Sales

Cake Orders in Nepali, Roman Nepali and English: A 98-Node n8n WhatsApp Agent That Can Never Mark an Order Paid

  • Level Advanced
  • Setup1–3 hrs (estimated)
  • Evidence Some evidence
  • CostNot stated in source
  • No-codeNo
n8nSource-linked

What they did

A developer in Nepal built a WhatsApp and website-chat ordering agent for a cake shop. Customers write in Nepali, Roman Nepali and English, often mixed, and send voice notes, photos and payment screenshots. The agent proposes orders, but code checks items, dates and delivery fees, prices come from the shop's own database, and a vision model only flags payment screenshots for the owner to confirm. A staff reply pauses the bot for that customer and the owner gets a Telegram alert when it needs help. It runs in production for the client.

How it went

n8n community post by the builder (Oct 26, 2026 UTC): 98 nodes across 5 workflows, 2 channels (WhatsApp via 360dialog and web chat), 3 languages, PostgreSQL, running in production for a client cake shop. No order volume numbers given

Worth knowing

  • n8n community post by the builder (Oct 26, 2026 UTC): 98 nodes across 5 workflows, 2 channels (WhatsApp via 360dialog and web chat), 3 languages, PostgreSQL, running in production for a client cake shop. No order volume numbers given.

Prompt

Cleaned version rewritten from the source — not a direct quote.

Act as the ordering assistant for my [shop] on WhatsApp. Customers may write in [languages] and send voice notes or photos. Look up items, prices, delivery zones and time slots only from my [menu sheet / database], never from memory. Collect the order details and reply in the customer's language. Never calculate the final total yourself and never mark an order paid: if a customer sends a payment screenshot, flag it for me to confirm. If a customer asks for a person, stop replying and alert me.

Recommended

Try it in Grok Bot

  1. Copy the prompt (button below).
  2. Open Grok Bot chat and paste it. Replace any [bracketed] placeholders.
  3. Connect the listed tools (Telegram, WhatsApp) as connectors when Grok Bot asks.
  4. If the task repeats, ask: "Make this a daily routine."

Alternatives: the same prompt works in n8n, Claude, ChatGPT Agent, OpenClaw.

Set it up

Tools: Telegram, WhatsApp · Stack: n8n

  1. Read the original source for full context.
  2. Connect the tools your agent needs.
  3. Paste the prompt and replace [bracketed] placeholders.
  4. Run once manually and check the output before scheduling it.

Generic suggested steps, not from the source.

Evidence

n8n community post by the builder (Oct 26, 2026 UTC): 98 nodes across 5 workflows, 2 channels (WhatsApp via 360dialog and web chat), 3 languages, PostgreSQL, running in production for a client cake shop. No order volume numbers given

Source: Web · Verified by us: not yet · Report broken or outdated

Related use cases

SalesReddit

Prospect B2B leads autonomously for 2 weeks (lessons)

An agent running 24/7 on a Mac Mini finds prospects, verifies emails, sends outreach, makes PDF briefings, and tracks the pipeline, keeping a lessons log so it improves over time.

  • Level Intermediate
  • Setup1–3 hrs*
  • Evidence Strong evidence
OpenClawSource-linked
SalesReddit

Research inbound leads when HubSpot fires a webhook

A HubSpot webhook alerts a 'Researcher' agent, which scrapes the prospect's site and LinkedIn for recent news. The trigger itself is a plain webhook, not an LLM.

  • Level Intermediate
  • Setup1–3 hrs*
  • Evidence Some evidence
OpenClawSource-linked
Inbox & communicationReddit

6-class email router for a 22-person consultancy

An IMAP trigger feeds an OpenAI classifier that routes 80–120 emails a day to Slack, Pipedrive and Zendesk. It hit 94% accuracy on 500 emails and saves 11 hours a week.

  • Level Intermediate
  • Setup1–3 hrs*
  • Evidence Some evidence
n8nSource-linked