Duolingo + Dot

Challenge

Duolingo wanted every employee (“Duo”) to make decisions with data—without forcing them to learn SQL, know which of 5,000+ tables to trust, or depend on busy Data Science and Data Platform teams. Non-SQL users routinely queued requests for complex analyses, creating bottlenecks and elongating the path from product question to answer. The risk wasn't just speed; it was quality. With many similarly named tables and evolving business definitions, well-intentioned self-serve could produce inconsistent results. Duolingo needed a front door to its warehouse that was both radically accessible and uncompromising on governance.

“Before Dot… anyone who wasn’t a regular SQL user would have to consult our Data Scientists and Data Platform teams to construct more complicated analyses.”

Lavanya Aprameya, Senior Software Engineer at Duolingo

Approach

Dot provides a text-based entry point to data directly in Slack, where Duos already collaborate. Users ask questions in natural language; Dot maps business terms to canonical tables and metrics, generates reviewable SQL, and—crucially—asks clarifying questions rather than guessing when a prompt is ambiguous. This reduces schema hunting and prevents misuse of look-alike tables.

Under the hood, Dot encodes Duolingo's data knowledge: approved sources, metric definitions, and join logic. It enforces permissioning and routes queries to the right warehouse assets, so non-experts don't have to memorize the data catalog. Results arrive in-thread, making it easy to validate queries, share outputs, and iterate. By shifting routine data pulls and first-pass investigations to Dot, analysts reclaim time for deeper modeling and experimentation while maintaining a single source of truth.

“Dot makes it so you don’t have to write raw SQL or comb through an entire data catalog… and it asks for clarifications when prompts are ambiguous.”

Lavanya Aprameya, Senior Software Engineer at Duolingo

Impact

Dot has unlocked self-serve analysis at scale—saving 12,000+ hours, delivering an estimated 18× ROI, and cutting time-to-insight by 99%. With Slack-native, governed access to 5,000+ tables, Duos move from questions to defensible answers in minutes, while expert teams focus on higher-leverage work.