Customer ServiceReplaces Human LaborVerified

Duolingo uses an AI chatbot (powered by a company called Decagon) to handle the majority of customer support requests for the Duolingo English Test. The AI fully resolves about 80% of chat inquiries without any human involvement, answering questions about test registration, scores, and institutional acceptance.

Details

Before adopting Decagon AI, Duolingo's English Test support team used a different AI vendor that deflected only about 30% of email tickets and failed to launch chat automation after a year of development. After switching to Decagon — with a reported go-live time of under one month — the 80% chat deflection rate was achieved immediately. The system automatically syncs with the Duolingo English Test FAQ database every hour to stay current. In early 2025, the DET team planned to expand the Decagon-powered system to email support. Note: No evidence was found of an equivalent AI chatbot deployment for the main Duolingo language app's customer support.

Products affected

Duolingo English Test

Sources & Evidence

Company Disclosure

How Duolingo Partners with Decagon

Decagon AI

https://decagon.ai/case-studies/duolingo

Other practices by Duolingo

Data AnalysisDuolingo uses a machine learning model to decide, in real time, whether to show each individual user an advertisement, a subscription offer, or neither — predicting which choice is most likely to generate revenue from that specific person at that moment. The system was credited with generating tens of millions of dollars in additional annual revenue.Data AnalysisDuolingo uses a machine learning algorithm to decide which push notification to send each user, at what time, and how often — optimizing for the messages most likely to get a lapsed learner back into the app. The system was published as research in 2020 and was shown to increase the number of daily active users by 0.5% and improve retention of new users by 2%.Data AnalysisWhen someone takes the Duolingo English Test at home, an AI system monitors the session by analyzing video, audio, typing patterns, and eye movements — looking for signs of rule violations across more than 75 different behavioral and environmental signals. AI flags potential issues, and then human proctors make the final call on whether to certify the test.

Have evidence about Duolingo's AI practices? Submit a report.

Report a Sighting →