monkmediaone.tech

Sovereign AI Deployments✦ Private LLM Infrastructure✦ Data-Sovereign Automation✦ Zero Data Leakage Workflows✦ Compliance-Ready Automation✦ On-Premise AI Agents✦ AI You Own. AI You Control.✦ Sovereign AI Deployments✦ Private LLM Infrastructure✦ Data-Sovereign Automation✦ Zero Data Leakage Workflows✦ Compliance-Ready Automation✦ On-Premise AI Agents✦ AI You Own. AI You Control.✦
Monk Media One Monk Media One Book a Discovery Call
Automate WhatsApp Document Delivery: Save 90% Staff Time with N8N
How We Automated Document Delivery on WhatsApp — 90% Time Saved

Discover how we built a WhatsApp Document Automation to eliminate manual document sharing, saving our financial client 90% staff time.

By Monk Media One Tech | AI Automation Agency, Ahmedabad, India

A full case study: N8N + WhatsApp Business API + cloud storage automation that eliminated manual document sharing for a financial services client.

Our client had 200+ active customers who regularly needed documents — account statements, invoices, certificates, policy documents, receipts. Every request came through WhatsApp. Every request required a staff member to find the right file, locate the right version, and send it manually.

It was taking 3–4 hours of staff time every single day. On busy days, customers waited 4–6 hours for a document they needed urgently.

We built a WhatsApp AI assistant that handles every document request automatically, in under 60 seconds, with zero human involvement.

Here's exactly how we built it.

The Problem (In Detail)

Before the automation, the process looked like this:

  1. Customer sends a WhatsApp message: "Please send my account statement for March"
  2. Staff member reads the message (if they're not busy)
  3. Staff navigates to the file storage system
  4. Searches for the customer's account
  5. Finds the correct document for the requested period
  6. Downloads it
  7. Sends it back on WhatsApp
  8. Marks it as done

Average time per request: 8–15 minutes. Daily volume: 15–25 requests. Total daily staff time consumed: 3–4 hours minimum.

On quarter-end, this jumped to 50+ requests per day. The team was overwhelmed. Customers complained. Document requests fell through the cracks.

The Solution Architecture

Customer WhatsApp Message
         │
         ▼
WhatsApp Business API (webhook)
         │
         ▼
N8N Workflow Engine
         │
    ┌────┴─────┐
    │           │
Identify     Identify
customer     document
(phone no.)  (intent NLP)
    │           │
    └────┬─────┘
         │
         ▼
Google Drive / S3 Search
(find the right file)
         │
         ▼
Send document via WhatsApp API
         │
         ▼
Log request to Google Sheets

Five components:

  1. WhatsApp Business API — receives messages and sends responses
  2. N8N — the orchestration layer connecting everything
  3. OpenAI GPT-4o-mini — understands the request and extracts intent
  4. Google Drive — stores all customer documents
  5. Google Sheets — logs every request for auditing

Step-by-Step Build

Step 1 — WhatsApp Business API Setup

You need a WhatsApp Business account through Meta's Cloud API (free) or a BSP (Business Solution Provider) like Twilio, Interakt, or Gupshup.

For Indian businesses, we recommend Interakt or WATI — they're Indian BSPs with INR billing, local support, and faster WhatsApp Business Account approval.

Once approved, you get:

  • A dedicated WhatsApp phone number
  • A webhook URL where messages arrive
  • An API to send messages back

Configure your webhook to point to N8N:

Webhook URL: https://your-n8n.yourdomain.com/webhook/whatsapp-incoming

Step 2 — N8N Webhook Trigger

In N8N, create a new workflow starting with a Webhook node:

{
  "node": "Webhook",
  "parameters": {
    "httpMethod": "POST",
    "path": "whatsapp-incoming",
    "responseMode": "lastNode"
  }
}

The webhook receives a JSON payload from WhatsApp like this:

{
  "entry": [{
    "changes": [{
      "value": {
        "messages": [{
          "from": "919876543210",
          "text": { "body": "Please send my invoice for February" },
          "timestamp": "1706784000"
        }],
        "contacts": [{
          "profile": { "name": "Rajesh Shah" },
          "wa_id": "919876543210"
        }]
      }
    }]
  }]
}

Extract the key fields using N8N's Set node:

  • phone_number: {{ $json.entry[0].changes[0].value.messages[0].from }}
  • message_text: {{ $json.entry[0].changes[0].value.messages[0].text.body }}
  • sender_name: {{ $json.entry[0].changes[0].value.contacts[0].profile.name }}

Step 3 — Intent Extraction with AI

Add an OpenAI node to understand what document is being requested:

System prompt:
You are a document request classifier. Extract the document type and time period from the customer message.

