Ecom Agent Lab

Work: AI agent demos and client projects

Six builds, labelled honestly. Five are working demos built on sample store data so you can see exactly how each automation behaves. One is a live chatbot built for a client. No invented results.

  • Vapi
  • n8n
  • OpenAI
  • Gemini
  • WhatsApp
  • Google Sheets
Title frame of the Sara demo video: AI Voice Agent for E-commerce Stores. Answers customer calls 24/7, checks orders live, handles returns, logs every call. Built with Vapi, n8n, OpenAI and Google Sheets.
Title frame from the Sara demo video (demo build).

Behind the flagship demo

Real frames from the demo video and the n8n build. Sample store, sample order, real workflow.

Live transcript frame from the demo: Sara says one moment, the n8n workflow is triggered live with order status for order 1001, and Sara reads back that it shipped with TCS
Live transcript: the n8n tool call fires mid-call and returns the order status.
The n8n workflow behind the demo: order lookup and returns during the call, call logging after it
The n8n workflow: one branch answers during the call, one logs it afterwards.
How it works frame: caller, Vapi voice agent, n8n order lookup, Google Sheets log
How it works, as explained in the video.
Google Sheets call log row created automatically after the demo call
The Google Sheets row written automatically after the call.

Demo build

Sara: AI voice agent for an e-commerce store (flagship demo)

The problem

Store phones ring while the team is packing orders or after hours. Most calls are the same few questions: where is my order, when will it arrive, how do I pay or return it. Missed calls become missed sales.

What it does

  • Answers straight away and asks for the order number
  • Checks the order live through an n8n workflow (product, courier, delivery date)
  • Confirms the cash-on-delivery amount and handles return requests
  • After the call, saves customer, order, intent and a summary to Google Sheets

One n8n workflow does two jobs: it looks up the order during the call, and it logs the call after it ends.

Stack

  • Vapi
  • n8n
  • OpenAI GPT-4.1
  • Google Sheets
Read the call transcript

Sara: Hi, thanks for calling ShopEase! This is Sara. How can I help you today?

Customer: Hi Sara, this is Ahmed. I'm calling to check on my order.

Sara: Sure, Ahmed! What's your order number?

Customer: Sure, it's one zero zero one.

Sara: Let me check that for you. [n8n looks up order 1001]

Sara: Your Wireless Earbuds Pro order is shipped with TCS.

Customer: Oh nice. So when will it arrive?

Sara: Your order will arrive tomorrow, September 27th.

Customer: Perfect. And it's cash on delivery, so I just pay the rider?

Sara: Yes, Ahmed, that's right! Just pay the rider forty-nine ninety-nine when your order arrives.

Customer: Great, that's all I needed, thanks Sara!

AI voice agent packages and pricing

Demo call (1 min 24 s). The caller is an AI test agent; the store and order are sample data.
Google Sheets call log row created automatically after the demo call: customer Ahmed, order 1001, intent order status, resolved yes, with an AI summary
After the call: the summary lands in Google Sheets automatically.

Demo build

WhatsApp AI agent for e-commerce (English + Roman Urdu)

The problem

Customers ask the same order and product questions on WhatsApp all day. Replies come hours later, and staff copy-paste order details by hand.

What it does

  • A WhatsApp message triggers the workflow
  • An AI agent reads the question and looks the order up in a Google Sheets order list
  • Handles order status and tracking, product questions, and returns or refunds
  • Replies in English or Roman Urdu, matching the customer
  • Never invents order details: if it can't find the order, it says so

Stack

  • WhatsApp Business API
  • n8n
  • OpenAI
  • Google Sheets

WhatsApp automation for your store

Illustration of the demo agent's behaviour · sample data

Demo build

Low-stock inventory alert agent

The problem

Best-sellers run out and nobody notices until customers complain or ads keep spending on products you can't ship.

What it does

  • Checks the inventory sheet every 30 minutes
  • Filters items at or below their stock threshold
  • Skips items already alerted today, so the team isn't spammed
  • Combines everything into one short alert written by Gemini
  • Logs every alert sent

Stack

  • n8n
  • Google Sheets
  • Google Gemini

Low-stock alert workflow

Workflow steps from the n8n build, drawn as nodes.

Illustration · sample data

Demo build

Missed-call recovery system

The problem

A missed call is often a lost customer. Following up by hand is slow and easy to forget, and double messages annoy people.

What it does

  • A missed call triggers a webhook
  • The caller's details are validated first
  • Duplicate check: one recovery message per caller per hour
  • OpenAI writes a short, friendly recovery message
  • Every outcome is logged in Google Sheets: sent, invalid, duplicate skipped or failed
  • An instant alert fires if any step fails

Built as Project #1 of my HVAC automation series on LinkedIn. The demo sends through Telegram because Twilio trials aren't available in Pakistan; a live store version would use SMS or WhatsApp.

Stack

  • n8n
  • OpenAI
  • Google Sheets
  • Telegram

Missed-call recovery workflow

Workflow steps from the n8n build, drawn as nodes.

Demo build

AI receptionist system (with a live demo line)

The problem

Small businesses miss calls and lose track of who called about what. Owners want every call answered, logged and flagged, without extra staff.

What it does

  • Workflow 1: after each call to the AI receptionist, the call is logged to Google Sheets and an email alert goes out
  • Workflow 2: outreach emails go only to leads that were approved first
  • Workflow 3: replies are tracked and an AI drafts a response, but it never sends automatically

Try it: call the live AI receptionist demo on +1 (930) 216‑3103. You'll speak to an AI agent, not a person.

Stack

  • Vapi
  • n8n
  • Google Sheets
  • Gmail
  • OpenAI

AI receptionist call logging

Workflow 1 of 3, drawn as nodes.

Illustration · sample data

Client project

Website AI chatbot for an SEO agency

The problem

A Texas-based SEO agency wanted its website to answer visitor questions and capture leads at any hour, instead of relying on contact forms alone.

What it does

  • Answers visitor questions about the agency's services using its business information and a searchable knowledge base
  • Qualifies leads for its digital marketing services
  • Guides visitors to book a free SEO audit or contact the team on WhatsApp
  • Remembers the conversation, so visitors don't have to repeat themselves

Built for a client's SEO agency website. The workflow is live. No performance numbers are shown because none have been measured and shared yet.

Stack

  • n8n chat trigger
  • Google Gemini
  • Conversation memory
  • Supabase vector store

Website AI chatbots

Illustration · sample data

Website chatbot conversation flow

Conversation steps of the client build, drawn as nodes.

Other builds

  • E-commerce AI social media automation: a Gemini agent that writes and posts content to Facebook and LinkedIn (demo).
  • WhatsApp auto-reply with memory: Gemini with Postgres chat memory, so the bot remembers earlier messages (demo).
  • LinkedIn growth agent: an n8n AI agent I use for my own LinkedIn engagement.

Want one of these for your store?

Book a free 15-minute audit. We'll pick the demo closest to your problem and plan how it would work with your store, your data and your team.

Book a free 15-min audit on WhatsAppEmail me

Prefer to hear an AI agent first? Call the live AI receptionist demo: +1 (930) 216‑3103 (it's an AI agent, not a person).

Free audit on WhatsApp Email