At Amazon Devices, I worked on a product launch where conversion looked strong, and we were sure we found a winning solution. Until we started reading user forums. People were asking questions about features that didn’t exist in the product they bought, because they hadn’t realized they selected the wrong model. I came to call this Accidental Conversion: when a product successfully drives an action before the user has fully made the decision behind it. The metrics said success; users were confused. For 2 decades, we taught products to remove friction. AI is accelerating that trend, making systems better at predicting, recommending, & completing actions. But it also creates a new blind spot: products can become increasingly good at getting users to “yes” before they actually decided. In this talk I'll share 2 Amazon cases where healthy metrics hid this pattern, what it costs when you miss it, and how to think about conversion differently in products where AI is doing more of the work.
Group Design Site Manager, Intuit