How Praktika Built a Profitable $50M ARR AI Business — Adam Turaev, Praktika

Used a report on English language penetration ('English first report', likely EF's proficiency index): Brazil was among the least English-proficient countries, English is a big income multiplier there, the population is large and people already learn English. Main worry was weak purchasing power and low willingness to pay for consumer subscriptions; went anyway, believing most early adopters were there. The app launched worldwide with no geo limits, English only, but all distribution effort focused on Brazil.

In Brazil, where people rarely pay for consumer subscriptions, every claim was about getting a better job rather than getting better at English. The ~$100/year subscription was justified against the cost of a human tutor; the '50X cheaper than a tutor' claim came out of this go-to-market and is now used across the category.

With no budget for big creators, used the creator search tool Modash to cap following at under 5,000 (big followings mean high fees) while filtering for creators currently getting lots of views, i.e. ones the algorithm is pushing. Found Gustavo, a Brazilian educational creator hitting tens of millions of views with no business email listed. Adam frames it as a reusable filter: high views relative to following means you reach creators before they peak.

Before outreach automation tools existed, the founder spent about a month full-time sending emails and messages to a long list of creators. Where a creator had no collaboration email (as with Gustavo), he DMed them directly on TikTok: 'we're building this app and it seems like we have a good match, so do you want to work together.'

Gustavo, a Brazilian who knew the pain point, made a video where an AI coach simply corrects a learner's mistakes in a short conversation, without pitching or showcasing the app. Adam disliked it and asked him to talk about the app; Gustavo said 'trust me' and Adam let him post. Standing rule since: do not tell influencers how to do their job; you hire them for their reach, so let them do it.

As soon as the correction video worked, Praktika had other creators in other countries remake the format with various angles, scaling it across the globe instead of staying focused on one country.

Ran a tiny paid budget behind the viral correction-format video in the month it broke out.

Right after the viral video, Adam hired Gustavo as an employee. He kept making videos, became the face of Praktika and runs its social media as an internal influencer, a separate stream from the contracted creator program.

In-house influencer lead (a creator from Poland who made videos for Praktika before being hired) runs the program worldwide; in-house influencer coaches, also ex-creators, each run a country and a roster of 20-30 contracted creators, rotating angles. Rationale: knowing what good content is is the key skill; management can be learned. Creators stay contractors, not employees. Runs on SideShift plus internal add-ons for analytics and reporting.

Typical deal: a fixed part plus a bonus tied to the number of views. Because coaches manage creators, terms are flexible: for strong talent they do whatever deal it takes to sign them. No amounts given.

The organic creator program and paid acquisition are separate streams so creators keep coming up with their own ideas and angles (the in-house team runs out of angles). When a creator video goes viral, it is shared with the paid team, who reproduce it with their own roster of internal influencers and content-creator contractors.

Unhinged, cursing avatars were going viral in upper-funnel content, so the team added them to the app. Users liked the look but did not understand why their avatar was cursing. Praktika removed them and decided against cursing avatars because it sells AI coaches, not AI companions.

After the cursing-avatar failure: product boundaries are agreed internally and written as a one-sheet description of what Praktika does and does not do. Growth content has only basic limits; creators can pursue pretty much any angle.

In onboarding, 'fear' was one of the most clicked reasons for wanting to practice, so Praktika built avatars to simulate talking to another person: they don't need to look fully human, just create the feeling that someone is watching. Pixar-style cartoons avoid the uncanny valley. Not built for marketing, though the avatars also show well in content. It was the most questioned part of the product during the first fundraise.

In 2022 web versions, pre-made Synthesia avatar videos were paired with hundreds of conversation trees hand-built by copywriters and content creators, so the experience felt like live AI conversation. Even then the avatar reacting to the user was central. Hundreds of customer interviews were recorded from that time.

There is no interface for setting a daily reminder; the AI tutor asks 'Hey, do you want to set a reminder? I can set this for you.' Part of making coaches more proactive.

Moved the app from Flutter to native iOS to stabilize it and give the best experience on the device; the team only hoped not to lose metrics. Development speed stayed about the same. Adam would go iOS native first and build Android after if starting again.