CVS is already asking your digital twin what you'd buy before you walk in the door.

The Summary

  • Simile raised $300 million at a $2 billion valuation to build AI models that simulate individual consumer behavior — not aggregates, but digital clones of specific audience segments.
  • CEO Joon Sung Park calls it "a foundation model of human behavior" that lets companies define an audience, then ask those simulated consumers any question about products, messaging, or experiences.
  • CVS, financial institutions, and entertainment companies are already using Simile to test everything from store layouts to marketing campaigns on digital twins before real humans ever see them.
  • This isn't predictive analytics. It's synthetic focus groups that never get tired, never lie about their income, and cost a fraction of traditional market research.

The Signal

Simile is building what amounts to a behavioral API for humanity. You feed it a description of a person or demographic, it spins up a simulation, and you interrogate that simulation about purchasing decisions, media preferences, or reactions to creative work. The company claims these aren't statistical models extrapolating from past behavior. They're generative simulations trained on enough human data to predict what a 34-year-old mother of two in suburban Atlanta would say about a new CVS loyalty program.

The valuation suggests investors believe this crosses a threshold. Market research is a $75 billion global industry built on surveys, panels, and A/B tests that take weeks and cost tens of thousands of dollars per campaign. Simile's pitch is simple: skip the humans, ask their twins, get answers in minutes. CVS is already using this to test store layouts and messaging, which means they're making real capital allocation decisions based on what simulated shoppers prefer.

"They can say who they want to talk to and what questions they want to ask. And really they can ask anything."

This is where it gets interesting for the agent economy. If Simile's models are good enough to replace focus groups, they're good enough to replace entire categories of human decision-making in product development, marketing, and retail operations. The immediate use case is research, but the obvious next step is agent-to-agent commerce. Why have a human sales team pitch a human procurement officer when both sides could deploy simulations to negotiate, test objections, and close deals while the principals sleep?

The entertainment and finance angles matter too. Hollywood could test rough cuts on simulated audiences segmented by age, politics, and viewing history. Banks could stress-test new financial products against simulated customers before risking real capital or regulatory blowback. Every industry that depends on understanding customer preference is a potential customer.

The Implication

If Simile works as advertised, the market research industry doesn't gradually digitize. It collapses. Companies that can simulate consumer panels for pennies on the dollar will stop paying for real ones. The humans who moderate focus groups, conduct surveys, and synthesize consumer insights will need new jobs.

The deeper question is what happens when your digital twin makes better predictions about your behavior than you do. Simile's models don't just reflect your preferences. They predict your responses to stimuli you haven't seen yet. That's not market research. That's behavioral forecasting at scale. Watch how fast this moves from testing store layouts to dynamic pricing, personalized manipulation, and agent-driven commerce where your AI negotiates with their AI and you only find out what you bought when it ships.

Sources

Business Insider Tech