A chat bubble that just sits in the corner gets ignored — most visitors never open it, so they never discover the assistant can actually help. This sample adds the engagement layer: the assistant invites the visitor in and hands them a few real questions to tap, so the goal — a first message sent — happens in one click.
A small, tailored invite appears on its own after a few seconds (or on exit-intent), so the visitor doesn't have to discover the chat themselves.
Three or four tappable questions, drawn from the business's knowledge doc, remove the blank-box freeze. One tap is the first message.
"Ask anything — if I can't help, I'll connect you with the team." Saying it out loud beats the "it's just a dumb bot" reflex that trains people to ignore chat.
Same one-paste floating bubble as Sample 2, plus a small config block: the proactive nudge text and the pre-generated starter questions. Drop it in once per site; the heavy chat iframe still lazy-loads only when a visitor actually engages.
<!-- eeZfi web chat — proactive engagement layer -->
<script>
window.eezfiChat = {
host: "https://YOUR-EEZFI-HOST",
businessId: "YOUR-BUSINESS-ID", // eeZfy hashtag or UniqueID
// Pops up on its own to invite the visitor in.
nudge: {
delayMs: 4500,
text: "Hi! Want to know if we take your insurance, or book a cleaning?"
},
// Pre-generated from the business's knowledge doc (editable in Admin).
// One tap sends it as the visitor's first message.
starters: [
"Do you take my insurance?",
"Book a cleaning",
"I have a toothache",
"Where do you park?"
]
};
</script>
<script src="https://YOUR-EEZFI-HOST/js/embed/eezfi-chat.js" async></script>
↘ This page runs the same experience inline so you can demo it without a host site.
When a business saves its knowledge document on Admin → AI Handle → Chat Service, eeZfi pre-generates three to five likely first questions and stores them with the business — so they load instantly with the widget, cost nothing per visit, and the business can edit them. The same document grounds the assistant's answers, so the question a visitor taps is one the assistant can genuinely answer well.
Success metric for this layer: first-message rate — the share of visitors who send a message. Nudge-shown, nudge-opened, and chip-tap events are logged so it can be tuned per business.
eeZfi · Front door to eZentral platform