Workflow · workflow
Duolingo designs LLM-powered AI video calls with Lily for language speaking practice
Using LLMs for language learning is not straightforward — simply instructing the model to speak a language with a learner is insufficient, and the LLM needs structured prompts with clear goals and predictable sentence structure.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Pre-call first question generation
While the call bell rings, the system uses a separate LLM prompt to generate the first question for that call.
Tools used
LLM
Outcome
Duolingo achieved more natural and appropriately leveled AI video calls by separating first-question generation from main call instructions and adding a mid-call evaluation step to detect when learners want to lead the conversation.
What failed first
When first-question generation instructions were combined with main conversation instructions in a single prompt, the LLM became overloaded and produced undesired results — either overly complex sentences or failure to include required vocabulary.
Grounding & classification
Source type: technical build writeup
11 fields verified against source quotes, 1 dropped as unverifiable.
ai agentconversational aipersonalizationchat transcriptknowledge basenamed customerproduction runtime claimedtools describedvendor confirmedworkflow describededucationtechnical build writeupagentic task execution