Adam Turaev
Co-founder and CEO, Praktika
Where to hear and read about Adam Turaev1 episode or test
Adam Turaev is co-founder and CEO of Praktika, an AI language tutor app that reached $50M ARR with a team of just over 40 people. The most useful thing to learn from him is his split between near-total creative freedom in growth and tight written guardrails in the product.
Do not tell influencers how to do their job.
How Praktika built a profitable $50M ARR AI business — Adam Turaev, Praktika
Adam Turaev: the story and the lessons
1How Adam Turaev got from enterprise chatbots to Praktika
Turaev started with chatbots, years before large language models. Three school friends built Cleverbots in 2017, automating customer and HR support for large international companies, sometimes up to 60% of tickets. Telegram's chatbot API inspired the idea.
They believed chatbots would become how people interact with machines. Building hundreds of chatbots taught the team how people expect a conversation to go and where they quit, which made conversation design a specialty.
He picked language learning because he felt the barrier himself. Turaev is not a native English speaker and calls English an income and opportunity multiplier. The three co-founders lived the language barrier and broke through it themselves.
The pivot followed a pain point, not the biggest market. He saw dissatisfaction with existing solutions and no one fixing it with conversational AI, and the team did not expect the market to be this big.
Praktika shipped rough early versions and improved as models caught up. Prototyping began in 2021. A 2022 web version used a Synthesia avatar and hundreds of conversation trees written by copywriters, the mid-2023 app launch was a rough one, and public launch came in 2024.
Praktika still has hundreds of 2022 customer interviews in which users say what they say now. Turaev's advice: "Build on the edge of technology."
Turaev aims at career-minded adults rather than casual learners. Praktika targets serious career English for people like its founders, who had moved countries. It teaches English only.
He praises Duolingo's product and says it is still in a league of its own, so Praktika avoids head-on competition with it.
2Why Turaev chose Brazil and sold a better job
Turaev picked Brazil from an English proficiency report, not intuition. The English First report showed Brazil as one of the least English-penetrated markets, with a large income gain from English. Praktika launched worldwide but marketed only to Brazil.
The team worried about purchasing power but judged the fundamentals strong enough.
Praktika sold a better job, not English lessons. In Brazil, where people rarely pay for consumer subscriptions, people paid about $100 a year. All claims were framed around career outcomes and compared with the cost of a tutor.
The "50X cheaper than a tutor" claim came from this go-to-market. The product itself teaches language, not job search, so the job promise lives in the marketing. In Turaev's words: "We didn't sell English to them. We sold them an opportunity to get a better job."
3How Turaev finds creators and lets them run growth
He searched for high views and low followers to find rising creators. With no budget for big names, Turaev used Modash to cap followers below 5,000 and filtered for accounts the algorithm was already pushing. He spent about a month full time messaging a long list, and a few worked out.
Gustavo listed no email, so Turaev sent a direct message on TikTok. He says the filter works beyond micro-influencers because the point is reaching someone before they peak.
A video Turaev disliked took ARR from $1M to about $12M. Gustavo, a Brazilian creator, made a video of a coach correcting a learner's mistakes. It got 10M views each on Instagram, TikTok and YouTube, and ARR rose from $1M to about $12M in a month.
Turaev wanted the app shown, but Gustavo said "Believe me, trust me." The source gives no breakdown of paid spend, other creators' results or retention, so the cause is not proven.
Turaev copied the viral format across countries before understanding it. Praktika reused the format with other creators in other countries and varied the angles. Early on, CPMs, CPIs and cost per trial and purchase were close to zero.
He says the team did not realize what was happening that month. The format has since become standard for language-app videos, and competitors use it too.
He hired Gustavo at once and drops his own approval filter. Turaev's rule since the beginning is not to tell influencers how to do their job. A founder's approval filter hides his own blind spots and can kill the winning video.
Failed experiments can be dropped quietly, since trends move fast.
Former creators run the program of more than 100 creators. The influencer lead began as a creator in Poland and now runs the program worldwide. Country coaches are ex-creators who each rotate 20 or 30 creators; leads and coaches are employees and creators are contractors.
Turaev says management is easy to sort out, but knowing good content is the key skill. The team uses SideShift with internal add-ons for analytics and reporting.
Pay is a fixed fee plus a view bonus, bent for top talent. Coaches can reshape a deal for a strong creator because they optimize for talent. Turaev's rule: "If there is a good talent, just do whatever to hire this person."
He says nothing about the structure is unusual.
