The Behavioral Health Crisis Is a Data Problem

The average patient gets seen and disappears. No signal, no follow-up, no data trail. Just a receipt and a co-pay. Lauren Larson, CEO of Videra Health, knows exactly what lives in that gap, and he has spent six years building AI to close it.Lauren came up through HireVue, where video-based AI interviewed 100 million job candidates and surfaced the best ones through behavioral signal, not resume gatekeeping. When the company sold in 2019, he took everything he learned about reading humans through a screen and pointed it directly at behavioral healthcare.The Problem Videra Is SolvingThe system rewards patients who are good at making appointments. The people who are actually in crisis, the ones who missed their last visit, the ones who stopped their medication because of nausea, the ones who are not sleeping, those people spiral quietly. Videra uses AI-powered check-ins via audio and video to reach those patients between appointments, collect behavioral data, and surface the ones who need intervention back to their clinical teams.The platform is not trying to replace providers. It is trying to make sure providers only get interrupted when it actually matters.Core Frameworks DiscussedPassive vs. Structured Assessment: Lauren emphasizes the difference between conversational AI that just listens and structured clinical AI that knows which questions to ask first. The opening prompt is everything. Random check-ins produce noisy data. Calibrated sequences produce signal.Observational Biomarkers at Scale: Rather than guessing which features predict a condition, Videra trains on as many features as possible and lets the model surface what matters. The goal is 30 to 40 observational biomarkers detected in a single two-minute session, tracking movement, voice, language, and facial affect over time.The ROI Problem in Healthcare Innovation: Cool technology does not get deployed unless someone can pay for it. Lauren learned this lesson early. Videra had to expand beyond assessment into clinical documentation, patient intake, and provider coaching before the sales motion started working.Bias Testing Through Model Cards: For every predictive model, Videra builds model cards that track false positive and false negative rates across demographic and intersectional groups. Not just men vs. women. Not just race. But black women vs. black men vs. white women, and so on. Then they monitor for drift over time.The Elevate Product: AI that listens to provider-patient conversations and gives clinicians direct, specific feedback on where their empathy broke down and what they could have done differently. The goal is not to replace human care. It is to make every clinician perform closer to their best.Founder Experiment: Build a Behavioral Signal Intake BotUsing a voice or text-based AI agent (Claude, GPT-4o, or a similar LLM with tool access), build a simple structured intake flow for your product that collects behavioral signal, not just preference data.Start with three seed questions designed to elicit emotional state rather than factual answers. Log the responses. After 10 interactions, review the transcripts and flag any response patterns that correlate with disengagement, churn risk, or user distress. Run that as a lightweight customer health model before you ever touch a clinical dataset.If your product drives human decision-making in any way, behavior is your biggest data layer. This experiment will show you how much you are currently leaving on the table.⁠https://viderahealth.com/⁠⁠https://www.linkedin.com/in/estesryan/⁠⁠⁠⁠⁠⁠https://aiforfounders.co⁠⁠⁠⁠⁠⁠⁠https://kitcaster.com/application⁠ ⁠⁠⁠⁠⁠⁠⁠https://ryanestes.info⁠⁠⁠⁠