Multi-modal input
When a task is easier to show than to type, accept whatever format fits, text, an image, a file, rather than forcing everything through one channel.
Relevant principles
What the research says
Accepting text, images and files lets users hand over the task in whatever form they already have it, which cuts the effort of getting started. Reduced friction is a known antecedent of trust, but there is no direct AI-trust study on input modality, so treat this as principle-backed. 1
Interactive demo
Bookkeeping assistant
What are we filing today?