Market research has always relied on something scarce: people's time and patience. Assembling panels, designing surveys, waiting for responses, analyzing data. A process that could take months and cost millions. Now, a startup is proposing a radical shortcut: replacing those people with digital versions of themselves.
Simile, founded in 2024 as a Stanford spin-off, trains AI agents with data collected from conversations with real humans to create what it calls "agentic twins": digital clones that replicate the preferences, personality, and behavioral patterns of specific individuals.
The company has just closed a $100 million Series A round led by Index Ventures, and already counts CVS Health and Gallup among its clients.
How it works
The process begins with interviews with real people. From this data, Simile builds an AI agent that acts as a digital representation of each individual. It then combines that information with behavioral and purchase data to enhance the model's accuracy.
Companies can access this bank of agents and ask them unlimited questions, without worrying about respondent fatigue or recruitment costs. Annual access costs between $150,000 and several million dollars, depending on use.
"I don't have to stop with the number of questions I asked. There is no fatigue," said Sri Narasimhan, vice president of customer experience at CVS Health.
The CVS case
CVS has been using Simile's agentic twins for over a year. Its database is built from 2.9 million responses from over 400,000 consented real people. The company calibrates this data with its own information, such as historical survey responses and customer support interactions.
In internal tests, CVS found that its digital twins replicated known findings with up to 95% accuracy. The company has used them to explore topics such as medication adherence, communication with customers about veterinary medicine, and the behavior of hard-to-reach populations, such as patients with chronic diseases.
CVS Health Ventures, the company's venture capital arm, is also an investor in Simile.
Plans include expanding its twin base to more than 100,000 agents and using them to evaluate store designs and new products. "The trade-off is clear: if we invest in agentic twins, we need less traditional panel research," Narasimhan said.
Gallup enters the game
Gallup announced an alliance with Simile to offer more than a thousand digital twins to their joint clients. Application areas include public policy research, trend analysis, workplace well-being, and job satisfaction.
"This allows us to go deeper in a more scalable way than we ever have been able to, with fewer financial barriers," said Joe Daly, global partner at Gallup.
Limits and precautions
The enthusiasm has nuances. Evan Brown, emerging technology analyst at Gartner, warns that the technology is still early. While marketing research is a low-risk area for experimentation, it is not close to fully replacing traditional processes.
CVS knows this. Narasimhan insists that the company will continue to validate twin results against real human responses, and that its researchers must detect inconsistencies and ensure the validity of the findings. He stated that:
"We are never going to stop talking to real customers"
Simile, for its part, implements controls such as role-based access and content monitoring to reduce the risk of hallucinations or inappropriate responses.
The future envisioned by Joon Park, co-founder and CEO of the startup, includes simulations where multiple agentic twins interact with each other in real-world scenarios. For now, the research market is the entry point. What comes next is, literally, yet to be seen.