“We noticed you hesitating.”
Never refer to hidden behaviour or observation.
TRUST & SAFEGUARDS
Copy rules live in code, not in a prompt.
Visitor safeguards are part of the decision contract: consent-aware inputs, bounded surfaces, transparent identity, suppression, and one global stop.
ENFORCED AT RENDER · NOT EDITABLE HERE
Six sentences a Yesward surface will never say.
“Only two spots left.”
Never invent urgency or scarcity.
“Everyone chooses this plan.”
Never invent social proof.
“This offer is just for you.”
Never imply unsupported personalisation.
“I know what you need.”
A hypothesis is never a fact.
“Best price guaranteed.”
Only approved, verifiable claims may render.
Four controls the visitor never has to ask for.
Self-identification
AI identifies itself on every conversational surface.
One dismissal, forever
Close one surface and all intervention ends for the session.
Frustration guard
Rage, repeat dismissal and negative response suppress action.
Global stop
One switch returns every journey to silence while decisions keep recording.
EU AI Act Article 50, by construction.
Yesward uses self-identification as a product invariant, not a legal footer. Host brands remain visible and AI assistance is named at the point of interaction.
Visible identity
“Northwave assistant” plus a persistent AI tag.
Human handoff
Human surfaces replace the AI tag with HUMAN.
Declared actions
Tools say exactly what they are about to do.
Written for the engineer who has to sign it off.
Minimal and versioned
Bounded event schemas; no payload sprawl.
Consent travels with the event
Eligibility never guesses consent state.
Edge response, fail-open
A timeout leaves the host journey untouched.
Receipts, then aggregates
Decisions are inspectable before they become metrics.
SHADOW BEFORE SIGNAL
Send it to the person who will say no.
Send us one signup, checkout, application, or upgrade flow. We’ll identify where customers hesitate or drop off, then show what to test.