Return JSON only:
{
  "document_type": "invoice|statement|certificate|receipt|policy|other",
  "time_period": "month name and year if mentioned, or 'latest' if not specified",
  "confidence": "high|medium|low"
}

If you cannot determine the document type, set document_type to "unclear".

User message: {{ $json.message_text }}

This converts "Please send my GST invoice for last quarter" into:

{
  "document_type": "invoice",
  "time_period": "Q3 2024",
  "confidence": "high"
}

Step 4 — Customer Lookup

Using the phone number, find the customer in your system. We use a Google Sheets lookup (simple) or a database query (for larger systems):

// N8N Code node
const phone = $input.first().json.phone_number;
const sheet_data = $input.all(); // From Google Sheets Read node

const customer = sheet_data.find(row => 
  row.json.phone_number === phone || 
  row.json.phone_number === phone.replace('91', '')
);

if (!customer) {
  return [{ json: { found: false, phone: phone } }];
}

return [{
  json: {
    found: true,
    customer_id: customer.json.customer_id,
    customer_name: customer.json.name,
    folder_id: customer.json.drive_folder_id
  }
}];

Step 5 — Google Drive Search

Using the customer's folder ID and the extracted document type/period, search for the right file:

// N8N Google Drive node — Search Files
// Query: name contains 'invoice' and 'Feb 2024' and parents in 'FOLDER_ID'

const doc_type = $json.document_type;
const period = $json.time_period;
const folder_id = $json.folder_id;

const query = `name contains '${doc_type}' and name contains '${period}' and '${folder_id}' in parents and trashed = false`;

If one file is found → send it.
If multiple files found → send the most recent one and note others are available.
If no file found → send a "not found" message and create a staff notification.

Step 6 — Send Document on WhatsApp

// N8N HTTP Request node — WhatsApp API send document

{
  "method": "POST",
  "url": "https://graph.facebook.com/v18.0/YOUR_PHONE_NUMBER_ID/messages",
  "headers": {
    "Authorization": "Bearer YOUR_WHATSAPP_TOKEN",
    "Content-Type": "application/json"
  },
  "body": {
    "messaging_product": "whatsapp",
    "to": "{{ $json.phone_number }}",
    "type": "document",
    "document": {
      "id": "{{ $json.whatsapp_media_id }}",
      "caption": "Here is your {{ $json.document_type }} for {{ $json.time_period }}, {{ $json.customer_name }}. If you need anything else, just ask!"
    }
  }
}

Note: You first need to upload the file to WhatsApp's media servers to get a whatsapp_media_id, then send that ID. The N8N workflow handles this in two steps.

Step 7 — Logging and Error Handling

Every request gets logged to Google Sheets:

Timestamp Customer Phone Request Document Found File Name Response Time
2024-02-15 14:32 Rajesh Shah 9198... invoice feb Yes INV_Feb2024_RS.pdf 4.2s

For failed requests (document not found, customer not in system), the workflow:

  1. Sends the customer a polite "working on it" message
  2. Sends a WhatsApp notification to the staff member responsible
  3. Logs the failed request for follow-up

The Numbers After 60 Days

Metric Before After
Average response time 4–6 hours 45 seconds
Staff hours per day on doc requests 3–4 hours 20 minutes
Customer complaints about delays 8–12/week 0–1/week
Requests handled automatically 0% 87%
Staff cost for this task ₹12,000/month ₹2,000/month

The 13% of requests not handled automatically are either unusual document types, new customers not yet in the system, or requests requiring judgment (e.g., "send me everything from last year"). These get flagged for human handling.

Extending the System

Once the base workflow was running, we added:

Outbound proactive notifications — When a new document is generated (monthly statement, renewal notice), the system automatically sends it to the customer on WhatsApp. Zero staff involvement.

Voice note transcription — Some customers send voice notes instead of text. We added a Whisper transcription step before the intent extraction.

Hindi/Gujarati support — OpenAI handles Hindi and Gujarati text naturally. We added a language detection step and adapted the response templates.

Request history — Customers can ask "what documents have you sent me this year?" and get a summary from the log sheet.

What This Costs to Build and Run

Build time: 5–7 working days for a mid-complexity implementation.

Running costs per month:

  • N8N (self-hosted): ₹1,200/month (VPS)
  • OpenAI API (intent extraction): ~₹500–1,500/month depending on volume
  • WhatsApp Business API: ₹0 for first 1,000 conversations/month, then ~₹0.50–0.70 per conversation
  • Total: ₹2,000–4,000/month

For a client that was spending ₹12,000/month in staff time on document requests alone, this system paid for itself in the first week.

We're Monk Media One Tech — AI automation agency, Ahmedabad. We build WhatsApp automation, N8N workflows, and AI systems for businesses across India and Canada.

Book a free discovery call: monkmediaone.tech/contact
📞 +91 88668 19349 | hello@monkmediaone.tech

Subscribe
Subscribe