Jun Park, co-founder and CEO of Simile, is building an AI that feels like a cheat code for market research. His company’s “foundation model of human behavior” doesn't just crunch numbers; it simulates the subjective quirks of human preference, taste, and value across vast market segments. Forget waiting months for focus groups or surveys. Simile lets you get a read on millions of potential customers, practically overnight.

Park told the TBPN podcast that Simile's models are “trying to be as human as possible to actually represent the human values, preferences, taste, all the subjective half of human brain.” They've already scaled to simulate "tens of millions" of people, with a goal of hitting a billion. This isn't theoretical: Fortune 10 and 50 retailers, financial institutions, and CPG giants are already using this tech. They're testing new products, validating messaging, and confirming market assumptions at a scale previously unimaginable.

Key Takeaways

  • Simile's AI creates sophisticated "foundation models" that predict human behavior by simulating subjective values, preferences, and tastes across various market segments.
  • The company’s models currently mimic “tens of millions” of individuals, aiming for a billion, providing access to market insights at an unprecedented scale.
  • Fortune 10 and 50 retailers, financial institutions, and CPG companies are already customers, using Simile to test new products, messaging, and validate market assumptions.
  • This technology democratizes access to robust market testing, allowing businesses of any size to “filter down to any population of their interest” and get tailored feedback.
  • Founders can use Simile’s AI Human Behavior Simulation Process to rapidly test product concepts, marketing messages, or even website mockups without expensive, slow user research.

The Simile's AI Human Behavior Simulation Process

Simile’s approach offers a clear, repeatable path to rapid market validation:

  • Step 1: Define Population of Interest: filter down to a population of their interest so they can describe the population to be whatever they want it to be let's say male living in California in their 30s
  • Step 2: Ask Questions: literally ask any questions that can be a behavioral environmental questions
  • Step 3: Provide Assets (Optional): send images if it's image asset videos if it's a video like advertisement. It can even be a product demo whether it's a Figma or real website
  • Step 4: Receive Feedback: our agents will actually traverse through those websites and Figma mock mocks and basically give feedback once they have seen it.

When This Works (and When It Doesn't)

This process shines when founders need rapid, scalable feedback on early-stage product concepts, marketing messages, or even UI/UX mockups. Jun Park notes it's used for “replicating what they know to be ground truth,” “test[ing] new markets,” and “new product message testing.” It truly democratizes market research, offering a scalable way to get access to people for feedback and insights previously reserved for companies with huge research budgets.

However, simulated behavior, while advanced, isn't a perfect substitute for real-world user interaction and qualitative insights. It's best for quantitative validation and exploring broad market reactions. It might fall short when deep emotional responses, subtle usability issues, or truly novel, un-simulatable interactions are critical. Relying solely on simulations without any real user touchpoints could lead to blind spots, missing nuanced feedback only real users can provide.

What to Do With This

Imagine you've built an MVP for a new productivity app targeting young professionals. Instead of costly focus groups or slow A/B tests, use Simile's framework. First, Define Population of Interest: target “young professionals, 25-35, living in major US cities, interested in remote work tools.” Next, Provide Assets: upload your Figma prototype or a short video demo of your app. Then, Ask Questions: “Would you pay a monthly subscription for this app? What feature would you use most? What’s missing?” Finally, Receive Feedback: review the simulated agents’ feedback and behavioral data. Use these insights to iterate on your app’s features and pricing strategy before writing more code or launching to real users. This gets you market validation in days, not months.