Companies that take too long to go to market will lose out. At the same time, you can't ship something where the output is wrong 5% to 10% of the time, which is what we see now with many LLM-powered products.
One customer of ours, for example, is training an AI accountant with an accuracy rate of 96%, compared to human tax pros who are accurate 94% of the time. That level of rigor is critical for bringing AI products to market, and all companies need to learn how to balance shipping fast with features that customers actually trust.
For example, many of our customers use Amplitude to personalize messaging with a feature we call predictive cohorts. We first built a tool that customers could use to identify non-converting users and send them a promotional message to push them to convert. After speaking with customers, we learned they’d prefer the ability to predict, based on behavior, which users would likely not convert.
With any new tech, there will be an inherent lack of consumer trust in the early days. Companies that peel back the curtain will secure trust and win more customers faster.