Copybot — from zero to 1,100+ paying customers

Problem

Copykiller runs a copywriting course that thousands of people have taken: copywriters, e-shop owners, marketers. They knew the method; they had no tool to apply it every day. Companies without a copywriter had the same gap.

Decision

Give the Copykiller course graduates real value in the form of an app for writing their copy. It was also about increasing the value of the product (the course) with a dedicated app and raising Copykiller’s revenue — customers don’t need to pay for ChatGPT and the like, they buy our solution instead.

What I built

The whole product: application, AI layer, cloud infrastructure, onboarding, pricing and operations. Copybot is owned by Copykiller; I led and built it. The entire pipeline from development to deployment was driven by AI coding agents under engineering supervision: domain model and tests first, agents generating code under those tests, human sign-off on architecture and anything touching data or money. It is the reference project for how I build software today.

Result

1,100+ registered customers; the product runs with minimal operations cost.

What I would do differently

On this project we tested a number of modern platforms and frameworks, and later ran into incompatibilities between them. Today I would be more restrained in those decisions and lean toward proven technologies.