Simile is a startup that sells AI models of real customer groups to large companies. It works like a survey whose respondents are computer models of real people. A team types a question, such as which flavor and price adults aged 21 to 34 would pick, and gets the share of that group choosing each option. A second model estimates how far off each answer is likely to be and shows a confidence label from Low to High. Simile trains the models on interviews, surveys and shopping records from consenting participants: the CVS Health model uses 2.9 million responses from more than 400,000 people.
A CVS Health vice president told a Wall Street Journal summit that traditional research reaching patients and healthcare providers usually takes six to eight weeks and took about 20 minutes with Simile. The executive said CVS Health checks the models against its survey panels. Chief executive Joon Sung Park told the Journal that access costs $150,000 to millions of dollars a year per customer. Simile reports that a result labeled High confidence has a 95% chance of coming close enough to real behavior to guide a decision, against 38% for Low.
Polling firm Gallup built test models from interviews with about 1,000 members of its survey panel and checks them against real panel results. It will not use simulated answers in its published population estimates. In July 2026 Simile raised over $200 million at a $2 billion valuation to work toward simulating all eight billion people.
Read More: MiroFish, a free tool that makes simulated AI characters react to a news item or draft policy
Sources:
- Simile: Announcing Our Series B
- Simile: Building confidence in Simile
- WSJ: Can AI Replace Humans for Market Research?
- WSJ: How Digital Twins Help CVS Health Move More Quickly and Confidently
- Gallup: Gallup Begins Research on Simulated Responses
Disclaimer: For information only. Accuracy or completeness not guaranteed. Illegal use prohibited. Not professional advice or solicitation. Read more: /terms-of-service
Reuse
Citation
@misc{kabui2026,
author = {{Kabui, Charles}},
title = {Simile: {AI} {Models} of {Real} {Customers} {That}
{Companies} {Question} {Before} {They} {Launch}},
date = {2026-11-09},
url = {https://toknow.ai/posts/simile-ai-models-of-real-customers-test-decisions/},
langid = {en-GB}
}
