The Global Journey of Speak: Revolutionizing Language Learning Through AI

In 2018, Connor Zwick found himself driving through Seoul, surrounded by skyscrapers that housed not offices, but English language academies. Billboards and taxi tops plastered with ads for English classes revealed a cultural obsession he hadn’t anticipated. This moment of realization changed everything for Zwick’s fledgling language learning app, Speak. “English language learning there was like an obsession,” he recalls. “There was such latent demand.” The traditional approach to language education in South Korea emphasized textbooks, grammar rules, and vocabulary lists over practical speaking skills. Students spent countless hours in classrooms led by instructors who often lacked fluency themselves, creating an environment where the fear of making mistakes prevented authentic learning. This ineffective yet expensive system represented not just a problem, but an opportunity for Zwick and his co-founder Andrew Hsu to create something revolutionary – an AI-powered language tutor that could provide judgment-free practice and personalized feedback to help learners become truly conversational.

What began as a vision has transformed into an impressive reality. Today, Speak offers an innovative voice-based AI coach that engages users in realistic role-playing scenarios – ordering at a restaurant, asking for directions, or making small talk – in six different languages including English, Korean, Spanish, Japanese, French, and Italian. Built primarily on OpenAI’s models, the app creates a safe space for learners to practice speaking aloud without fear of embarrassment. “No one’s going to know if you said something embarrassing,” Zwick points out, highlighting one of the key psychological advantages of learning with AI rather than human tutors. Users can follow structured lessons or create their own practice scenarios through prompts, with gamification features like leaderboards and streaks to encourage consistent practice. This approach has resonated with users worldwide, resulting in 15 million downloads and recently surpassing $100 million in annualized revenue – a significant milestone for the company now valued at $1 billion.

The path to this success wasn’t straightforward. When Speak was founded in 2016, raising capital proved challenging as investors showed little interest in either language learning or AI applications. Some were even suspicious, questioning whether the app was secretly designed to harvest user data. The founders’ backgrounds as Thiel fellows who had dropped out of prestigious universities (Harvard for Zwick and Stanford for Hsu) to pursue their vision speaks to their commitment to innovation outside traditional pathways. Their passion for AI led them to unconventional learning methods – sneaking into classes at UC Berkeley and Stanford, even attending lectures by OpenAI co-founder John Schulman without being detected. This dedication to understanding cutting-edge technology positioned them perfectly for the AI revolution that would eventually transform their industry. “They were AI-first before it was actually cool to be AI-first,” notes Khosla Venture Partner Sven Strohband, an early investor who remembers when Speak’s technology could understand foreign accents but wasn’t yet capable of full conversations.

The company’s decision to focus initially on South Korea proved strategically brilliant. Rather than competing immediately in crowded Western markets, Speak used South Korea as what Accel Partner Ben Quazzo calls a “sandbox” – a demanding testing ground in one of the world’s most competitive language learning environments. Success there validated their approach and provided a foundation for expansion to other markets including Japan and Taiwan. The enterprise segment has also become increasingly important, with 500 companies including KPMG and HD Hyundai now offering Speak subscriptions to their employees, primarily in South Korea. This dual approach – serving both individual consumers and corporate clients – has created multiple revenue streams as the company continues its global expansion, including a recent push into the U.S. market that began in June 2024.

As Speak enters the American market, it faces formidable competition, particularly from language learning giant Duolingo, which reported $724 million in revenue last year and projects growth to over $1 billion by the end of this year. Duolingo has also embraced AI, offering video calls with its AI character Lily as part of its suite of learning tools. Yet Zwick believes Speak offers something fundamentally different. While acknowledging Duolingo’s strength in teaching grammar and vocabulary, he positions Speak as focused on conversational fluency and pronunciation – helping learners become comfortable actually speaking the language. “People speak five to ten times more on Speak than on other language learning apps,” he claims, emphasizing that the difference is more than cosmetic. In a pointed comparison, Zwick describes competitor apps as “Mobile games that also teach you language… a healthier alternative to Candy Crush,” while positioning Speak as “something a little bit more serious” – a genuine educational tool rather than entertainment.

The future looks promising for Speak as AI technology continues to advance. Having raised approximately $160 million from investors including Khosla Ventures, Accel, and the OpenAI Startup Fund, the company is well-positioned to enhance its offerings as underlying AI models improve. Features that were impossible when the company was founded, like accent correction, are now part of the user experience, with the app automatically updating lesson plans based on individual performance. “We’re not selling like a dopamine hit or engagement bait app,” explains Quazzo. “We’re selling an education journey to get you to an end outcome.” This focus on substantive results rather than superficial engagement distinguishes Speak in an increasingly crowded edtech landscape. As the technology continues to evolve, Speak’s early bet on AI and its commitment to helping people become genuinely conversational in new languages may well position it as a leader in the next generation of language learning tools – transforming not just how we learn languages, but how effectively we can communicate across linguistic boundaries.