Human Tutors Team Up With AI: How Preply's Lesson Insights Personalizes Language Learning!

Hey there, it's Shii-chan! Today I've got a warm little story about AI and human teachers teaming up.

OpenAI News openai.com

What was announced?

Over on OpenAI's News, there's a new case study about how Preply, the world's largest marketplace for online language learning, is using OpenAI. Preply connects more than 100,000 expert tutors with learners across more than 180 countries, in more than 90 languages.

The star of the story is Lesson Insights, a feature built on the OpenAI API. After every 1:1 lesson, the AI analyzes what happened and automatically generates feedback on grammar, vocabulary, and pronunciation.

Why it matters

Learning a language is deeply human — it needs conversation, confidence, motivation, and cultural understanding. Preply's tutors bring all of that, but they were also spending time on repetitive work like writing lesson plans and notes. Learners, meanwhile, often want a clear sense that they're actually improving.

So instead of using AI to replace people, Preply used it to strengthen the tutor-learner relationship: let AI handle the admin, and let humans focus on what humans do best.

What changes

Within minutes of a lesson ending, the learner and tutor get a personalized report right in their chat thread. It includes:

• A summary of the key lesson topics
• Grammar corrections and explanations
• Vocabulary highlights and translations
• Pronunciation feedback
• Recommended next steps

And it doesn't stop there — those insights feed straight into Preply's self-learning exercise engine, which generates homework tailored to each learner. One human lesson turns into practice that keeps building over time.

One tutor, Michelle, said prep and homework used to take hours, and now much of it happens automatically:

"Before I started using Preply's AI feature, I would spend hours and hours prepping for classes and creating homework. But now that time has been cut by more than half."

Dive Deep

Here's a bit more on how it works. Lessons happen in the Preply Classroom, and with the learner's consent they're recorded and transcribed. Insight generation is scheduled a few minutes before the session ends, so the feedback is ready for the tutor and learner to review together the moment the lesson wraps up. I love that it's all consent-based.

Internally, Preply rolled out ChatGPT Enterprise across the company, running enablement sessions for more than 600 employees in New York, Kyiv, London, and Barcelona. Weekly active usage climbed from 60% to 95%. They've also embedded custom GPTs like a Brand Voice GPT into daily work to keep everything on-brand.

On the engineering side, Codex is doing a lot of the heavy lifting: around 94% of Preply's engineers use Codex and AI coding assistants for code generation, PR reviews, and debugging, which frees them up to build new learning experiences.

Wrap-up

Let me recap the key points:

• Preply built Lesson Insights on the OpenAI API to auto-generate grammar, vocabulary, and pronunciation feedback from 1:1 lessons
• Those reports flow into a self-learning engine that turns them into personalized homework
• Internally, ChatGPT Enterprise weekly usage went from 60% to 95%, and about 94% of engineers use Codex
• The numbers look strong: a 4.7/5 satisfaction rating from more than 300k ratings, and a 70% product market fit score

Emily wrapped it up nicely:

"We're only at the beginning."

It's not "humans or AI" — it's humans in the lead, with AI backing them up. If you're thinking about how to fold AI into a language or education product without losing the human touch, this one's for you.