A startup is building digital copies of shoppers for companies to quiz

A startup called Simile is building AI simulations of people that businesses can quiz about products, ads and more. CEO Joon Sung Park described the idea on the Sourcery podcast, which released on Monday. Simile says it has raised $300 million in total at a $2 billion valuation.

How it works

Park said customers use a web page to pick who they want to talk to and what they want to ask. The system then simulates what those people might do. He called it a foundation model of human behavior, meaning one large AI system trained to predict how people act.

CVS has worked with Simile to study how shoppers respond to store layouts and messaging. Park said the company also works with financial institutions and entertainment companies.

Where the data comes from

Park said partnerships with groups such as Gallup, which can reach real people who agreed to take part, give Simile data from more than 40 million people worldwide. He also said the company trains its models on behavioral data, including transaction and mobility data.

That means purchase records and movement patterns can feed tools that predict what people like you will do. Retailers such as Amazon and Walmart already use AI shopping assistants and customer data to personalize what shoppers see.

Several things are not clear from the interview. It does not say where the transaction and mobility data comes from, how accurate the simulations are, or whether any clone is tied to a specific person. It is also unclear what choices shoppers get about being included.