload dialog: the files are uploaded to the Workspace and, as soon as each one has been processed, its placement suggestion is created automatically, with nothing further to run. For files that are already in the Workspace, open the ⋮ menu next to Upload on the Dataroom tab and select Scan Workspace. Both are offered only once the Dataroom has a structure, because that is what the AI matches the files against.
   Scan Workspace looks at every unplaced file; files that are already in the Dataroom are left alone.11.Menu next to Upload22.Scan Workspace
2. Wait for the scan. It runs as a job in the background and reports its progress at the bottom of the page; a large batch can take a few minutes. When it finishes, the Review suggestions button at the top shows a red dot, and in the Workspace each file with a proposal shows a marker in its Index column with a tick to accept and a cross to reject right there. The filter next to the search field narrows the list to Pending suggestion, so you can see at a glance what the AI proposed and what it skipped.
   A suggestion is a proposal, not a placement: the file is still only in the Workspace until you accept it.11.Workspace filter22.Index column with the suggestion33.Review suggestions
3. For anything more than a couple of files, select Review suggestions to open Review Index suggestions. The list on the left names every file with a proposal; selecting one shows on the right where it would go, whether that target is a folder or a placeholder, the name the file will carry in the index, and the AI rationale: a short explanation of what the document is and why that slot was chosen. Search by filename, path or reasoning to find a particular file.
   The rationale says what the AI read in the document. Accept places the file at the target; the tree below lets you choose a different one first.11.AI rationale22.Target33.Accept
4. Decide each suggestion:
   • Accept places the file at the proposed target with the proposed name. Tick several rows and use Accept selected for a batch.
   • To move it somewhere else, pick a folder or placeholder in the tree on the right and select Save target, then accept.
   • Edit the Filename in Index if the proposed name is not what buyers should read. The name follows the project's AI file naming convention; the original filename stays in the Workspace.
   • Reject drops the proposal and leaves the file in the Workspace.
   
   A suggestion whose target no longer exists, for example because the placeholder was deleted, is marked Needs retarget and cannot be accepted until you choose a new target.

How the AI decides

The AI reads the processed text of each unplaced file and compares it with the Dataroom: folder names, placeholder names and, above all, the Placement instruction on each placeholder. A file gets at most one suggestion, and only when one slot is a clearly better fit than the rest; a file that could go in two places, or matches nothing, is left without a suggestion rather than guessed. That is why a scan over a hundred files typically proposes slots for most of them and leaves a handful for you.

Three things follow from this. Add or reword placeholders and instructions first, then run Scan Workspace again; a rescan only looks at files that still have no placement. A suggestion made before the structure changed may point at a target that has since been deleted or moved, and is marked Target missing or Placeholder missing until you retarget it. And a file the AI keeps skipping is not stuck: place it by hand with Place in Dataroom; see Place documents in the Dataroom.

Check the result

Accepted files show their new index reference in the Workspace's Index column, appear under their placeholder on the Dataroom tab, and the completion counter at the bottom of the Dataroom rises. The Review suggestions button no longer shows a dot once every suggestion has been handled.

If something goes wrong