Run autofill
Open a file in the case review modal, pick a template in the right-side Autofill rail, click Autofill, and the finished PDF lands as its own row in the case Documents list, tagged with the template output category.
Autofill takes a case and a template, resolves every field value from its source, then writes the values onto the template to produce a finished PDF. The fill step is deterministic: same case + same template + same source data = the same PDF, every time. Extraction (the AI step) is non-deterministic; the fill that follows is not. (One exception: templates with Document reflows enabled add an AI placement step at fill time, covered under Troubleshooting.)
How values flow
Each field's inputs (Question, Extracted, or Static) resolve into case.data; its outputs then act, a write zone stamping a value onto the PDF or a transform deriving a new one. The full field model lives in Build a template. What matters at fill time: each write zone reads case.data[placement.dataKey] and stamps it; source tabs (workspace document sources, blank pages, template references) set page order, with pages from referenced case files stitched in; and named PDF form fields are placed by name and never move.
The five modes share one resolver and renderer, differing only in steps run: Extract Only (ingest or Re-extract, no PDF), Fill Only (fill from existing data; a Generate <template> row builds the deliverable with no incoming file), Extract + Fill, Update Case (write extracted values back), and Batch Fill (one batch template across many selected cases, each filling one row into a single combined PDF saved to every case, no extraction). Two "batch" terms meet here: a batch field is a field type holding a variable number of rows, while Batch Fill is the operation, one run across many selected cases each filling one row of that field.
Run it on a case
The review modal pins an Autofill rail to the right of the preview, the canonical entry point, holding the template picker and one Autofill button. Clicking it swaps the preview to a prompt: form questions, an optional Special AI Instructions textarea (this run only), and Run Autofill (if the template has no form questions and no prefilled rows, the prompt is just the textarea and the button, which you can leave blank to run the template's default logic). The output PDF appears as its own row in the case's Documents list, tagged with the template's output category. The kebab (...) menu owns Re-extract and Share.
Some rows may fill themselves. If the template came from an uploaded fillable form, boxes that match information the workspace already collects arrive prefilled from the case and carry a small From case data tag. Review them like any other answer: edit or clear any of them before running. These values always reflect the case's current information at run time, and the app does not save a separate copy, so updating the case updates the next run on its own.
On mobile, the kebab opens a bottom drawer with the picker plus Autofill, Extract/Re-extract, and Share.
You can also trigger runs from the case's Documents list: the files sit under a Fill from template header (which also holds the sort and Upload controls). Expand any file to see the templates you can run on it as a short list of rows, the ones built on that document first, each with a chevron and a Generate button (Extract for read-only templates). The chevron (or clicking the row) opens that template's inline form, where you can review or type inputs before running. If a template still needs answers or extraction data that hasn't arrived, clicking its button opens the same form first, where you type the values or upload the missing source document. Templates that build a deliverable from saved case values alone sit under the Generate from template group at the bottom of the list. See Cases and data for the full Documents layout.
To run from the editor instead, the toolbar's split Run button tests against the template's stored sample documents, while its dropdown's Run with your documents… option fills against PDFs you upload (the filled PDF downloads automatically and appears in the preview pane): use it to test a new template or generate a one-off. On Pro and Scale these editor runs create no case; on Free and Lite each one is saved as a temporary case in your dashboard for 14 days. For bulk runs, select multiple cases in the dashboard, pick a batch template in the batch action bar, and click Autofill.
Troubleshooting
Fields empty despite case values: check the input source. Question must be typed at run time; for Extracted, confirm the write zone points at the input or transform that produces the value (a write zone stamps whatever value its key points at in case data).
The template I want is far down the list. Every template is offered, ordered by how well it matches the file. If the one you want is not near the top, it is probably not built on this document: open it and check the tabs above the page (build a template).
Fields in the wrong place: placement is exact by default. If content shifts between fills (variable-length sections pushing fields around, a paragraph growing, a signature block moving onto the next page), enable Document reflows in the template's details dropdown (click the template name in the editor toolbar) so write zones re-find their anchor with AI. If an anchor can't be found, the value stamps at its authored position rather than guessing.
Missing pages: pages aren't included unless a source tab references them. Open the document bar above the editor preview, then the active tab's chevron popover (multi-source layout) or ⋯ menu (single-source layout) and check the "Include pages from source" boxes.