A mother apologised to her AI. An AI's face is a setting, not decorat…
An AI's face triggers trust reflexes: we thank, apologise, watch less. What that changes when the AI acts for you, with Meta, Claude and OpenAI as examples, and a grid to decide the face.
You say "thanks" to ChatGPT without thinking. A face, a name and a smile push further: we apologise, we forgive, we watch less. Meta just gave one to its assistant Muse, and an episode tells of a mother who apologised to it. An AI's face is not decoration: it is a trust setting, and the more the AI acts for you, the riskier it gets. What you walk away with: a grid to decide an AI's face according to what it may do alone, and a five colleague test. What changes (designer, PM, PO, scrum master, Design System lead): the character is chosen with the list of actions, no longer by taste. Why now: Meta, Claude and OpenAI each make a choice about a face, or none, for their AI. The principle holds whichever AI you put in your product. · · · <!-- order:freeform --> A dated example: Jolly, Meta's…