Simile $200 million funding values AI research startup at $2 billion
Simile raised $200 million at a $2 billion valuation as it builds AI-generated consumer panels, though its accuracy claims remain unverified.
By Marcus V. Thorne · Markets Editor
· 3 min read
Simile’s $200 million funding round has valued the Palo Alto artificial-intelligence startup at $2 billion. The Series B, announced July 30 and led by Greenoaks, followed a $100 million Series A announced about five months earlier, according to The New York Times and TechCrunch.
Simile sells simulated consumer panels to companies seeking to test products, brands, marketing, pricing and other decisions. Its central proposition is that AI-generated agents, built using research collected from people, can produce usable indications of how a selected population might respond before a company conducts, or alongside, conventional customer research.
Greenoaks led the latest financing. Index Ventures, Hanabi, A*, Bain Capital Ventures, CVS Health Ventures and Definition also participated, according to the Times. The company’s reported clients include CVS Health, Deloitte and Wealthfront. Simile also works with Gallup, according to reporting republished by The Star.
How do Simile’s AI consumer twins work?
Simile calls its simulated users “agentic twins.” They are not individual digital replicas used for personal assistance or customer-service personalisation. Instead, they are software agents intended to represent groups of people in a research exercise.
The company initially conducted two-hour interviews with 1,000 participants selected to represent a broader population, then used that material to develop a model of human behaviour. It compared the model’s answers with later responses from the human panellists, The Star reported. Simile subsequently added input from a few million people worldwide through Gallup and other partners.
Customers select a target population through a web interface, and Simile runs a simulation involving panels from hundreds of thousands to millions of agents. Its charges depend in part on the number of agents modelled, with larger panels costing more, chief executive and co-founder Joon Sung Park said.
CVS Health used data from 400,000 AI-generated agents while examining ways to improve prescription filling and adherence, according to the Times. That application illustrates the intended use: rapidly testing possible explanations or interventions for a defined customer group rather than collecting every response from a live survey panel.
Can AI consumer panels replace customer surveys?
The available evidence does not establish that they can replace conventional surveys. Simile says its technology can predict behaviour and offer results more quickly and cheaply than traditional studies. Mihika Kapoor, its head of product, said the system’s responses were 85% to 99% accurate, varying by scenario and population. Those figures are company statements, not results from an independent audit or benchmark disclosed in the reporting.
That distinction is material because market research is designed to measure behaviour that can be difficult to predict. TechCrunch described Simile’s broader stated ambition to simulate all eight billion people accurately and honestly as implausible, while characterising simulated users for research as a promising area.
Simile is competing in a developing field that includes Helm, Artificial Societies, Rehearsals and Aaru, according to The Star. Park has said Simile also plans “dynamic simulations” that follow agents over time, such as modelling product use during a week.
The funding gives Simile capital to develop that approach, but it does not independently validate its product claims. MLQ reported that the company had not provided enough underlying commercial validation data for outside parties to broadly test its performance assertions.
This story draws on original reporting from NYT DealBook.
