Artificial Intelligence
An AI feature earns trust at every stage, from the first impression to the follow-up. The principles below walk through that lifecycle and what each moment asks of the design.
Before the interaction
Set expectations before a merchant engages. Explain what the feature does in terms of outcomes, not technology. Be upfront about what data it uses.
During the interaction
Always acknowledge that input was received and processing is underway. Never leave a blank or frozen state.
Merchants must be able to review output before it takes effect. Every suggestion should be previewable, editable, and dismissable.
AI-generated content should never replace human-generated content without review or approval.
When something goes wrong
Acknowledge errors directly. Give the merchant a clear path to correct, dismiss, or retry from a clean state.
Writing for AI
Write in plain, conversational language. Avoid technical jargon about the underlying model or infrastructure. Speak to the merchant's goals, not the system's mechanics.
Use active voice. It makes AI responses feel direct and accountable.
Never oversell what a feature can do. Be honest about limitations upfront rather than letting the merchant discover them through failure.