I Started Building a Simple AI Tool as an Experiment

Here’s What’s Happening So Far

Created with Gemini Nano Banana 2

I didn’t set out to build an app.

A friend told me people in her niche hate marketing. Most of what they send out comes straight from vendors. The problem is obvious — those emails all sound the same and don’t sound like the person sending them.

That stuck with me.

I kept asking the same question. What if you could take that generic vendor content and turn it into something that sounded like you, without having to write it from scratch?

Where this started

I opened a chat and, using WisprFlow, talked out the idea exactly as it was in my head. No structure. No cleanup first. It was a rambling mess.

The AI didn’t make it more complex. It did the opposite. It kept asking me to narrow the scope and focus on one outcome instead of trying to solve everything at once.

The idea shifted fast. From “some kind of marketing tool” to something much smaller. Paste vendor content, get a draft email back that sounds like you.

This felt doable instead of overwhelming.

Granted, this can be done by simply pasting the generic content into a prompt and asking for it to be rewritten in one’s voice. But that’s not the point here. This is practice so I can see how to build with code on a local host.

What I’m building right now

I’m keeping the scope tight on purpose.

The flow is simple. You paste vendor content. You describe the tone you want. You click a button. You get a draft email you can use. No login. No saved data. No extra features yet.

That wasn’t an accident. Every time I thought about adding something, I pulled it back and asked what the smallest useful version could be.

That question has been doing most of the work.

The part I thought would slow me down

I expected the technical side to be the sticking point. Mostly because I don’t come from a development background.

What’s happening instead is I’m learning just enough to move forward. I’m not trying to understand everything at once. I’m asking very specific questions, getting an answer, applying it, and moving to the next step.

When something breaks, I paste the error into the chat and ask what it means. That alone has made this feel way more accessible than I expected.

If you’re wondering if you could do this

The short answer is yes, but not by trying to figure it all out upfront.

What’s working for me is starting with a small idea and letting the process unfold one step at a time. I’m not building a full product. I’m testing a simple use case and seeing where it leads.

What I didn’t expect

Two things have stood out.

First, I had to get a basic sense of how things connect. Frontend. Backend. API. I’d heard these words before. Now I’m seeing how they relate to each other in a real project.

Second, I’ve had to get comfortable moving forward without full clarity. There are moments where I don’t fully understand what I’m doing, and I keep going anyway because I know the next step is enough.

That shift has been bigger than anything technical.

Where I’m stuck right now

I’ve got pieces working, but they’re not fully connected yet. The backend can generate the email. The interface is there. But getting them to talk to each other has been the challenge.

The issue is something called a CORS error. Basically, the browser is blocking the connection between the two parts of the app.

I haven’t solved that yet, and I’m okay with that.

The bigger takeaway is I’ve gone from a rough idea to something that’s partially working. I can see how the rest will come together.

One simple action you can take

If you want to try this yourself, start with one small idea and open a chat.

Write something like: “I have a rough idea for a simple tool. Help me turn it into the smallest version I can test. Ask me questions to clarify it.”

Then follow the questions and resist the urge to make it bigger.

That’s how this started for me, with the help of Ellen Britt, PA, Ed.D., PHom, who is guiding me on this new path.

What matters next

Building it is one part. What matters more is whether someone would use it.

My next step is to show this to a few people and see how they respond. I want to see what makes sense to them, what doesn’t, and where they hesitate.

That’s the part that will shape what happens next.

I’m inspired by many people who are using AI to build useful tools for themselves and others.

My AI Builder Crushes on Substack

Have you, or are you, building tools with AI? I’d love to hear from you… what are you creating? What has your experience been like? Please share, I want to follow your journey!

Ellen Britt and I are putting together a small beta group of people who are interested in learning how to build AI-powered apps from scratch. We’re working on the details, and if you’re interested in learning more, get on our “first to know” list here. No obligation, just info so you can see if it’s something you’d like to learn.

Click here to get on the “first to know” list.