MenuFR

Internal product study · 2026

Local Document Assistant

A local business assistant that turns mixed documents into structured information, checks what matters and prepares outputs with human review.

Concept, interfaces and data designed by VALENQO. The documents shown are synthetic. File D-042 is a separate example from the one on the home page (CMD-024).

How a case moves

From document to output, every step stays readable.

Follow case D-042: values are extracted with their source, discrepancies are flagged, a person decides, and only what has been approved makes up the output.

Sources

DOC-AForm

  • Field 01Reference: D-042
  • Field 04Date: 12/03/2026
  • Field 06Estimated quantity: 10

DOC-BTable

  • Line 06Packaging · 4
  • Line 07Supplies · 12
  • Line 08Transport · 1

DOC-CNote

  • § 01Request received early in the month.
  • § 02supplies batch grouped delivery

Record D-042Human review

  1. ReferenceD-042DOC-A · field 01Approved
  2. Date12 March 2026DOC-A · field 04Approved
  3. Quantity12 unitsDOC-B · line 07ApprovedApproved after review
  4. DescriptionBatch of suppliesDOC-C · § 02ApprovedCorrected

Output

VALENQO / D-042

Case summary

Reference
D-042
Date
12 March 2026
Quantity
12 units
Description
Batch of supplies

Sources: DOC-A, DOC-B, DOC-C

Human review completed

05 / 05The output is assembled from approved fields only, with a record of the review.

Intake

Different sources, one working case.

The assistant receives files in different formats and keeps every source identifiable. All processing stays inside the team’s local environment.

  1. DOC-A

    Form

    Entered fields and attached files.

    Received
  2. DOC-B

    Table

    Data spread across several rows.

    To extract
  3. DOC-C

    Note

    Information written as free text.

    Received

A form, a table and a note: three structures the assistant brings back to one data model.

Structure

Every field keeps its source.

Parsing and extraction propose a structured record. Provenance stays visible so the team can check any value in its original document.

Case D-042
Reference
D-042
DOC-A · field 01
Date
12 March 2026
DOC-A · field 04
Quantity
12 units
DOC-B · line 07
Description
Batch of supplies
DOC-C · § 02

Source → extracted field

A field is not automatically true: the user can open its origin, correct the extraction and keep a record of that decision.

Checks

Exceptions go to a person.

Simple rules check required fields and discrepancies between sources. When document search or an AI model helps, its suggestion still requires approval.

  1. Reference confirmed

    D-042 / reference

    Approved

    The value is present in the form.

  2. Two values to compare

    D-042 / quantity

    Approved after review

    A person compared the sources and confirmed 12 units.

  3. Wording corrected

    D-042 / description

    Corrected

    A person adjusted the wording before generation.

Every check states what it verifies, and every decision states who made it.

Generation

An output prepared after review.

The output document assembles approved fields only. It remains editable, and its status shows that a human review took place.

VALENQO / D-042Human review completed

Case summary

Reference

D-042

Date

12 March 2026

Subject

Batch of supplies, 12 units

Sources

DOC-A, DOC-B and DOC-C

Local architecture

A complete workflow inside a local boundary.

The interface, extraction services, data model and generation stay in a controlled environment. Retrieval or AI can assist a step without becoming the mandatory starting point.

Local boundary

  1. 01Intake and parsing
  2. 02Structured data
  3. 03Rules and useful search
  4. 04Human review
  5. 05Generation and export

The same breakdown adapts to a workstation, an internal server or a private cloud environment.

Technology in service of real work.

This study shows one way to design business software: make sources readable, leave decisions to the team and produce a verifiable result.

Let’s talk about your workflow