I Built a SaaS That Turns Any Website Into a Support Agent — Here's What I Learned

SiteWhisper — Fri, 24 Apr 2026 17:41:41 +0000

Three months ago, I had an idea. Six weeks ago, I shipped it. Today, I have 240+ people using it. Here's what I learned building SiteWhisper.

The Problem Nobody Was Solving

My friend runs an e-commerce store. Every day, she'd get the same questions over and over:

"Do you ship internationally?"
"What's your return policy?"
"Can I customize this product?"

She'd answer them over email, losing hours. Her support team would drown. Some customers would just... leave, because they couldn't get an instant answer.

I looked at the chatbot market. There were options:

Expensive tools ($500+/month)
Complicated setups (weeks to implement)
Free but they hallucinate (making up answers)

Nobody was solving for "I want to answer visitor questions from my actual website content, cheaply, easily."

So I built SiteWhisper.

The Architecture (Non-Technical)

Here's what it does: You give us your website URL. We crawl it. Your chatbot learns from what's actually there. When a visitor asks a question, the chatbot answers from your real content — no hallucinations, no guessing.

Copy one line of code into your website. Done.

The whole setup takes two minutes.

What I Got Hilariously Wrong

Assumption #1: People want lots of features.

When I started building, I thought users would want:

Custom logic
Advanced workflows
Multiple language models
Deep integration options

I spent two weeks building all of this.

Then I showed it to five early users.

They said: "Can you just... make it simpler?"

I was building for a market that didn't exist. I was solving for "what would an engineer want" instead of "what does a busy business owner need."

The lesson: Talk to users before you build. I cut 40% of the features I'd planned.

Assumption #2: Everyone wants the latest AI.

I was obsessed with using the most cutting-edge language model.

What I found: Users didn't care about AI being "smart." They cared about it being accurate.

They'd rather have a bot that correctly answers "What's your shipping cost?" from their FAQ than a bot that tries to be clever and gets it wrong.

We shifted our entire focus from "make the AI smarter" to "make the AI more accurate to what's on your site."

It was the best decision we made.

Assumption #3: This is a developer tool.

I thought the market was developers who wanted to build chatbots.

Actually, the market is non-technical business owners who want to reduce support tickets.

My whole go-to-market changed. We stopped writing technical docs and started writing guides for WordPress users and Shopify store owners.

What Actually Worked

Speed of setup. No single thing matters more than being able to set it up in two minutes. If it takes 30 minutes, people abandon it.

Honesty about limitations. We're not promising to replace your support team. We're promising to handle the easy 40% of questions so your team can focus on the hard 60%.

Free forever. People don't trust new tools. But they'll try free ones. Once they're using it, they see the value.

Actually solving a problem. Every feature we ship answers a specific user request. No vanity features.

The Launch

I posted on Product Hunt. It wasn't fancy. Just a straightforward post: "I built a chatbot that takes 2 minutes to set up and costs $0."

First day: 80 signups.
By day three: We were trending.

The interesting part? Most signups came from small business owners, not engineers. That told me we'd actually found the right market.

The Surprising Parts

Traction ≠ Revenue.
We have 240 users. 60 are paying ($19-49/month). But the real value? The free users are evangelizing it. They're telling friends. That word-of-mouth is worth more than I could afford in advertising.

Different markets than expected.
I thought: SaaS and e-commerce.
Actually using it: Agencies, service businesses, content creators, even nonprofits.

Each segment has different needs. Once I stopped trying to serve everyone and focused on one segment, things got clearer.

People care about results, not how it works.
Nobody asks me "what AI model are you using?" Everyone asks "will this reduce my support tickets?"

Marketing lesson: Sell the benefit, not the technology.

The hardest part isn't building, it's listening.
Shipping was fast. Listening to feedback, iterating, killing my favorite features because users didn't want them — that was hard.

What's Next

We're not profitable yet. But we're sustainable (costs are low, runway is long).

I'm thinking about:

How to help users get more value (better onboarding, more integrations)
When to charge more aggressively (or if to stay free forever)
Whether to hire or stay solo (I think solo for now)
How to compete when bigger companies copy this (they will, and that's okay)

The Thing I Wish I'd Done Differently

I wish I'd talked to 20 potential customers before building anything.

Instead, I built what I thought people wanted. Luckily, I built something close to what they actually wanted, so I recovered. But those first two weeks of building unnecessary features? I'd do that differently.

The shortcut is: Talk to customers first. Build second. Ship third.

For Fellow Builders

If you're thinking about building a SaaS:

Solve a problem you or someone you know has. Don't build to an abstract market.
Ship fast. You'll learn more from users than from thinking.
Stay small. Profitability > growth for as long as possible.
Be honest. Your product isn't for everyone. Find who it's for and serve them.
Listen ruthlessly. Users know better than you.

Most startups don't fail because the product is bad. They fail because the founder wasn't willing to listen when users told them it was wrong.

The Real Metric

I don't measure success by revenue or users.

I measure it by: "Is this actually helping people?"

The emails I get saying "We cut support tickets in half" or "Our chatbot captured a lead at 3am" — that's the metric that matters.

The rest is just business.

Article End

Author Bio (typically auto-filled but include)

Avinash is the founder of SiteWhisper. He's built and sold a SaaS before, and is currently building in public. You can follow his journey at [twitter link or website].

Call-to-Action (Optional, Dev.to allows)

Try SiteWhisper for free (no credit card required): https://sitewhisper.online

Forem: SiteWhisper