n8n automation & AI front desks · Kolkata, working worldwide

Most leads wait hours for a reply. Mine get one in 4 seconds.

I'm Drishti Sen. I build small-business automations in n8n, and I write for a living, so the replies they send still sound like a person. Every system on this page is live, and every result shows the date of the real run behind it.

Watch the 42-second demo See pricing

Or . It's one of the systems.

Lead Capture Automation run #4 · 16 Jul 2026
Succeeded in 3.933s
3.933 s

A replay of run #4, 16 Jul 2026. It lasts 3.933 s, exactly as long as the run took from form submit to sheet row, owner alert and the lead's confirmation email.

3.9s

form lead → sheet row, owner alert and the lead's reply email

run #4 · 16 Jul 2026
2.2s

WhatsApp question → answer sent to the customer

run #2319 · 28 Sep 2026
18.7s

one brief → 9 content drafts, digest sent, calendar filled

19 Jul 2026
9.8s

customer inquiry → AI reply, sent and logged

run #8 · 17 Jul 2026
20 workflows live on my own n8n · 618 production runs in the 7 days to 28 Sep 2026, 2 failed Built with n8n · Google Sheets · Gmail · Gemini · WhatsApp Cloud API · Apify

Live demo · WhatsApp front desk

A message lands at 1:08 a.m. The reply leaves 2.2 seconds later.

My WhatsApp front desk for a fictional Kolkata gym, recorded on 28 Sep 2026. It answers from the owner's Google Sheet in English, Hindi or Bengali, books the free trial, and puts the lead in the owner's inbox. Nobody has to be awake.

Real screen recording · typing trimmed, every reply time real · Meta's test number

Gym Front Desk — WhatsApp · run #231928 Sep 2026 · 01:08:20 IST
  1. 0.036sWhatsApp Triggermessage in
  2. 0.036sRead Message6 ms
  3. 0.042sLoad Info · Google Sheet473 ms
  4. 0.515sPrepare Turn7 ms
  5. 0.523sAsk Gemini1,043 ms
  6. 1.566sParse & Remember8 ms
  7. 1.574sSend Reply · WhatsApp659 ms
  8. 2.234sLog Question · Google Sheet972 ms
  9. 3.208sSave Lead · Google Sheet1,001 ms
  10. 4.209sEmail Owner · Gmail676 ms
✓ Succeeded in 4.885s
Only from the sheetIf the sheet doesn't say, it says so and offers to pass the question on.
Their languageEnglish, Hindi or Bengali, including Hinglish typed in English letters.
Official WhatsApp APIOn the owner's own Meta account. Replies only to people who message first.
Try the web version, no signup Try the web version, no signup WhatsApp front desk pricing

Systems I've built · and still run

Six systems. All live in my n8n.

Each card is a real workflow; the chips are its actual steps. Open any card for the build notes and, where I have one, the original screenshot of the run.

Live · try it

AI front desk, on the web and WhatsApp

Answers customers at any hour from the owner's own Google Sheet. It never guesses a price: if the sheet doesn't say, it says so, offers a human and flags the gap. Booking requests land in a Leads sheet and the owner's inbox.

WhatsApp Trigger→Read Message→Load Info→Ask Gemini→Send Reply→Log Question→Save Lead→Email Owner
Try the web demo → Try the web demo → Watch the WhatsApp run
2.2 squestion → reply sent · run #2319, 28 Sep 2026
How it's built

The problem. Chatbots that answer everything are a liability: they invent prices and policies. Chatbots that answer nothing are a phone tree. Small businesses need the middle: instant answers from their own facts, and honesty about everything else.

The build. n8n with a chat or WhatsApp trigger, code-based memory and guards, and Gemini under a prompt that forbids inventing facts. Contact details are only accepted if the customer actually typed them. Every question is logged with an answered-from-the-sheet flag, so the owner sees exactly where the sheet has gaps.

Where it started. The first version ran on Telegram, with production runs of 8.6 and 9.6 seconds tested from a real phone: the opening-hours question answered exactly from the sheet, the yoga-mats question honestly deflected.

The Telegram version: the real conversation and the n8n production log side by side
The Telegram version: the real conversation and the production log in one frame.
Live

Lead capture

Any form posts to a webhook. The lead lands in a Google Sheet, you get an alert, and the lead gets a personal confirmation before they've closed the tab.

