Lark AI EDU — training bots inside e-learning courses
Employees rehearse difficult conversations with a voice bot; managers get a scored report instead of a quiz
Each conversation scored against 5 rules on a 0–20 scale with written justification, computed once at the end
Voice conversation with speech recognition and synthesis, optionally with an avatar
Three applications — the in-course widget, an organisation manager panel and a platform admin panel
Challenge
Corporate training ends with a multiple-choice test that says nothing about how someone handles a real conversation. Rehearsing with a trainer works, but does not scale to hundreds of people.
Solution
A widget embedded in the course starts a voice conversation with a bot — speech-to-text, model response and speech synthesis, optionally with an avatar. Afterwards the system scores the conversation against five rules on a 0–20 scale with written justification; managers see team results, can override a score and track progress over time. Admins build bots from the company’s own knowledge base.
Technology & facts
- Model-usage accounting per organisation, with cost control on the backend side
- Gamification — badges and rankings computed on a schedule for participants