How Vietnamese entrepreneur utilizes AI to teach English to 50 million learners worldwide

In January she was recognized as a Technology Pioneer at the World Economic Forum in Davos, Switzerland, among 200 startup founders leading innovations in their fields. Over nearly a decade Elsa (English Language Speech Assistant) has become one of the most widely used English-learning tools with over 50 million users globally.

Apple recently featured Elsa in its New Year 2025 highlights on Vietnam’s App Store, describing it as a prime example of AI-driven education.

"Elsa Speak employs a custom-built AI model to analyze users’ pronunciation at the syllable level, providing detailed feedback and improvement suggestions."

Elsa's CEO Van Dinh Hong Vu at the World Economic Forum 2025. Photo courtesy of Elsa

Despite excelling in English during her school years, she struggled with communication when she first studied abroad. She found that a lack of access to native speakers and real-time feedback made pronunciation difficult to refine. "For this challenge, technology is the only solution," she says.

When Elsa was first developed in 2015, AI was not yet widely applied in language learning, and was primarily used in virtual assistants. Initially the team considered Google’s speech recognition API, but it was designed to interpret commands despite errors, making it unsuitable for pronunciation training.Vu then partnered with Dr. Xavier Anguera, an expert in speech recognition and AI, to create a proprietary system.

Using deep learning, it could detect even subtle pronunciation mistakes with over 95% accuracy. The team later introduced Speech Analyzer, a tool offering real-time feedback and detailed corrections. As generative AI gained traction in late 2022, Elsa launched Elsa AI, a feature allowing users to engage in personalized simulated conversations.

"Learners can practice real-life dialogues tailored to their specific needs, whether for work or daily life," Vu says.

User interface of the Elsa app on iPhone. Photo by VnExpress/Luu Quy

The AI-powered tutor enables users to speak naturally without scripted responses, suggesting conversation ideas in both English and their native language when needed, she says. It also provides instant feedback on grammar, vocabulary, pronunciation, and fluency, helping users gradually refine their skills, she says.

"With just a smartphone, learners can receive one-on-one AI tutoring, bridging the shortage of English teachers and offering a personalized learning experience anytime, anywhere."

Although Elsa was launched in the U.S., Vietnamese users were among the first to access its latest features. In its early days 80-90% of users came from Vietnam, but as the platform expanded, this share has declined to around 20% as it gained traction in global markets. By 2024 Elsa had been downloaded in 195 countries and territories and had more than 50 million users.

It gained global recognition in 2017, when Forbes named it one of four AI-powered companies transforming the world.Vu believes perceptions of AI have shifted, with users increasingly viewing it as a time-saving tool that enhances efficiency in education and at the workplace.

"This aligns with Elsa’s vision: leveraging AI to make learning more efficient, reducing costs and time, and allowing users to focus on what truly matters."

As generative AI continues to evolve, she envisions English-learning apps becoming even more adaptive, catering to individual learners’ needs in real time. The company aims to develop an AI-powered communication tool that helps users speak confidently in any situation, enhancing both language proficiency and critical thinking skills.

"Each learner will have a customized journey that integrates English learning with other essential skills."