Pulse surveys use Jev to bucket and classify open-ended answers
Instead of dumping all open-ended responses into an LLM for a summary, Pulse's Text Insights uses Jev to bucket and classify each answer, enabling real-time filtering, grouping and drill-down.
pretty proud of how we built "Text Insights" on pulse surveys. analyzing open ended text response questions is tricky. most platforms shove all the responses into an LLM and get a summary. in practice this isn't that useful and you can't slice and dice data easily. instead we built a custom pipeline using jev (@typesafeai) that buckets and classifies every response. this allows you to filter and group responses in real time and you can easily go deeper with Windy whenever you want.