Receive Lead→Format Lead Data→Append row in sheet→Send Lead Notification→Send Confirmation to Lead
3.933 sform → both emails · run #4, 16 Jul 2026
Why it matters

Firms that tried to contact a web lead within an hour were nearly seven times as likely to qualify it as those that tried even an hour later, yet only 37% of 2,241 U.S. companies audited replied within the hour (Harvard Business Review, 2011).

n8n execution view: Lead Capture Automation succeeded in 3.933 seconds, every node green
The run in n8n: Jul 16, 01:54:42, succeeded in 3.933s.
Live · automation + writing

Voice-matched content engine

One blog post, transcript or voice note in; 3 LinkedIn posts, a newsletter section and 5 short posts out, drafted against a voice profile I write from the client's real writing. Nothing publishes without approval.

New Content Input→Extract Input→AI Ghostwriter Engine→Parse Content Pack→Send Content Digest→Save to Content Calendar
18.754 s1 brief → 9 pieces · 19 Jul 2026
The run, and a spec demo

Latest live run: one 250-word finance brief in, nine publish-ready pieces out, digest delivered and calendar filled, all marked Draft for review.

When a finance blog posted a job, I pushed a cost-of-living brief in their niche through this engine and attached the run to my proposal the same day, instead of just claiming I could build it.

n8n execution of the Voice-Matched Content Engine, all nodes green, 18.754 seconds
The execution: every node green, 9 items into the calendar. Jul 19, 10:10:02.
Live

AI inquiry assistant

Classifies every incoming message as general, booking or complex. Routine ones get an instant reply in the business's tone; anything sensitive goes to the owner as a ready-to-edit draft instead of being sent.

New Inquiry→AI Classify and Draft→Log Inquiry→Route by Complexity→Send AI Reply/Alert Owner
9.756 sinquiry → reply sent and logged · run #8, 17 Jul 2026
Published

Carousel factory

A brief goes in. AI writes the slide copy, each slide is rendered to an image, and the set arrives as an email digest, logged as Draft for review. No Canva.

New Carousel Request→AI Carousel Writer→Split Slides→Render Slide Image→Email Carousel Digest→Log Carousel Slides
6 slideswritten, rendered to images and logged, from one brief
See the workflow
The Carousel Factory workflow in n8n, eleven nodes from brief to logged slides
Carousel Factory in n8n: 11 nodes, from brief to logged slides.
Behind the scenes

My own business runs on 20 of these

  • Lead Gen Engine. Every week, pulls new local businesses from public Google Maps listings and filters them.
  • Outreach Engine. Writes a personal first email for each one, as a draft I read before it goes.
  • Reply Checker. Checks my inbox every 15 minutes and drafts an answer to each prospect who replies.
  • Mission Control. A watchdog that keeps an eye on the rest and tells me when something fails.
618production runs in the 7 days to 28 Sep 2026 · 2 failed

How a project goes

From your message to a system that runs.

No discovery documents, no six-week onboarding. Four steps, and you watch it work before you pay the second half.

01Trigger

You send one link or one problem

A form that leaks leads. An inbox that's always behind. No brief needed.

02Call

Free 20-minute call

I tell you exactly what I'd build and what it costs, then send a one-page quote the same day.

03Build

I build it on your accounts

Your n8n, your Google account, your API keys. Flat price, 5 days to 2 weeks.

04Test run

You watch a real run succeed

End-to-end test before handover. Then plain-language docs and a week of free fixes.

Your accounts, your system

Built on your n8n, Sheets and API keys. No dependence on me, no subscription to me.

Tested end to end

You watch a real run succeed before handover, screenshots included.

Plain-language docs

A walkthrough your team can follow without me on the call.

A week of free fixes

If something breaks in week one, it's mine to fix.

Why me and not a cheap monthly tool? The tools automate volume, but they sound like machines, and your customers can tell. I'm a ghostwriter who builds automation: the workflows run themselves, and the words still sound like you.

Pricing

Flat, scoped, no surprises.

50% to start, 50% after you've watched a real test run. The advance can be paid by UPI or bank transfer. Custom workflows are quoted after a free 20-minute call.

