Building an AI Tool That Saved 5 Hours a Week

The situation

Alex, the owner, liked to reply to Facebook comments personally. It mattered to him. These were customers asking questions, and he wanted to answer them himself.

When ad spend was low, this was fine. A few comments a day. No problem.

But as we scaled the ads, the comments scaled too. Hundreds per week. Alex was drowning. He'd spend hours scrolling through the Ads Manager trying to find and reply to everything.

Something had to give. Either he stopped replying, or we found a better way.

What I built

I built him a custom tool. From scratch. In a day.

It pulled all the comments from every ad into one place. He could see them all, sorted by ad. Click on a comment, see the context, reply, done.

But here's where it got interesting.

The tool used AI to draft responses for him. Not generic chatbot rubbish. Responses modelled on hundreds of his own previous replies.

I trained it on his tone of voice. How he actually talks to customers. The way he phrases things. His personality.

When a comment came in about a specific product, the AI would pull in the relevant product information. It knew what the customer was asking about.

Alex could review the draft, tweak it if needed, and post with one click.

What made it work

It was built for him. Not a generic tool. Not a plugin. Something designed around how he actually worked.

The AI sounded like him. I trained it on his real responses. Customers couldn't tell the difference.

It removed friction. He wasn't scrolling through Ads Manager anymore. Everything in one place. Draft ready. One click to post.

Results

Five hours a week saved. At least.

Alex got to keep doing what he cared about, replying to customers, without it eating his whole day.

The comments still got answered. The personal touch stayed. The ads kept scaling.

Why this matters

I've been using AI tools since before ChatGPT launched. I'm not interested in AI for its own sake. I'm interested in solving problems.

AI in the wrong hands is just as bad as someone who doesn't know what they're doing. But when you understand the problem properly, and you build something tailored to it, AI can be genuinely useful.

This tool worked because I understood Alex's workflow. I understood what he was trying to protect. And I built something that helped him do it better.