← Back to portfolio
Field Note Micro-interaction Hover State Copilot Home Page

Icon Metamorphosis

Logged · September 2026 Surface · Microsoft Copilot Duration · 0:62

This is not what I expected this next level example of intelligence in LLMs to look like, taking the AI race to unprecedented heights, to the point of mimicking how magnets attract metals.

As a product manager, spotting and surfacing uncommon UX phenomena has become second nature to me, the same instinct that drives good engineering and QA work: noticing the small, easy-to-miss details that most people scroll straight past.

Just take a look at the following one I captured not too long ago. I promise you'll scroll back up more UX-informed (and AI-informed) than you were before, just like I did after capturing this.

Recorded observation — Copilot personalized home page, hover behavior on the flyout panel

On Copilot's personalized "Home Page," there are two popovers contained within a single container sitting below the suggestion panel. Each one displays a "Try it" icon rendered as a bright, lit yellow lightbulb, anchored into a green background.

Layout as observed
Popover A
Popover B
Two popovers, one shared container, sit below the suggestion panel

Here's the fascinating part: the longer your cursor lingers over one of these flyout panels, whether positioned to the left of the first one or below the second, the lightbulb icon automatically metamorphoses.

It sheds its yellow-on-green look and transforms into a peachy-orange identity-card icon, set within a soft cream card body and border. Simultaneously, the prompt text beside it shifts too, changing from "Try it" to "Tell me more."

A small, easy-to-miss detail, but a genuinely delightful piece of micro-interaction design, a hover-triggered transition that nonchalantly goads the user from a passive invitation ("Try it") to a warmer, more conversational one ("Tell me more"), reinforced by a coordinated color and iconography shift that feels deliberate rather than random.

Very fascinating phenomenon, indeed. As AI systems get folded into more of these everyday touchpoints, moments like this raise a real question for SMEs, engineers, and other stakeholders alike:

As we celebrate how far AI has come, how do we make sure a human in the loop stays responsible for tracking and tracing these metaphysical phenomena, and who gets credit (and remunerated) for noticing them first?

This field note published as a Confluence page for internal reference
Also documented internally in Confluence — same observation, formatted for team reference

Notes like this: system-thinking exercises and the UX fails I QA-test in the wild land roughly every two months. Subscribe to the newsletter ↓