He keeps organic creators and paid acquisition as separate streams. When a video goes viral, the paid team reproduces it with its own internal influencers or contractors. Creators are never briefed by paid acquisition.
The in-house team runs out of angles, and each new creator brings a new way of thinking. Turaev says: "Those two are separate streams. We don't mix them up."
4Turaev's product guardrails and the cursing avatar failure
Cursing avatars went viral in ads but failed inside the app. Users liked the avatar's look but asked why their coach was cursing. Turaev calls it the biggest failed experiment of the year.
He thinks the brand's "good guys" image made them a poor fit. Praktika is an AI coach, not an AI companion.
Product limits fit on one page while growth limits are almost absent. Product boundaries are agreed internally as a one-sheet description of what is and isn't done. Creators can pursue pretty much any angle, with only basic limitations.
Turaev calls the difference between the two pretty dramatic.
Turaev built avatars to ease speaking anxiety, not for marketing. Fear of speaking to another person was one of the most clicked reasons in onboarding. The avatar gives learners a feeling of being watched so they can rehearse.
It does not need to be perfectly human, and the Pixar style avoids the uncanny valley. Investors questioned the avatars most in the first fundraising round. He says: "It was never with marketing or distribution in mind."
Turaev lets the job to be done drive product choices. Praktika scores learners on the CEFR scale at onboarding and offers a judgment-free place to break the language barrier. Turaev says this makes it a coach with a method, unlike a general voice model that is a sparring partner.
He argues that focusing on the outcome puts context, harness and opinionated user experience first, which is what separates products in the AI era.
5How Turaev runs Praktika lean
Praktika reached $50M ARR with a little over 40 people. The company has almost half a million paying subscribers, so revenue per employee tops $1M. Turaev says the company is profitable.
The jump from $1M to $12M ARR came before this stage.
He prioritizes quality over inference cost because model prices keep falling. A model that was 50% cheaper improved unit economics, and a successor two or three weeks later was 50% cheaper again. Turaev expects models to eventually run on the device.
His view: "You can focus on quality and be efficient." A tutor only needs to be smart enough to be useful.
Moving from Flutter to native iOS was his biggest win of the year. Turaev only hoped not to lose metrics, but conversion and engagement improved across the board, and development speed stayed the same.
Starting again, he would go iOS native first and build Android after.
He has the tutor handle settings through conversation. There is no reminder screen. The tutor asks whether to set a daily reminder and sets it.
Turaev sees this as how the product will work in future, with Praktika's coaches becoming more proactive and moving beyond language to other subjects.
How Praktika built a profitable $50M ARR AI business — Adam Turaev, Praktika
Praktika's CEO explains how an AI language tutor app reached $50M ARR with a team of just over 40 by selling career outcomes in Brazil and letting micro-influencers run creative. A single creator video he did not like drove roughly 30M views and took ARR from $1M to about $12M in one month.
Turaev started with chatbots, years before large language models. Three school friends built Cleverbots in 2017, automating customer and HR support for large international companies, sometimes up to 60% of tickets. Telegram's chatbot API inspired the idea.
They believed chatbots would become how people interact with machines. Building hundreds of chatbots taught the team how people expect a conversation to go and where they quit, which made conversation design a specialty.
He picked language learning because he felt the barrier himself. Turaev is not a native English speaker and calls English an income and opportunity multiplier. The three co-founders lived the language barrier and broke through it themselves.
The pivot followed a pain point, not the biggest market. He saw dissatisfaction with existing solutions and no one fixing it with conversational AI, and the team did not expect the market to be this big.
Praktika shipped rough early versions and improved as models caught up. Prototyping began in 2021. A 2022 web version used a Synthesia avatar and hundreds of conversation trees written by copywriters, the mid-2023 app launch was a rough one, and public launch came in 2024.
Praktika still has hundreds of 2022 customer interviews in which users say what they say now. Turaev's advice: "Build on the edge of technology."
Turaev aims at career-minded adults rather than casual learners. Praktika targets serious career English for people like its founders, who had moved countries. It teaches English only.
He praises Duolingo's product and says it is still in a league of its own, so Praktika avoids head-on competition with it.
Turaev picked Brazil from an English proficiency report, not intuition. The English First report showed Brazil as one of the least English-penetrated markets, with a large income gain from English. Praktika launched worldwide but marketed only to Brazil.
The team worried about purchasing power but judged the fundamentals strong enough.
Praktika sold a better job, not English lessons. In Brazil, where people rarely pay for consumer subscriptions, people paid about $100 a year. All claims were framed around career outcomes and compared with the cost of a tutor.
