Automation does not mean giving up control. AI can interpret what the customer is asking for, but prices, taxes, inventory, discounts and commercial terms should come from controlled business sources.

A customer sends a WhatsApp message: “Hi, I need a quote for 25 laptops for our office. Preferably 16 GB of RAM, delivery in Miami. Can you send me a price?” In many companies, that message starts a manual chain: someone reads it, interprets the request, searches for products, checks inventory, calculates prices, prepares a document, sends it and then follows up.

A well-designed workflow can turn that conversation into structured data, retrieve real company information, apply commercial rules and generate a draft quote. AI is particularly useful for transforming unstructured language into structured fields that other systems can process (OpenAI, 2024).

Vallenoval practical principle AI interprets; the system retrieves; rules validate; humans control exceptions.

1. What are we actually automating?

The objective is not simply to build a chatbot. We want to turn informal customer messages into information that the sales process can use: intent, product, quantity, specifications, destination, required date and missing data.

Field Detected information
Intent Request a quote
Product Laptop
Quantity 25
RAM 16 GB or more
Storage 512 GB SSD
Destination Miami
Required date Not provided

2. Recommended architecture

An implementation can use the WhatsApp Business Platform as the incoming communication channel, a webhook to receive events, n8n as the orchestration layer, an AI model to interpret the message, and an ERP, CRM, product catalog or database as the source of pricing and inventory information.

CUSTOMER ↓ WHATSAPP BUSINESS PLATFORM ↓ WEBHOOK / TRIGGER ↓ N8N ↓ AI: EXTRACT DATA ↓ VALIDATE ↓ CATALOG / ERP / CRM ↓ BUSINESS RULES ↓ DOES IT REQUIRE APPROVAL? ↓ QUOTE → WHATSAPP / EMAIL → CRM

3. Receive and interpret the request

If the customer writes: “We need 25 laptops for the sales team. Minimum 16 GB RAM and 512 GB SSD. Can you send us a quote including delivery in Miami?” AI can classify the intent and extract the relevant fields without inventing information that the customer did not provide.

{ "intent": "quote_request", "product": "laptop", "quantity": 25, "ram_gb": 16, "storage_gb": 512, "delivery_city": "Miami", "required_date": null, "missing_data": ["required_date"] }

4. Detect missing information

If the customer only writes “I need a quote for 10 computers,” the system should not assume the specifications. It should record which information is missing and ask a specific question before continuing.

Design rule Unknown data does not mean data that AI should guess.

5. Retrieve real pricing and inventory

Pricing should come from a controlled business source: a managed spreadsheet, database, ERP, CRM, ecommerce API or inventory system. The AI model should not become the company's official pricing database.

SKU Product Price Stock
LAP-PRO-16 Laptop Pro 16/512 $899 42
LAP-BUS-16 Laptop Business 16/512 $979 31
LAP-EXEC-32 Laptop Executive 32/1TB $1,249 12

If the authorized price is $979 and the customer requests 25 units, the subtotal is $24,475. That calculation can be performed deterministically and does not need to be estimated by the AI model.

6. Apply business rules

IF stock < requested quantity → stop automatic quote delivery IF requested discount > 5% → request approval IF margin < minimum margin → human review IF transaction exceeds defined threshold → approval required before sending

7. What should be automated and what should be reviewed?

Level 1 — Direct automation

  • Classify customer intent.
  • Extract product and quantity.
  • Register the lead.
  • Detect missing information.

Level 2 — Rule-based automation

  • Search the product catalog.
  • Verify inventory.
  • Calculate subtotals.
  • Apply predefined commercial limits.

Level 3 — Human approval

  • Exceptional discounts.
  • Special payment terms.
  • High-value transactions.
  • Margins below the minimum threshold.
  • Ambiguous cases.
Vallenoval practical principle The greater the financial impact or the difficulty of reversing a decision, the greater the level of human control should be.

8. Where does n8n fit?

n8n can operate as the orchestration layer. It can receive the event, send the text to the AI model, validate the response, query external systems, apply conditions, generate the document and register the result in the appropriate business system.

9. What happens after the quote is sent?

The process does not need to end with a PDF. The workflow can record that the proposal was sent, schedule follow-up activities, detect customer responses and update the CRM. This makes it possible to measure response time, quoted value, acceptance rate, requested products and reasons for lost opportunities.

10. Three implementation levels

Initial: WhatsApp + n8n + AI + controlled price sheet + quote template.

Intermediate: WhatsApp + n8n + AI + CRM + product database + PDF + approval workflow.

Advanced: WhatsApp + AI + CRM + ERP + inventory + pricing + signature/payment + analytics.

11. Three questions to answer before automating

1. Where does the truth come from? Prices, inventory, taxes, terms and discounts should have clearly identified authoritative sources.

2. What can the system do automatically? Define limits for transaction value, margin and discounts.

3. When should a person intervene? Define the exceptions before putting the workflow into production.

Conclusion

Automating a quote request received through WhatsApp is not about connecting WhatsApp to an AI model and allowing it to respond freely. A stronger business architecture combines AI to interpret language, corporate systems as trusted data sources, deterministic rules to apply company policies, and people to control exceptions.

The opportunity is not simply to automate WhatsApp. It is to connect the customer conversation with the business process that happens behind it.

AI FOR BUSINESS

Is there a process in your company that could be automated?

If your sales, customer service or operations teams repeatedly move information between WhatsApp, email, spreadsheets, inventory systems or CRM tools, Vallenoval can help you evaluate where AI and automation could create practical value.

Evaluate a process in my company

Sources and references — APA 7th edition

  1. Meta Platforms, Inc. (n.d.). Webhooks for WhatsApp Business Platform. Meta for Developers. View source .
  2. n8n GmbH. (n.d.). n8n Docs. View source .
  3. n8n GmbH. (n.d.). WhatsApp Business Cloud node. n8n Docs. View source .
  4. OpenAI. (2024, August 6). Introducing Structured Outputs in the API. View source .