Simile raises $100M from Index to build society-scale AI simulations — predicts 8 of 10 earnings call questions

Summary

Simile has raised $100M from Index Ventures to build foundational models of human behavior that simulate society at individual and population scales. The company, led by co-founder and CEO Joon Sung Park, positions itself as a tool for Fortune 500 companies to interact with AI agents representing real people in order to predict outcomes and test decisions before execution.

The core product lets customers talk to millions of simulated agents built from real human data. Simile works with vendors to collect rich behavioral data on individuals, then builds high-fidelity simulations that capture people's values, preferences, and tastes rather than just rational decision-making. Park emphasizes that this focus on modeling human irrationality is fundamentally different from what frontier AI labs pursue.

Earnings call simulation

Simile helps executives prepare for earnings calls by simulating individual analysts and modeling how CEOs and CFOs might respond to their questions. The company predicts 8 out of 10 actual questions analysts will ask on average, then simulates how the conversation flows. This lets companies test different response strategies without public exposure.

Commercial customers

CVS uses Simile to simulate hundreds of thousands of its customers to test new products, store layouts, and concepts faster and more granularly than traditional human panels. Gallup uses Simile to create digital panels of its respondents when they are unavailable, extending the reach of its surveys without additional human recruitment.

Moat and differentiation

Generic frontier models fine-tuned on earnings call transcripts will only yield surface-level predictions. Simile's advantage comes from combining data science with product design. The founding team—Park, Michael Bernstein, and Preston Liang—brings both AI research credentials in generative agents and foundation models alongside a disciplined approach to data collection. Park frames the moat as tight alignment between modeling and product frontiers: better models of human behavior immediately yield better customer insights.

Scale trajectory

Simile currently models hundreds of thousands of people. Customers have appetite to simulate millions. Park's long-term vision is to model 8 billion people, though he acknowledges both data and computational efficiency as open questions.

Philosophical framing

When pressed on whether perfect prediction kills agency, Park pivots to interpretability. He cites the Oracle from The Matrix: you are not here to make a decision, you have already made it; you are here to understand why. Park argues Simile's value is not just prediction but traceability, allowing policymakers and product teams to understand how societal outcomes unfold so they can serve populations better.

Angel investors include Andrew McCarthy. The round reflects investor confidence that human behavior modeling grounded in real data and product intuition can move faster and more reliably than generic model fine-tuning for enterprise decision-making.