By Belle Lin
AI agents can now be used to predict and simulate human behavior -- a capability that startup Simile is tapping to help large companies like CVS Health and Gallup replace people for polling and market research.
Simile, which recently raised $100 million in Series A funding led by Index Ventures, is part of a new crop of startups using artificial intelligence to upend the way businesses have traditionally gathered feedback from customers and conducted market research.
Market insights is a roughly $150 billion industry that has long relied on conventional consulting and market research companies. But AI is sweeping rapidly through the sector, and the AI-simulated "people" offer customers quick, affordable access to market research -- essentially a shortcut instead of going the human route, said Seema Amble, a partner at Andreessen Horowitz.
Simile uses data gleaned from chats with human beings to train AI agents, who then become the digital twins of those people, said Joon Park, the Palo Alto, Calif.-based startup's co-founder and chief executive.
The AI agents are essentially digital clones of real individuals, who are interviewed to gather their preferences, personality and other traits. Simile then combines that data with participants' behavioral and purchase data to ensure "generalizability and visibility into people's thoughts and behaviors," Park said.
Previously, businesses contracted with consulting or market research firms to learn about their customers -- a costly process that could take months. Now, they can query Simile's online bank of agents, access that can cost between $150,000 to millions for each customer annually, Park said.
Customers can then ask infinite questions of their AI people, making the work of customer research cheaper and faster. The company has developed its own AI model for behavior prediction and combines it with various open-source models.
Simile was spun out of Stanford University in 2024 after Park said he saw immediate interest from large companies in simulation technology. He and two of his co-founders were among the authors of a 2023 paper on simulating human behavior.
AI-generated people have also been put to work in other areas like clinical trials, focus groups, even modeling clothing.
Over the past year, CVS has used Simile's AI-simulated people, or "agentic twins," to answer questions about its customers rather than solely asking panels of human participants.
CVS said its digital twins are based on 2.9 million responses from over 400,000 real people with their consent. The company then calibrates those responses with first-party data, such as historical survey responses and support interactions. In testing, CVS found that agentic twins replicated known findings with up to 95% accuracy.
CVS Health Ventures, the venture capital arm of the health giant, is an investor in Simile. The firm declined to disclose the amount of its investment.
Sri Narasimhan, CVS's vice president of enterprise customer experience and insights, said the company thinks of the agentic twins as "always on" compared with humans. "What I think was a really big unlock for us is we could go deeper," he said. "It's not like I have to stop with how many questions I asked. There's no fatigue."
As one example, CVS has tested how the agentic twins adhere to medication guidelines. By going deeper in questioning, CVS reaffirmed that people are commonly concerned about things like being able to reach a pharmacist and manage refills.
The company has also simulated hard-to-reach populations like providers and patients with chronic conditions. As well, it has tested the efficacy of customer messaging for pet medicine -- finding that people don't consider giving their pets medication a chore, and are looking for services like coordination with veterinarians, Narasimhan said.
CVS plans to expand its current roster of AI-simulated people to over one hundred thousand agents, and query those agentic twins in areas like store layouts and new product designs. "The trade-off is probably, if we're going to invest in the agentic twins, we don't need as much of that traditional market panel research," Narasimhan said.
The digital people are also finding their way into surveys and polls.
Gallup is launching a partnership with Simile in which the companies will make over 1,000 AI-generated digital twins available to their joint customers, said Joe Daly, a Gallup senior partner and member of its management committee.
Some of the areas Daly expects customers to be interested in are policy research, trend analysis and corporate research. The two companies are also planning to delve into topics like health, well-being and job satisfaction, he said.
"This allows us to go deeper in a more scaled-out way than we ever have in the past, with less of a financial barrier," Daly said.
In the future, Park said he expects Simile customers will require "multi-agent simulation," where agentic twins interact with each other. This will eventually occur in real-world settings to simulate more complex scenarios, he said.
For now, market research is one of the easiest places for AI-generated people to start making an impact, according to Evan Brown, an emerging tech analyst at Gartner. Testing marketing messages with AI-simulated people is a relatively low-stakes task compared with healthcare settings, Brown said, and offers some leeway in case the AI goes awry.
"It's still early in the sense of, you're not going to be replacing traditional processes with [AI agents]," he said. "You still want to be going out and collecting that data yourself."
Simile said it employs guardrails such as role-based access controls and monitoring for unsafe or sensitive content.
To help prevent AI hallucinations or other inaccuracies, CVS's Narasimhan said the company backtests the agentic twins with real people's responses, and relies on its researchers to flag possible inconsistencies and ensure the validity of results. And just like humans require an expert's touch in asking questions, so, too, do AI agents, he said.
"It's important for us to constantly have that flow of testing and making sure it's tested against real humans," Narasimhan said. "We're never going to stop talking to real customers."
Write to Belle Lin at belle.lin@wsj.com
(END) Dow Jones Newswires
March 06, 2026 07:00 ET (12:00 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments