Dan Zabrotski Is Betting That AI Will Change What Human Teaching Is Worth

The software engineer behind a zero-commission language marketplace is confronting an uncomfortable possibility: AI may take some lessons from human tutors while increasing the value of the work that still requires a person.

Dan Zabrotski built a language-learning platform around a promise to human tutors: Speekl would not take a percentage of what they earned. The software engineer and technical entrepreneur personally built the platform, including its payment infrastructure, security, compliance systems, and core product. Tutors set their own prices and keep what they earn, while students book lessons through the platform. Now Zabrotski is developing another option inside the same business, an AI language tutor that he believes some learners will choose instead of a person.

“I would be doing tutors a disservice if I pretended AI was only going to sit quietly beside human teaching,” Zabrotski says. “For some students, the technology can already handle a meaningful part of what they came to accomplish.”

That view challenges one of the more reassuring ways artificial intelligence has been discussed in education. AI is frequently presented as a practice tool that leaves the traditional role of the teacher largely untouched. Zabrotski believes the division may become less tidy, particularly for learners who want conversational repetition and frequent practice. An AI tutor can be available in the middle of the night, requires no scheduling, and lets someone struggle through an unfamiliar language without another person listening.

“The embarrassment factor is easy to underestimate,” Zabrotski says. “People understand intellectually that mistakes are part of learning. That does not mean they are comfortable making the same mistake five times in front of somebody else. With software, that social pressure disappears.”

The immediate question in many professions has been whether machines will replace people. Zabrotski sees another possibility developing alongside that debate. Technology may absorb the most repeatable portions of a job while forcing professionals to become clearer about what a person contributes that inexpensive software cannot readily reproduce.

Speekl places Zabrotski unusually close to that question. Its human marketplace has grown to 250 tutors and 20,000 students across 12 countries. At the same time, he is building the company’s conversational AI tutor so learners can move between software and human instruction according to what they need. Zabrotski is also the sole engineer behind Speekl, which means the same person making decisions about the company’s approach to teaching is responsible for building the systems that put those decisions into practice.

Routine conversation practice is one example of where he expects the balance to shift. A learner can repeat phrases, work through vocabulary, and continue practicing without purchasing another hour of a tutor’s time. Zabrotski argues that human instructors become more useful when the work moves beyond repetition into areas where judgment and personal understanding shape the result. He points to cultural nuance and exam preparation, as well as the ability to recognize when a student is losing confidence or preparing to give up.

“Human tutors should be doing the work AI is bad at,” he says. “That is where paying for a person makes sense. Asking tutors to compete with software on the cost of basic repetition puts them in the wrong competition.”

Zabrotski’s willingness to say that carries particular weight because Speekl was originally built in response to frustrations he encountered while learning English online himself. He had studied with tutors through other platforms and objected to how much of their lesson income could go to the marketplace connecting them with students. He designed Speekl so the human marketplace would operate without taking a commission from tutors.

That history makes his current position less comfortable. A founder who spent years building a platform around tutor economics could easily insist that AI will remain secondary to human teaching. Zabrotski takes a different approach because he believes ignoring changing learner preferences would leave tutors less prepared for the shift.

“I have every reason to want human tutoring to remain strong,” he says. “But reassuring people does not change what students decide to buy. If learners are beginning to use AI for work they previously paid a tutor to do, the useful response is to understand why and decide where human teaching becomes more valuable.”

His perspective is also shaped by the way he works as an engineer. Zabrotski began teaching himself computer science and programming at fourteen, deliberately choosing C++ as his first language because he wanted to understand how software worked at a deeper level. He was working professionally by sixteen and has since built software across fintech, big data, transportation, food and beverage, fitness, and geospatial technology. That experience allows him to approach the current shift as someone who understands both the business decisions and the technical work required to turn them into a functioning product.

Affordability remains one of the strongest arguments he makes for the technology. Daily practice with a human tutor is beyond the budget of many language learners, while AI can make far more frequent conversation possible without multiplying the cost of instruction. For Zabrotski, greater accessibility changes where human instruction is most useful. The tutor’s role becomes more concentrated around situations where judgment, cultural knowledge, accountability, or personal encouragement materially affect the learner.

“Daily language practice should be available to people who could never afford daily lessons with a tutor,” he says. “There are also moments when having another person involved changes what the student does, how long they persist, and how well they understand what they are trying to learn.”

Learners will ultimately decide where that boundary falls. Speekl is being built to give them access to both forms of instruction, allowing their choices to reveal when software is sufficient and when another person is worth paying for.

For Zabrotski, that makes the debate less theoretical. He is building the AI product beside the human marketplace he created for tutors, giving learners the opportunity to decide which parts of language education still justify paying for another person’s time.