FiveMinChat

A bit of background

I run a web agency in Kolkata, India. Have been doing it for over a decade. We build websites, plugins, dashboards - the usual stuff. It pays well, keeps a team employed, and honestly isn't the kind of thing that makes for an interesting IndieHackers post.

This isn't that post. This is about the thing I built on the side.

The problem I couldn't unsee

After building hundreds of websites for small businesses, I started noticing something that bothered me more every time I saw it.

A dental clinic. A real estate agent. A coaching center. We'd spend weeks getting their website right - good design, fast load, everything where it should be. They'd be genuinely excited. And then within a month they'd be right back to answering the same five questions on their phone, manually, every single day.

"What are your charges?" "Do you cover my area?" "Are you open on Sundays?"

The website looked great. But when a customer visited at 9pm with a real question, it had nothing to say.

I sat with that for a while. AI chatbots existed, sure. But every one I looked at was built for companies with IT departments and onboarding budgets. The small business owner I was thinking about — the dentist in Ballygunge, the plumber in Houston - doesn't have any of that. They just want it to work without having to read documentation.

So I decided to build something that would. I called it FiveMinChat. The name is the promise: five minutes from signup to a live, trained chatbot on your website. No developer needed. No long setup.

Building it while the agency kept running

I won't romanticise this part. It was messy.

Client deadlines don't move because you're excited about a side project. The agency kept running - meetings, deliverables, team stuff - and FiveMinChat got the early mornings, late evenings, and the scattered hours in between.

What helped was keeping the infrastructure lean from day one. I self-hosted Qdrant on the same VPS already running the agency's n8n workflows. OpenAI's gpt-4o-mini for chat, text-embedding-3-small for retrieval. Supabase for the database. A vanilla JS widget that loads with one script tag. The whole thing cost almost nothing to run, which meant the unit economics held even at $5.99/month - my lowest plan.

I wasn't trying to build the most impressive stack. I was trying to build something that worked reliably and didn't require a second mortgage to operate.

The part where I got it completely wrong

I assumed building the product would be the hard part. I was wrong.

The hard part was figuring out who I was actually talking to - and how to say something they'd care about.

My first three weeks of outreach were pretty dismal. I was sending emails to "small businesses with websites" which, as a target audience, is essentially everyone and no one. The emails were too long, too polished, and too focused on explaining what the product did. Nobody replied.

Then I narrowed down. Independent plumbing and HVAC contractors in the US. Owner-operated. Usually the person who answers the phone is also the person who does the job. I started researching each company individually before writing a single word - reading their website, checking their reviews, looking for one specific thing that was true about their business and hadn't been said to them before.

That changed everything. Not the product. Just the message.

What I learned about cold email the hard way

The emails that actually got replies were short, specific, and a little uncomfortable.

Not uncomfortable in a pushy way. Uncomfortable like when someone points out something true that you'd half-noticed yourself but hadn't said out loud.

The best subject lines I wrote weren't clever. They were just honest. "Your site says 24/7 - does it answer questions after 5pm?" works because it names a real tension the owner already feels. "AI chatbot for your plumbing business" doesn't work because it sounds like every other pitch they've ever ignored.

The body was never more than 6-8 lines. No feature list. Just: here's the specific thing I noticed, here's the one problem it creates, here's what I built to fix it. And then one simple ask - not a demo call, not a 30-minute conversation. Just: "Reply 'Yes' and I'll send you a private link where you can chat with a bot trained on your own website."

That last bit matters more than anything else. They're not watching a generic demo. They're watching their own business answer questions correctly, in real time, on their actual website. That's the moment people get it. Everything in my outreach is built around getting someone to that moment as quickly as possible.

Where things actually stand

I'm going to be honest here because I think the polished-numbers posts don't help anyone.

Outreach is working. Conversations are happening. People are trying it.

The thing I'm still figuring out is converting trials to paid customers. Getting someone to try it has been easier than I expected. Keeping them engaged long enough to reach that aha moment — where the chatbot answers a real customer question on their live website and they see the value with their own eyes — that's the work in front of me right now.

My theory is the aha moment has to happen in the first session, before they close the dashboard and forget it exists. So that's what I'm designing around.

I'll share proper numbers when there are numbers worth sharing.

The weird advantage of not going all in

I know the narrative is supposed to be "I burned the boats and went all in." That makes a better story. But it's not what I did.

The agency is still running. It pays the bills. FiveMinChat gets built in the gaps.

There's something I've found genuinely useful about this, beyond just financial stability. When you're not betting everything on one outcome, you make calmer decisions. You don't rush pricing because you need revenue this month. You don't say yes to every feature request because you're afraid of losing a customer. You can actually think.

That might change as FiveMinChat grows. But right now the dual track is working and I'm not going to pretend otherwise.

What I'm working toward

Fifty paying customers. That's the number I've set as the first real milestone - not because it's a meaningful business yet at that scale, but because it's enough to stop making decisions based on assumptions and start making them based on patterns.

Beyond that, I want to understand which segment sticks around and which churns. Right now I'm selling to any SMB with a website. That's fine for early validation. It's not a strategy you can build a company on. The next few months are about finding the answer to: who actually keeps coming back?

If you've navigated trial-to-paid conversion in an SMB product, I'd genuinely love to hear how you thought about it. That's the problem I'm living in right now.

And if you run a small business and your website goes quiet after hours - that's exactly who FiveMinChat was built for. Seven-day free trial, no card needed, starts at $5.99/month.

fiveminchat.com