PROJECT: AI-POWERED CONVERSATIONAL DOCUMENT CLOUD ACCESSIBLE VIA WHATSAPP
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
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