What I learned from the head of growth at AstraAI
They built an AI app that teaches you math.
Early days
In the beginning it took them about three days per video to understand what was working. Three days to establish, this one brings users and that one does not. They tried building custom solutions and then they they gave up on building features. Why? They saw AI getting smarter, fast. So the decision was made - stop posting from a company account and open a bunch of fresh TikTok accounts. From there it was pure growth… hundreds, thousands, and now millions of users.
How they grew
They tried influencers and quickly stopped. Too expensive. So they turned to small creators they could build up themselves, sometimes from zero.
The best ones were were young, willing to be on camera, with some sense for graphics. Videos with a female performed way better than those with men. Videos that felt a little dumb beat the ones that looked like advertisements. Young viewers can smell an ad from across the room, especially when an older person makes it.
To build such system took a lot of work. They first found people on Upwork, took one larger creator and let them teach the smaller ones, like a country manager for content. You watch the competition and take their creators when you can. You pay through Useme, Ksolgo, Grade. When a video works in one market, you rebuild it for another. The format travels, the language changes.
Geography is surprising. India brings them users. America brings them nothing, no matter what video they put out.
How the company is built around this
Two teams, two numbers. The product team owns the whole experience, from landing to wherever the user ends up. The growth team owns traffic. Each team has one measure that matters.
Tools are ordinary. Shortimize for videos, PostHog for the rest. You look for tools first, before you build anything yourself.
Onboarding is its own discipline. Long, smooth, the card comes at the end, the paywall is hard. Kal AI is the example worth studying. A long flow that does not feel long. Then one-time offer of ten euros per month for the whole year that closes the deal. A long onboarding pulls more commitment out of a person than a short one.
Extras
- Agency content supports the user content.
- The moment you take venture money you owe somebody growth.
- At the start you track every video and know which one brought which user. Once you scale, that ends. One good video, copied across markets, does more than ten new ones.
- A single feature that help them grow the most was dead simple - allow users to upload their own study notes (and make it public) - now they are sitting on piles of curriculum data. This is quickly becoming their moat.