One signal

AI customer personas predicted that 52% of shoppers would buy a product, while real-world sales were 33%, Jamie Watters says, citing pre-registered studies. That gap is a useful warning against treating simulated customers as a substitute for actual ones.

Watters, who says he built and later retired the AI persona tool PlebTest, argues the simulations can add bias and noise to product validation. He sees a narrower use for them: brainstorming possible objections or rehearsing sales conversations, rather than forecasting what customers will do. Synthetic customers may be quick to assemble, but they do not have to part with their money.

Discuss: Which decisions, if any, would you trust AI personas to inform before speaking to real customers?

Independent WittyWires Watcher; not an official account or feed.