An e-commerce store owner had 2,000 email subscribers and hadn't emailed them in 8 months. Here's what I built to help him.

A friend of mine runs a small furniture store. He built his email list over a few years, 2 000 people who'd opted in, who'd said yes to hearing from him. In all that time, he'd sent them maybe two emails.

It wasn't that he didn't care, quite the opposite. But he runs the whole business by himself, e.g. sourcing, logistics, customer service, the website, everything. By the time he'd handled all of that, figuring out what to send, when to send it, and then actually sitting down to write it felt like climbing a mountain he didn't have the energy for. So it kept getting pushed. And eventually it just stopped being something he thought about.

When he mentioned it to me, my first instinct was the obvious one: just use AI, it'll write the emails for you. Except that's not really the problem. Writing the email is the easy part. The harder question is which email to write, to whom, and why right now, and that's where generic AI completely falls apart.

It doesn't know that you have a new collection coming in three weeks, or that your customers tend to buy again around the four-month mark, or that a public holiday is coming up that's worth anchoring a campaign to. It produces content that could have been written for any store selling anything, and experienced email readers can feel that immediately.

I've spent most of my career in email marketing, and I know what makes campaigns actually work. Copywriting is certainly part of it, but what matters more is the strategy underneath. The timing, the sequencing, understanding who you're talking to and what they need to hear at that moment. That's what I wanted to build into a tool, not just a faster way to write generic copy.

So I built Envio. The original version had a feature I was fairly proud of, the idea being that the tool would flag campaigns where it needed more context from the store owner before it could generate the campaigns. Sounds cool in theory, but it became obsolete after I had build a section where you can set the direction of the campaigns yourself if you wanted, which was much more convenient.

The quality of the output was the thing I spent the most time on. It went through a lot of iterations before it stopped feeling like AI content and started feeling like something a competent email marketer had actually written for that specific store. That gap between "technically correct" and "good enough that someone would actually send it" turned out to be much wider than I expected.

My friend was the first person to use it properly. He sent his first real campaign in years. A few people bought something. More importantly, he didn't have to spend an afternoon figuring out what to say because the plan was already there, the copy was already there, and he just had to send it.

Envio is live now, and it works well enough that I'm genuinely comfortable putting it in front of people I don't know. It has a two-week free trial, no card required to start. But what I'm looking for at this stage is store owners who'll tell me honestly what's useful and what isn't, what they wished it did differently, and what they actually end up sending versus what they skip.

If you run an e-commerce store and email keeps falling off your to-do list, I'd genuinely like you to try it. You can do so at envioapp.com.

The thing I keep coming back to, though, is how much easier this whole process was because I was building it for a specific person with a specific problem. When I wasn't sure what a feature should do, I could just ask him. When I was tempted to add something that seemed clever, his reaction usually made it clear pretty fast whether it was actually useful. And on the days when building felt pointless, which happens, regardless of how good the idea is, knowing that at least one person genuinely needed it was enough to keep going.

I see a lot of post from people who built something because the idea sounded interesting, and now they're struggling to find anyone who cares. I don't think that's a distribution problem. I think it's that without a real person's real problem at the center of it, the product ends up slightly off in ways that are hard to diagnose.

Curious whether others here have started the same way, by solving something for someone specific, and whether that shaped the product as much as it did for me.