Automation

Any single automation

$300₹15,000
5 days

Pick the one workflow that's costing you most: a form nobody watches, an inbox that's always behind, a report you rebuild by hand. Built end to end on your own n8n and Google accounts.

Start here

Never Lose a Lead stack

$500₹25,000
7 days

Lead capture and an AI assistant, built to work together. Every enquiry logged, you notified, the lead answered in seconds; anything sensitive comes to you as a draft.

New

WhatsApp front desk

$450₹22,000
7 days + Meta's approval

The front desk on your own WhatsApp number, on Meta's official API. It only replies to people who message you first; Meta bills its message fees to you directly. See it working

Chat assistant, website or Telegram

$350₹18,000
5 days

Answers only from your own Google Sheet, admits what it doesn't know, and flags the exact question so your answers grow. The assistant on this page is one.

Voice-matched content engine

$400₹20,000
7 days

One input, nine drafts in your voice, delivered as a digest and logged to a content calendar. Nothing publishes until you approve.

Optional human edit: +$150/mo+₹7,000/mo

Full stack, all four systems

$900₹45,000
2 weeks

All four builds as one set: shared sheets, one consistent voice, one handover session. The cheapest price per system.

Writing

LinkedIn profile & resume makeover

$50–80₹2,500–4,000
48 hours

Headline, About and experience rewritten so a recruiter gets what you do in the first two lines. ATS-safe resume, before and after side by side.

LinkedIn content sprint, 8 posts

$150–200₹7,000–10,000
2 weeks

One 20-minute call to pull out your stories, then eight posts in your voice. You approve every one, and you keep the drafts either way.

The writing side

Automation that still sounds like a person.

I write for a living. Two short posts in the style I write for clients: hook in the first line, one specific observation, plain language, a question to close.

Drishti SenGhostwriting sample
Contrarian insight

Most LinkedIn advice is about posting more.

Posting more is the easy part. The hard part is saying the one thing only you would say. Without it, volume just trains people to scroll past your name.

What would you say that nobody else in your field would?

Drishti SenGhostwriting sample
Behind the build

The hardest thing I've built is a bot that says “I don't know.”

Answering is easy: any AI will answer anything. Stopping is the work. Admitting the gap, and handing the question to a human. Most bots skip that step, and their customers pay for it.

Would you trust a bot that never says “I don't know”?

Drishti Sen
Drishti Sen · Kolkata

About

I build for the failure case first.

Most automation fails in one of two ways. It breaks quietly, and nobody notices until a month of leads has gone missing. Or it works too well: it answers a pricing question it should never have touched, in a voice that leaves your customer feeling handled by a machine.

So every system here has a designed path for “I don't know”: the bot that admits it and emails you the exact question, the classifier that hands anything sensitive to a human, the content engine where nothing publishes until you approve it. Then I write the voice, because a workflow that runs perfectly and sounds like a robot has only solved half the problem.

I'm a working ghostwriter who builds automation, based in Kolkata and working with clients anywhere. If I can't show you it running, I don't claim it.

Kolkata · ISTIndia and abroadReplies within hoursEvery number dated

Questions

Before you message me.

Do I pay you every month?
No. You pay a flat price for the build. The only running costs are the tools themselves: your n8n plan, a little AI usage, and for WhatsApp, Meta's own message fees. I list each one in the quote.
Whose accounts does it run on?
Yours. I build on your n8n, your Google account and your API keys, so you own everything and nothing depends on me.
What if something breaks?
You get plain-language documentation and one week of free fixes after handover. If something breaks in that week, it's mine to fix.
Will WhatsApp ban my number?
The front desk uses Meta's official WhatsApp Cloud API on your own Meta account, and it only replies to people who message you first: no broadcasts, no bulk messages, none of the unofficial tools that get numbers banned.
Which languages can the assistant reply in?
The demo replies in English, Hindi and Bengali, including Hinglish and Banglish typed in English letters. It answers in the language the customer writes in.
How do we start?
Send one link or one problem. We have a free 20-minute call, I send a one-page quote the same day, and I start once the 50% advance is in.

Start a project

Send me one link or one problem.

A form that leaks leads. An inbox that's always behind. A blog post that should be a week of content. I'll tell you exactly what I'd build and what it costs, and I'll often run a free sample through a live system before you pay anything.

Email me WhatsApp LinkedIn

drishtisen16@gmail.com · I reply within hours (IST)