Outset raises $17M Series A from 8VC for AI-moderated research — Microsoft and Nestlé are customers

Summary

Outset, a YC Summer 2023 alumnus, closed a $17M Series A led by 8VC partner Jack Moshkovitz. The company sells AI-moderated research to enterprises including Microsoft, Nestlé, and Weight Watchers, targeting a $140 billion market currently dominated by traditional research firms like Ipsos.

Outset replaces two inadequate research methods. Mass surveys produce shallow data. One-on-one human interviews don't scale. Outset's AI conducts video interviews with voice, screen share, and multimodal input across hundreds of participants simultaneously, then synthesizes results into segmented, quantifiable reports. A client can interview 500 people in a day or two, something physically impossible with human researchers.

Voice over text and avatars

The product launched with text-only interviews in 2023 via an early Weight Watchers pilot. Voice and video have since proven meaningfully richer. A full voice-to-voice mode exists but carries just enough latency to remind participants they're talking to a machine. The company experimented with AI avatars and found participants rejected them due to uncanny valley discomfort. The current approach uses AI voice without a face, and participants accept it. They often share more candidly with an AI than with a human researcher because social desirability bias disappears. The Weight Watchers use case, which centered on personal weight loss habits, illustrated this directly.

Real-time infrastructure

Outset runs primarily on OpenAI models via Azure, chosen for reliability and enterprise compliance. Nestlé-scale customers demand both. Interviews must process in real time rather than batch overnight, so the system handles multiple clients running 500 simultaneous interviews during peak hours. Gemini serves as a fallback alongside direct OpenAI access. Azure's model router, announced at Microsoft Build, has not yet rolled out.

Synthesis as product

Outset frames its output layer as deep research for primary research, an end-to-end agentic workflow that converts hours of conversational video into quantified breakdowns, highlight reels, and slice-and-dice analytics. The pitch is that qualitative data at this scale has never had a real analytics layer on top of it.

Customer segmentation

Outset deliberately targets enterprises and turns away early-stage founders who want to use the product to avoid talking to customers directly. The guidance is to return once the company has a few hundred people and is scaling research operations. Consumer businesses trying to learn from millions of users are a better fit at smaller scale.