Preply deploys OpenAI API to automate lesson feedback while preserving human tutoring
Preply's Lesson Insights uses GPT to generate personalized language-learning feedback post-lesson, freeing tutors from administrative work.
Last verified:
Automating the administrative layer of language tutoring
According to the OpenAI Blog, Preply, the world’s largest online language-tutoring marketplace, has deployed an OpenAI API-powered feature called Lesson Insights to generate personalized feedback after each 1:1 lesson. The marketplace connects more than 100,000 tutors with learners in 180+ countries, offering instruction across 90+ languages. Lesson Insights analyzes lesson transcripts to produce structured reports covering grammar corrections, vocabulary highlights with translations, and pronunciation guidance—work that previously consumed tutor time on administrative documentation. Learners and tutors receive the feedback within minutes, directly in their chat thread.
Why Preply chose OpenAI’s infrastructure
Preply evaluated multiple AI models before selecting OpenAI as its technology partner. According to Voloshyn, Preply’s Co-founder and CTO, OpenAI’s models delivered superior performance on language-learning tasks while meeting the company’s requirements for speed, reliability, and production-scale deployment. The partnership extended beyond the customer-facing Lesson Insights product: Preply introduced ChatGPT Enterprise across its entire 600-person organization spanning offices in New York, Kyiv, London, and Barcelona. Weekly active usage of ChatGPT Enterprise grew from 60% to 95% following company-wide enablement sessions, demonstrating rapid internal adoption.
Preserving human connection while reducing tutor workload
Preply’s framing of Lesson Insights reflects a deliberate philosophy: AI augments rather than supplants human teaching. Voloshyn stated that personalization represents the core opportunity in language education—identifying patterns and nuances in individual learner progress that are difficult for humans to track manually. By automating transcript analysis and feedback generation, tutors reclaim time previously spent on repetitive documentation, directing that capacity instead toward real-time interaction, motivation, and cultural scaffolding that define effective language instruction.
The technical execution prioritizes learner agency: lessons are recorded and transcribed only with explicit learner consent, and feedback generation is scheduled a few minutes before the lesson ends, allowing tutor and student to review insights together while the session context remains fresh.
Why This Matters
Preply’s implementation of Lesson Insights illustrates a pragmatic middle ground in AI-education adoption: using LLM inference to compress routine post-session documentation rather than to replace classroom instruction. For tutoring marketplaces and language-learning platforms, the win is twofold—learner engagement improves when progress is immediately visible and actionable, and tutor retention improves when administrative overhead diminishes. If Preply’s adoption metrics hold (600-person org reaching 95% ChatGPT Enterprise usage), the model demonstrates that knowledge workers in education sectors view AI-assisted workflow tools as productivity enablers, not threats. Competitors in the edtech space will likely benchmark Preply’s approach as they weigh similar investments in transcript-to-feedback pipelines.
Frequently Asked Questions
How does Preply's Lesson Insights feature work?
After each 1:1 lesson, OpenAI's API analyzes the recorded and transcribed session to generate structured feedback on grammar, vocabulary, and pronunciation. Both tutor and learner receive the report within minutes in their chat thread.
Does this replace human tutors?
No. According to Preply Co-founder Dmytro Voloshyn, the feature is designed to strengthen the tutor-learner relationship by automating repetitive administrative tasks, not to replace human instruction.
How has Preply adopted AI internally?
Preply rolled out ChatGPT Enterprise across its 600-person workforce (offices in New York, Kyiv, London, Barcelona), growing weekly active usage from 60% to 95% after internal enablement sessions.