PROJECT: AI-POWERED CONVERSATIONAL DOCUMENT CLOUD ACCESSIBLE VIA WHATSAPP

Job ID: 40210757

Budget: €250 – €750 EUR

PROJECT: AI-POWERED CONVERSATIONAL DOCUMENT CLOUD ACCESSIBLE VIA WHATSAPP
1. Overview

The project consists of creating an intelligent document cloud, accessible primarily through WhatsApp, where users can ask anything related to the company (machines, documentation, spare parts, regulations, internal data, etc.), and the AI automatically returns the correct information, either as a text response or by delivering the exact PDF document required.

This is not a traditional app and not a simple chatbot.
It is the living memory of the company, organized in the cloud and accessed conversationally.

2. Entry Point: WhatsApp

WhatsApp is the only access channel.

The phone number identifies the user.

There are no usernames or passwords.

The system automatically recognizes:

The phone number

The associated company

The assigned role

Users can send:

Text

Photos (plates, parts, documents)

PDFs

Audio messages

Videos

3. Roles and Permissions

Each phone number has an assigned role, which defines the type of documentation it can access.

Example roles

Operator: operator manuals, technical datasheets

Technician: technical manuals, exploded views, schematics

Administration: insurance, registration documents, legal documentation

Management: full access

External / Client: limited access

Permissions are applied by document type, not by folder.

4. Cloud Structure

There is no folder per machine.
Folders are global and organized by document type:

/FOTOS_PLACAS
/FOTOS_MAQUINAS
/CE
/MANUALES_OPERARIO
/MANUALES_TECNICOS
/DESPIECES
/ELECTRICOS
/HIDRAULICOS
/FICHAS_TECNICAS
/SEGUROS
/PERMISOS_CIRCULACION
/EXCEL


All documents for all machines coexist in these folders.

5. Internal Entity: “Machine”

Even without a machine-specific folder, the system creates an internal Machine entity to relate all information.

Main fields

Brand

Model (normalized)

Serial number / chassis number / internal ID

Name variants

Serial ranges (if applicable)

This entity is used to filter and relate documents precisely.

6. Intelligent Indexing of the Cloud

All content is automatically indexed.

PDFs

Internal text

OCR per page (for scanned documents)

Tables

References

Page numbers

Images

OCR of visible text

Automatic classification (plate, part, schematic, general photo)

Excel

Sheet reading

Columns

Cells

Relationships between data

Each document is linked to one or more machines with:

Confidence level

Evidence (page or text where model/serial appears)

7. Machine Identification via Plate Photo

This is the primary identification method.

Flow

User sends a photo of the identification plate

OCR extracts brand, model, and serial

Data normalization

High-confidence machine identification

Automatic filtering of all relevant documents

If the serial number is unclear, the system works with brand + model and may request additional confirmation.

8. Document Search (Example: CE Certificate)
Ideal case

User:

“I want the CE certificate for this machine” + plate photo

System:

Identifies the machine

Searches in /CE

Selects the correct PDF by model and/or serial

Sends the CE PDF directly via WhatsApp

If multiple options exist, the system asks for the serial number or a new plate photo.

9. Spare Parts Module via Photo
Key rules

A plate photo is always mandatory

Without plate identification, no exact reference is returned

Alternatively, the user may manually enter model and serial

Full flow

User sends plate photo

System identifies the machine (MachineKey)

User sends part photo
(recommended: one loose part + one mounted photo)

System searches only compatible manuals and exploded views

Part identification using:

OCR on the part

Visual similarity (shape)

Mounting context

Comparison with exploded diagrams

Result

Exact part reference

Exploded view PDF

Exact page

Item number (if available)

Visual evidence

If confidence is low, the system requests an additional photo.

10. Excel Search via Natural Language

Users can ask questions such as:

“How many hours can a bus driver work per day?”

“What insurance does machine 1696 have?”

The AI:

Searches relevant Excel files

Extracts the required data

Responds in natural language

No file opening required.

11. Company Phones and Contacts

From WhatsApp, users can ask:

“Company phone numbers”

“I want to call insurance”

The system returns:

Correct person

Role

Direct phone number

12. Languages and Translation

Automatic language detection

Responses in the same language as the user

Audio and video

Automatic transcription

Translation

Response in configured language

13. Document Delivery

Documents are delivered:

As a PDF attachment via WhatsApp, or

As a secure, time-limited link

Optionally:

Summary

Relevant page

Key extracted data

14. Development Phases
Phase 1 – Internal Use (MVP)

WhatsApp interface

Roles and permissions

PDFs and Excel

Plate-based identification

Document delivery

Translation

Phase 2 – Advanced

Spare parts via photo

Fault diagnostics

ERP integration

Multi-company (SaaS)

15. Final Definition

An AI-powered intelligent document cloud, accessible via WhatsApp, where the system understands what the user needs, searches across all company documents, and delivers the exact file or correct information, with full permission control and traceability.

16. Development Philosophy: Scalable and Evolutive Project

This project is not conceived as a finished product, but as a living platform, designed from day one to grow, improve, and integrate new functionalities continuously.

Core principles

Modular development: each function (documents, spare parts, Excel, translation, ERP, etc.) is an independent module

Scalable architecture: ready to grow in documents, users, and companies without rebuilding the system

Continuous improvement: the system learns from real usage

Future integrations considered from the initial design

This document defines the BEGINNING

The scope described here represents:

A solid functional starting point

A minimum viable foundation for internal validation

A first operational product that already delivers real value

It is not the final state of the system.

17. Evolution Roadmap (Mid-Term Vision)

Once internally validated, the system is prepared to progressively incorporate:

New document types

Improved image-based part recognition

Advanced fault diagnostics

ERP and external system integrations

Automated supplier ordering

Usage analytics and performance optimization

Multi-company SaaS model

Each phase will be driven by real usage and detected needs, not by a closed development plan.

18. Key Message for the Developer

The goal is a flexible, well-structured, and scalable development, where the initial objective is not to build everything, but to build the foundation properly, knowing the system will grow day by day.

The developer must clearly understand that:

This project will evolve continuously

New data sources and integrations will be added

The architecture must enable change, not block it