Our Process

At Vernacular, we use AI to understand people — not just what they believe, but why. Every interview is guided by empathy, depth, and respect for cultural nuance.

Representative Sampling from Trusted Data

We build samples from the highest-quality voter file and demographic data available, designed to capture variation across geography, race, age, education, language, and partisanship — not just nationally, but down to the communities that matter for your specific questions.

AI Voice Agents, Real Phone Calls

Our AI voice agents place live phone calls and conduct semi-structured interviews. They listen, follow up, probe deeper — adapting in real time to what the respondent says. These aren't robocalls or scripted surveys. They're genuine conversations, conducted in the respondent's preferred language, that surface the reasoning and emotion behind people's views.

From Transcripts to Estimates: MRP at Scale

Every interview produces a full transcript. We extract themes, sentiment, and reasoning patterns using large language models, then combine these with respondent demographics in multilevel regression and poststratification (MRP) models. The result: population-representative estimates of attitudes, beliefs, and issue priorities — broken out by geography, demography, and voter segments — from a method that captures why people think what they think, not just which box they'd check.