> ## Documentation Index
> Fetch the complete documentation index at: https://docs.penbox.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Build your first Document Intelligence template
> Step through a complete Document Intelligence template using Penbox Intelligence — validation, typed extraction, and a live test, with a bank RIB as the worked example.
This guide walks you through building a **simple, production-ready Document Intelligence template** powered by **Penbox Intelligence**: natural-language validation, structured fields the model must respect, and an in-app test loop so you trust the output before anyone uploads a real file.
We use a **bank RIB** (*Relevé d'Identité Bancaire*) as the example document — the same pattern applies to invoices, IDs, contracts, or any PDF or image you need to understand at scale.
## What you'll build
When you finish, you will have a **fully functional Document Intelligence template** that:
* Uses **Penbox Intelligence** to **validate** uploads against criteria you write in plain language, so bad or irrelevant files are caught early
* **Extracts** a strict schema of **typed** values (text, number, date, or list) you can reuse in cases, forms, and automations — not loose blobs of text
* Ships with a **verified test run**: you have already seen validation, confidence, and field-level results on a real sample
**Estimated time:** 10–15 minutes
***
In your Penbox workspace, open **Templates** in the sidebar, then choose **Document Intelligence**.
Click **+ New**.
Give your template a name that will read clearly in your workspace — for this example, use *RIB extraction* — then click **Create**.
The editor keeps **configuration** and **Try it** in one place, so you can define the template and prove it on a sample file without switching context.
Under **Validation**, tell Penbox Intelligence what must be true for this document type. That grounds extraction: the model checks the file against your rules before you trust the fields.
For our RIB example, you might use:
*A RIB is valid if it contains the following fields: bank name, bank logo, IBAN account number, bank account holder.*
Under **Fields to Extract**, click **+ Add field** for each value you want from the document.
For each field:
1. **Choose a type** — **Text**, **Number**, **Date**, or **List**. Penbox Intelligence returns values in that shape, so downstream steps and APIs stay predictable.
2. **Name the field** — mirror the wording on the document when you can (e.g. *IBAN account number*) so the model anchors on the right region.
3. **Add an optional hint** — steer the model when the layout is ambiguous (e.g. for French IBANs: *An IBAN bank account starts with FR*).
Repeat for every field your template must output. For the RIB walkthrough, add **four** text fields and hints like below — then open **Configure** and confirm the full schema matches what you want in production.
| Field | Type | Hint (example) |
| ------------------------ | ---- | ------------------------------------------------- |
| IBAN account number | Text | An IBAN bank account starts with FR |
| BIC | Text | *Optional* |
| Bank account holder name | Text | Extract the full name: given name and family name |
| Bank name | Text | Name of the bank where the account is hosted |
On the left, under **Test Document**, upload a representative file (PDF, JPG, or PNG). Here we use a sample **RIB**; for your own template, pick a file that matches the document type you described in validation.
With your test file in place, click **Try it** to run Penbox Intelligence end-to-end on that upload. After you tweak validation or fields, use **Retry** to re-run without leaving the editor.
Open the **Results** tab: you get a validation verdict, image-quality signal, and each extracted value with confidence — the same signals your live flows can rely on once the template is attached to a **File** element.
If the output matches the source document, your template is ready to ship. If not, iterate in **Configure** — tighter validation, clearer field names, and sharper hints usually close the gap in one or two cycles.
***
## You're done
You have built and battle-tested a **Document Intelligence** template on **Penbox Intelligence**: human-readable validation, a strict extraction schema, and proof from a real upload. Swap the RIB for any document class you need — the workflow is the same, and the template is ready to plug into your Penbox flows.
***
## What's next?
Create or update templates programmatically via Pen Connect.
See how Document Intelligence fits into recent Penbox releases.