The Confidence Curve

I’ve been having a bit of a debate with another member of my AI/ML incubation team at work over how confident we should be when surfacing in-product intelligence. Once we start making attempts at enhancing and augmenting user notes, should these attempts surface more as suggestions, or should we act confidently and then ask for forgiveness if necessary?

My teammate believes in assertively trying things out and then monitoring user behavior around what we did and adjusting from there.

While I am all for incorporating user feedback, I am much more tentative about the initial user contact - I want us to acknowledge that there will inherently be some-to-a-lot of uncertainty in any prediction we might make. And especially with sensitive content like personal notes, brusquely acting on user content feels like the quickest road to feelings of creepiness and distrust in our product.