Simile

Simile develops AI simulation software that models how people — customers, patients, employees or citizens — are likely to respond to a decision, allowing organizations to run scenarios at scale before committing budget or shipping a change. The company describes its product as the first AI simulation of society, populated by generative agents grounded in real humans, and is building a foundation model intended to predict human behavior across situations and scales. Reported use cases include simulating a product launch, a price change or a campaign, rehearsing earnings calls, modeling litigation outcomes, and testing policy changes.

According to the company, each simulated population begins with real people: proprietary algorithms design large-scale studies on an ongoing basis, and the resulting data is combined with behavioral and transactional datasets to train the models. Simile says it validates weekly against real humans through more than 7,000 evaluations spanning subpopulations and enterprise use cases, measuring distributional distance, effect sizes and accuracy against observed behavior, and that a separate confidence model tags each result with a predicted accuracy level. Models are retrained continuously on new behavioral, macroeconomic, pricing and policy data, and agents are described as consuming media in real time. Enterprises can also opt in to contribute their own first-party data — loyalty, balance or telemetry records — to train custom models.

Coverage of the company reports tens of millions of simulations run for Fortune 100 organizations, revenue growth of 5x in the five months following the February 2026 public launch, and headcount above 50. Named customers and partners include CVS Health, Wealthfront, Deloitte, Gallup, Banco Itaú, Suntory Beverages & Food and Garnett Station Partners.

Founding story

Simile was founded by Joon Sung Park, a Stanford PhD graduate whose doctorate was advised by Michael Bernstein and Percy Liang, after his 'Smallville' research project in which 25 AI agents lived simulated lives in a virtual town. The resulting paper, 'Generative Agents: Interactive Simulacra of Human Behavior,' won the Best Paper Award at UIST 2023 and became the basis for the company. Park co-founded Simile with Bernstein (Stanford HCI professor), Liang (director of Stanford's Center for Research on Foundation Models) and Lainie Yallen; all four worked together at Stanford before spinning the company out.

Business model

Simile sells enterprise access to its simulation platform and foundation models, positioning itself against traditional market research and consumer insights spend. Customers can also contribute opt-in, customer-governed first-party data to train custom models. Sources do not disclose pricing or contract structure.

Enterprise commercial relationships with large organizations; sources report revenue grew fivefold in the five months after launch but do not disclose absolute revenue or pricing mechanics.

Traction

Reported traction includes tens of millions of simulations for Fortune 100 enterprises, 5x revenue growth in five months, more than 50 employees, and named customers and partners CVS Health, Wealthfront, Deloitte, Gallup, Banco Itaú, Suntory Beverages & Food and Garnett Station Partners. CVS Health is reported to have built 100,000 AI patient twins from 2.9 million records to test pharmacy experiences, and Wealthfront's head of user research states the platform expanded qualitative research scope 15x. Simile was also selected for the Disney Accelerator.

Latest developments

On 30 July 2026 Simile announced a $200 million Series B at a $2 billion valuation led by Greenoaks, with Index Ventures, Hanabi, Bain Capital Ventures, A*, Factory and CVS Health Ventures returning and Definition joining. The capital is intended to fund training of the core human-behavior foundation models, expansion of simulation compute infrastructure, and growth of commercial engineering teams across healthcare, financial services and media. The company also recently launched a product allowing organizations to act on predictions directly, and was selected for the Disney Accelerator.