Simile Raises $100M to Build AI That Simulates Human Decision-Making

SAN FRANCISCO: Stanford spinout Simile emerges from stealth this week with $100 million in Series A funding to build what it calls the first AI foundation model for simulating human behavior. Index Ventures leads the round, with Bain Capital Ventures, A*, and Hanabi Capital participating. AI luminaries Fei-Fei Li and Andrej Karpathy also back the company.

Founded by Stanford computer scientists Joon Park, Michael Bernstein, Percy Liang, and Lainie Yallen, Simile trains its model on interviews with hundreds of real people, transaction records, and behavioral science literature.

The result is a platform that builds high-fidelity digital twins of individuals, AI agents that replicate how specific people think, prefer, and decide. The model took seven months to develop and, in independent testing, reproduces survey responses with 85% of the accuracy of the actual humans it models.

The commercial applications span industries. CVS Health uses Simile to model which products customers will buy before committing to shelf placements. Telstra, Australia’s largest mobile provider, is also among early customers.

On the corporate side, Simile tells Bloomberg its model correctly forecast eight out of ten analyst questions on a simulated earnings call, giving executives a rehearsal tool that doesn’t require actual analysts. The platform can also stress-test product launches, UI changes, and policy announcements against simulated audience reactions before anything ships.

The founding team’s academic roots trace directly to Smallville, a landmark 2023 research project that placed 25 AI agents in a simulated environment and demonstrated that generative agents could reproduce both individual and group behavior with measurable accuracy.

Simile treats simulation fidelity as a scientific discipline, a differentiator it argues separates it from competitors building synthetic research tools without the same empirical grounding.