The "50X cheaper than a tutor" claim came from this go-to-market. The product itself teaches language, not job search, so the job promise lives in the marketing. In Turaev's words: "We didn't sell English to them. We sold them an opportunity to get a better job."
He searched for high views and low followers to find rising creators. With no budget for big names, Turaev used Modash to cap followers below 5,000 and filtered for accounts the algorithm was already pushing. He spent about a month full time messaging a long list, and a few worked out.
Gustavo listed no email, so Turaev sent a direct message on TikTok. He says the filter works beyond micro-influencers because the point is reaching someone before they peak.
A video Turaev disliked took ARR from $1M to about $12M. Gustavo, a Brazilian creator, made a video of a coach correcting a learner's mistakes. It got 10M views each on Instagram, TikTok and YouTube, and ARR rose from $1M to about $12M in a month.
Turaev wanted the app shown, but Gustavo said "Believe me, trust me." The source gives no breakdown of paid spend, other creators' results or retention, so the cause is not proven.
Turaev copied the viral format across countries before understanding it. Praktika reused the format with other creators in other countries and varied the angles. Early on, CPMs, CPIs and cost per trial and purchase were close to zero.
He says the team did not realize what was happening that month. The format has since become standard for language-app videos, and competitors use it too.
He hired Gustavo at once and drops his own approval filter. Turaev's rule since the beginning is not to tell influencers how to do their job. A founder's approval filter hides his own blind spots and can kill the winning video.
Failed experiments can be dropped quietly, since trends move fast.
Former creators run the program of more than 100 creators. The influencer lead began as a creator in Poland and now runs the program worldwide. Country coaches are ex-creators who each rotate 20 or 30 creators; leads and coaches are employees and creators are contractors.
Turaev says management is easy to sort out, but knowing good content is the key skill. The team uses SideShift with internal add-ons for analytics and reporting.
Pay is a fixed fee plus a view bonus, bent for top talent. Coaches can reshape a deal for a strong creator because they optimize for talent. Turaev's rule: "If there is a good talent, just do whatever to hire this person."
He says nothing about the structure is unusual.
He keeps organic creators and paid acquisition as separate streams. When a video goes viral, the paid team reproduces it with its own internal influencers or contractors. Creators are never briefed by paid acquisition.
The in-house team runs out of angles, and each new creator brings a new way of thinking. Turaev says: "Those two are separate streams. We don't mix them up."
Cursing avatars went viral in ads but failed inside the app. Users liked the avatar's look but asked why their coach was cursing. Turaev calls it the biggest failed experiment of the year.
He thinks the brand's "good guys" image made them a poor fit. Praktika is an AI coach, not an AI companion.
Product limits fit on one page while growth limits are almost absent. Product boundaries are agreed internally as a one-sheet description of what is and isn't done. Creators can pursue pretty much any angle, with only basic limitations.
Turaev calls the difference between the two pretty dramatic.
Turaev built avatars to ease speaking anxiety, not for marketing. Fear of speaking to another person was one of the most clicked reasons in onboarding. The avatar gives learners a feeling of being watched so they can rehearse.
It does not need to be perfectly human, and the Pixar style avoids the uncanny valley. Investors questioned the avatars most in the first fundraising round. He says: "It was never with marketing or distribution in mind."
Turaev lets the job to be done drive product choices. Praktika scores learners on the CEFR scale at onboarding and offers a judgment-free place to break the language barrier. Turaev says this makes it a coach with a method, unlike a general voice model that is a sparring partner.
He argues that focusing on the outcome puts context, harness and opinionated user experience first, which is what separates products in the AI era.
Praktika reached $50M ARR with a little over 40 people. The company has almost half a million paying subscribers, so revenue per employee tops $1M. Turaev says the company is profitable.
The jump from $1M to $12M ARR came before this stage.
He prioritizes quality over inference cost because model prices keep falling. A model that was 50% cheaper improved unit economics, and a successor two or three weeks later was 50% cheaper again. Turaev expects models to eventually run on the device.
His view: "You can focus on quality and be efficient." A tutor only needs to be smart enough to be useful.
Moving from Flutter to native iOS was his biggest win of the year. Turaev only hoped not to lose metrics, but conversion and engagement improved across the board, and development speed stayed the same.
Starting again, he would go iOS native first and build Android after.
He has the tutor handle settings through conversation. There is no reminder screen. The tutor asks whether to set a daily reminder and sets it.
Turaev sees this as how the product will work in future, with Praktika's coaches becoming more proactive and moving beyond language to other subjects.