Monday, 9:00 AM. An external auditor sends a request: every supplier invoice in the September sample, one row per invoice, as a CSV file by Friday. Supplier, invoice number and date, purchase order, payment terms, due date, currency and total.
No system holds that list in this shape. The answers sit in the documents themselves — invoices in English, Spanish and Czech, plus a paper copy from the morning mailroom scan. Usually that means retyping, or a ticket for IT to build a report.
Our new video guide, Audit-Ready, On Demand, shows a third way. It runs just under six minutes, you can watch it at the top of this page, and it follows the accounts payable team of Keldmoor, a fictional distributor, from the request to the reply. Three steps, no code. The audit is only the example: the same steps give you any list, in the shape you need, from any set of documents.
Step one: describe the columns you need
It starts in Model Manager. A model defines which fields DocReveal extracts from documents and in what structure. Canonical models describe common document types; custom models are yours, shaped to your own systems or to a single request.
The team creates an empty model called Audit sample and adds eight fields — exactly the auditor’s columns. Each field gets a type — text, a number, a whole number, yes or no, or a nested object or list, such as the line items of an invoice — and dates get the Date format.
The most important part is the description. It goes to the AI that reads the documents, so it says in plain words what the field means: the supplier’s legal name as printed on the invoice; the days to pay as a number; no due date if the invoice does not state one. A check looks over the structure, and publishing shows exactly what changes before anything is saved.
Step two: pick the documents
In File Manager, the team filters by document type and selects the eight documents — across folders, including the paper copy and a freight invoice split out of the mailroom scan. One click on Reveal Group, and Reveal Manager opens a new run with all of them.
Under Mapping models, the team adds Audit sample. Every file was read when it arrived, so nothing is parsed again: DocReveal reuses the existing result and fills the new structure. A board follows every file, and in the video the run is done a few minutes later, with all eight results written to Analytics.
Step three: one table, one file
In Analytics, the new dataset opens in the Data Explorer. Pick the eight fields as columns, run the query, and there it is: eight rows, one per document. American, Spanish and Czech dates now share one format. Net 45, 30 días and 30 dní became plain numbers. Dollars stay dollars, euros stay euros.
Sorted by invoice date, the table shows something worth knowing: INV-88163 appears twice — the PDF and its paper copy from the mailroom, caught before anyone pays it twice. And where a freight invoice states no due date, the cell stays empty. Nothing is invented.
Exports turns the same query into a CSV file with the columns in the auditor’s order, and Download saves it to your computer. Behind every row sits an audit layer: the run, the model version, the time and the full extracted record. At 9:19, the reply is on its way.
The audit is one request among many. The same three steps build a supplier list for another system, a register of contracts and their notice periods, or the answer to a one-off question from management — whatever columns your business needs.
Next month, it is already set up
The model stays. When next month’s invoices arrive, they go through the same model: same columns, new rows. Nobody rebuilds anything: the next list is a new run, not a new project.
Let an agent do it
Once the model exists, the rest can also run without a person at the keyboard. DocReveal speaks MCP, the open standard that AI assistants and agents use to work with other systems.
An AI agent connected over MCP, with a key you control, can do what the team did by hand: list the documents and your models, start a mapping run with your custom model, and read the rows back through governed, read-only queries — no free-form SQL, nothing changed. Starting a run counts as a sensitive action, so it follows the policy you set: run automatically, ask for confirmation or wait for a person’s approval. Every call is logged. How keys, approvals and the audit work is the story of Agents under control.
Inside DocReveal, the same tools are available to the agents you build in Agent Manager, and those agents can start on a schedule, when a mailbox receives something or when an event arrives. In Integration Manager, an integration can start on its own as soon as documents are mapped. Your agent asks for the list, DocReveal returns the rows, and the agent passes them on — to a report, a spreadsheet or another system it is allowed to use.
What to take away
- You define the structure. Your columns and your names, described in plain words — no code, no IT ticket.
- Any set of documents. Any language, any layout, paper copies included; results that already exist are reused.
- One table you can check. One format for dates and numbers, empty where a document says nothing, and an audit trail behind every row.
- Ready for agents. Once the model exists, an AI agent can run the rest over MCP — under your rules.
Keldmoor, its suppliers and its auditor are fictional demo data; the screens are the real product.
Your columns. Every document.