Product · Tables
A document per row. A question per column.
A table is the shape the answer was always going to take: the pile down the side, the questions across the top, and a cell where each pair meets. Write a question once and every document answers it — typed, so a date sorts as a date, and cited, so every cell says where it was read.
Free credits on sign-up. No card, no subscription.
| Row | Documents | Supplier | Due Date | Total Amount | Line Items |
|---|---|---|---|---|---|
| Harlow Plant inv 0412.pdf | Harlow Plant Hire Ltd | 1 Oct 2026 | 1,240.00 | Excavator hire, Delivery | |
| Northgate Supplies 7781.pdf | Northgate Supplies | 14 Oct 2026 | 386.50 | Timber, Fixings, Delivery | |
| Calder Electrical receipt.jpg | Calder Electrical | 30 Sep 2026 | 92.00 | Cable tray | |
| Wharfside Facilities 2291.pdfCredit note CN-118.pdf | Wharfside Facilities | 21 Oct 2026 | 2,015.00 | Cleaning, Waste removal |
Typed, so it sorts.
Every column has an output type — text, a number, a date, yes or no, one of a fixed set, a list, a picture or several pictures — and the type is what turns a column of answers into something you can work with. A due date sorts as a date. A total filters as a number.
Filters are built from each column’s type, so a date column offers date comparisons and a number column numeric ones, and conditions join together. They read what a cell shows: a figure somebody corrected by hand is the figure a filter sees.
A filter is a way of looking at the table, not a setting on it, and it is not saved with the table.
| Document | Supplier | Due Date | Total Amount |
|---|---|---|---|
| Harlow Plant inv 0412 | Harlow Plant Hire Ltd | 1 Oct 2026 | 1,240.00 |
Asking again costs the asking, not the reading.
Uploading reads each document once, and that reading is what a page is charged for. A column added a year later is charged per cell and nothing more, because its pages were read on the way in.
Change a column’s question and its answers are marked stale rather than quietly rewritten — you decide when to rerun them, and a run is quoted in credits before it starts. A rerun leaves cells corrected by hand alone, and a cell that fails is not charged.
One row can be several documents.
A contract and its amendments are one agreement. Group them into a bundle and the bundle becomes the row, answered as one subject — and each answer’s citation names the document it came from, so a rent that was varied points at the deed that varied it.
A bundle holds up to fifteen documents. The first three are included in a cell’s rate, and each one past the third adds to it.
What else a table does.
| Feature | What it does |
|---|---|
| Auto mode | Answers each new row added to the table — a file dropped onto it, an email, a connector sync — without anyone pressing Run. It does not rerun rows already answered. |
| Answer language | Writes a table’s prose answers in another language. Dates, numbers and categories stay as they are, so they still sort. |
| Templates | Start from a template built for your document type and edit its questions, or save your own for the whole organisation. |
| Export | XLSX or CSV, with a row range and the columns you choose. Exporting costs no credits. |
| Sharing | Granted per workspace, at Read, Read & Write or Manage — to the whole organisation, to a role, or to one person. |
What it will not do.
- It does not keep your filters. A filter narrows what you are reading and is gone when you leave; the table itself is unchanged.
- An export is a snapshot. It carries the answers, not the citations or the pictures behind them, and it does not update when the table does.
- Auto mode fills the table; it does not watch it. It answers new rows as they are added and nothing more — no schedule, and no reminders.
A table is one part of it. Extraction is how each cell gets answered, citations are how you check one, and integrations are how documents arrive without anyone uploading them.
Start with the pile you already have.
Signing up comes with free credits. No card, no subscription.