Salesforce agrees to buy Listen Labs to run real and simulated customer interviews in its CRM

Listen's two products go wrong in different places. An interview returns what a recruited participant actually said, so its errors come from who was recruited ^[9]. Salesforce cites a panel of more than 50 million people who can be interviewed in more than 120 languages ^[4]. Fifty million is the size of the pool. The number of people in a typical study, and whether a given sample matches the market being asked about, are separate figures ^[4].

A twin's errors come from extrapolating past the behavior it was grounded in ^[5]. For a twin's answer about a new product to carry over, two things have to hold. The idea under test has to resemble what those customers already did. Someone also has to have measured the gap, by putting the same question to twins and to real participants and comparing the answers. Salesforce's announcement does not explain how the simulations will be validated ^[8]. The obvious place to run that comparison is Listen's own record of more than one million interviews ^[11].

I think the interview half is the stronger engineering bet, because every answer traces back to a person. It also came from a real problem. In January 2023 an AI avatar app the founders had built drew 20,000 users overnight, and they did not know who those users were or why they had come ^[19]. A hackathon project to interview them became the foundation for Listen ^[19].

Aman Naimat, president of Salesforce AI Labs, said Listen had "reimagined" market research "with AI agents that can conduct meaningful research and uncover context behind customer and user feedback" ^[16]. He also said: "We look forward to expanding access to these agentic capabilities through Salesforce." ^[16]

Listen's only published revenue figure is a ratio: 15-fold growth in annualized revenue in the nine months after launch ^[11]. Fifteen times an undisclosed base is still an undisclosed number. The valuation path comes from press reports. The Stack put Listen's last valuation at more than $500 million after a $69 million Series B in January ^[15]. On September 9, TechCrunch reported that Listen had signed and then walked away from a $125 million Series C term sheet at a $1.5 billion valuation ^[12]. The same report put the proposed Salesforce deal at around $2 billion ^[13]. According to The Stack, Business Insider first reported talks at that figure ^[14]. If $2 billion holds, Salesforce is paying at most four times the January valuation ^[1] and about a third more than the term sheet Listen turned down ^[2].