How autofill works
Autofill takes a case and a template, resolves every field value from its source, and writes the values onto the template to produce a finished PDF.
Last updated
Autofill takes a case and a template, resolves every output value from its source, and stamps 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. When extraction runs first, the AI step is non-deterministic; the fill that follows is not.
Core concepts
The five execution modes
All five modes share the same resolver and renderer. They differ only in which steps run.
- Extract Only — a file is ingested or you click Re-extract. AI reads the document and writes values into the case. No PDF is produced.
- Fill Only — you pick a template on a case that already has data. The template fills from
case.data, static values, and any form answers. Output: a filled PDF. - Extract + Fill — you upload a file with a template active, then run autofill. Extraction runs first, then the fill. Output: updated case data plus a filled PDF.
- Update Case — you run a template whose extracted fields write back to the case. The PDF stamps as usual, and the resolved values are saved into
case.data. - Batch Fill — you select multiple cases and run one template across all of them. One output per case. Extraction does not run.
Field source + output kind
Every field has a source — question (user fills via form), extracted (AI pulls from document or context), or static (fixed value). Each field then declares one or more outputs, and each output's kind decides how its value is produced:
input— typed by the user on a question fieldstatic— fixed at template design timezone— AI/OCR reads a rectangle (read zone) you drew on the source documentprompt— derived from a natural-language AI prompt running over other case datacomputed— derived from a formula referencing upstream outputsbinding— pulled from non-document context: email metadata, the email body, system clock, or company profile
Static-kind outputs and question-source fields are excluded from AI extraction — the AI never sees them.
Output assembly
Each output's value lands in case.data under its key (e.g. case.data.vendor_name). At fill time, each write placement on the template reads case.data[placement.dataKey] and stamps that value into the PDF. The template's source tabs (workspace document sources, blank pages, template references) define the page order; additional pages from case files referenced by those tabs are stitched in. The result is a single PDF.
Walkthrough: filling a vendor cover sheet
You have a case with extracted invoice data. You want to generate a cover sheet PDF.
- Open the case in the dashboard.
- Click Autofill in the file rail (desktop) or in the action menu (mobile).
- Pick the Vendor Cover Sheet template.
- Review the resolved values in the preview.
- Click Run Autofill. The PDF lands in the case's Outputs section.
- Click the new output to download.
Common patterns
Quick Fill for one-offs
The Quick Fill card in the template studio sidebar runs a template against a single uploaded PDF without creating a case. Use it to test a new template or generate a one-off document. See Lite autofill.
Batch templates for bulk runs
Select multiple cases in the dashboard and click Autofill in the batch action bar. One output per case.
Form questions for ad-hoc values
When a template needs a value that isn't on the case (like "which package to send"), set the field's source to question. Autofill prompts you at run time and uses the answer.
Output naming patterns
The template's output filename pattern controls what the generated file is named. Tokens like {vendor_name} and {permit_number} are replaced with field values.
Troubleshooting
Some fields are empty in the output even though the case has values.
Check the field's source. If it's question, you have to type the value at run time. If it's extracted, the placement's data key must match an output key the field actually produces — case.data[placement.dataKey] is the lookup path.
Autofill shows "no template available." Templates appear only when the case has a file in a category the template accepts. Check the case's file categories and the template's input categories.
The output looks wrong — fields in the wrong place. Field placement is exact. Open the template studio and check the placement coordinates.
The output is missing pages from the source files.
Check the source tabs in the document bar above the template editor's preview. Pages aren't included unless a source tab references them. Open the active tab's chevron popover (multi-source layout) or ⋯ menu (single-source layout) and verify the "Include pages from source" checkboxes.