I almost built the wrong thing entirely.
For months, I had this idea sitting in the back of my head. A tool that would analyze a female founder's website and give her an honest outside perspective — the kind of feedback I give clients in paid audits, but in an automated, free format.
Simple concept. Except I had convinced myself I needed to build an AI agent to do it.
An AI agent. Not a straightforward automation. An agent. Because that's what everyone on LinkedIn was talking about in late 2024, and somehow my brain decided that if I was going to do this, I had to do it the complicated way.
So I spent months not building it. Researching. Overthinking the architecture. Having low-key anxiety at night about whether I even knew enough to build what I imagined.
The moment it clicked (and embarrassed me a little)
One day, in a comment reply to someone's LinkedIn post about "agentic fever," I wrote this:
I remember when I decided to build an AI-powered Website Scanner for my ICP, it took me months of postponing it and overnight anxiety. Once I started, I realized I don't even need an AI agent for my use case, but more of an AI Automation. For a non-techy mortal it sounds the same, but it is not.
And that's the whole story, honestly. The anxiety was useful. It forced me to stop and ask: do I actually need an AI agent, or do I just need a working solution?
Turns out: working solution.
What I actually built (and how)
The technical stack ended up being embarrassingly simple compared to what I had imagined:
- A Webflow form — collects the website URL, a one-sentence business description, who their ideal client is, their biggest website frustration, and their name and email
- Jina Reader API — fetches the submitted website's content as clean markdown (free, no authentication needed, just a GET request to
r.jina.ai/[url]) - Claude API via Zapier — takes the website content + form answers and generates a personalized 4-section mini-audit report
- Gmail via Zapier — sends the report directly to the person who submitted the form
That's it. Four steps. Webflow → Jina → Claude → Gmail.
I built the whole thing in about two days.
The real intellectual work was the prompt — a system prompt that contains my entire audit framework, scoring guidelines (1–10 clarity score), tone instructions, and constraints about what the mini version should and shouldn't include. That took longer than the automation itself, which honestly feels right. The thinking matters more than the tools.
Zapier Copilot vs. actually debugging it myself
Here's something no one tells you about building with AI-assisted automation tools: the AI that's supposed to help you build sometimes creates more confusion than it solves.
I used Zapier Copilot (their AI assistant) to help set up the automation. And it was... fine. It got me 80% of the way there. But when errors happened — and they did — I found myself ignoring Copilot and going straight to Claude to debug the JavaScript code.
One error that took me embarrassingly long to solve: Cannot read properties of null (reading 'url') — thrown by the Email by Zapier step. Copilot kept suggesting generic fixes. Claude looked at the actual code, understood the data flow, and pointed to the exact line where the response parsing was failing.
Lesson: AI tools are not interchangeable. Some are better at generating code. Some are better at debugging it. Know which is which before you need them.
I built it. Then I hid it.
This is the part I'm slightly embarrassed about but also the most important part of the story.
After the automation worked — like, actually worked, emails going out, reports being generated, everything functioning — I sat on it for two weeks.
I told myself it needed more testing. Maybe the prompt needed refining. Maybe the email format wasn't quite right. Maybe I should add one more section.
My coach Ricardo cut through it in about 30 seconds:
"You built a complex lead magnet that required learning new technologies. You are making decision after decision. You are just now stuck in the thing that messes with most of us: putting yourself out there."
He was right. It wasn't about the tool. It never is.
The first people I sent it to
I did the thing you do when you're not ready to go public: I shared it quietly with a few women I trusted.
Anna's response: "The scanner is very cool. It left me wanting more, which is the point."
Sadia's response: "It is already very rich."
The fear was louder than the reality. It usually is.
A few weeks later, Anna (who later became my client) shared the scanner with her own community. Unprompted. Before she'd even paid me for anything.
That's what a good lead magnet does. It travels.
What the scanner actually does (and doesn't do)
The Website Clarity Scanner is a free tool. You give it your URL, one sentence about your business, who your ideal client is, and your biggest website frustration. It uses that context — plus an actual analysis of your homepage — to generate a personalized mini-report covering:
- First Impression (5-second test) — Is it immediately obvious who this is for?
- Positioning Clarity — Is there a specific value proposition, or generic coaching-speak?
- SEO/AEO Snapshot — Can Google (and AI search tools) actually understand what you do?
- Conversion Path — Is there a clear next step, or are there 7 competing CTAs?
What it deliberately doesn't do: give you implementation advice, prioritize what to fix first, or tell you about technical errors. That's what the paid Website Clarity Audit is for.
The mini version diagnoses. The full version treats.
The actual lesson
I spent months not building a tool because I was imagining a more complicated version of it. Once I stopped trying to build the impressive version and started asking "what's the simplest thing that actually solves the problem?", I built it in two days.
The hiding was about visibility, not readiness. It usually is.
If you're sitting on something because it's "not ready yet" — honestly, ask yourself: is it not ready, or are you not ready to be seen?
Want to try the scanner? Check it out here. If the report makes you go "oh sh*t, this is exactly what's wrong" — that's when the paid Website Clarity Audit might be worth talking about.