Most live chat waits. The widget sits in the corner of the page and a conversation only starts when a visitor decides to open it. That serves people who already know what they want to ask, and it misses everyone else: the person comparing your plans for the third time, the one who stalled halfway through setup, the one who leaves without ever naming what stopped them. A proactive message reverses the order — your team opens the conversation, at a moment it chooses, for an audience it defines.
Why waiting for the first message costs you conversations
An inbound-only queue is a self-selecting sample. The visitors who write to you are confident enough to type a question to a company, patient enough to wait for an answer, and convinced enough that the answer will be worth the wait. Everyone whose hesitation is smaller than that threshold simply closes the tab. Nothing arrives in the inbox, so nothing looks broken.
That gap is expensive precisely because it is invisible. Support metrics describe the conversations that happened: first-response time, resolution time, satisfaction. They cannot describe the question that was never asked. A team can hold a five-minute first response all quarter and still lose most of the people who needed one sentence of reassurance before they would commit.
None of which is an argument for interrupting everyone. A message that appears on every page for every visitor is closer to a pop-up advertisement than to support, and people learn to dismiss it without reading. The useful version is narrow: a specific audience, on a specific page, after enough time on that page to suggest genuine consideration rather than a passing click.
What makes a proactive message worth reading
Start with the headline, because it is the only part guaranteed to be read. It should name the situation the visitor is actually in — comparing plans, importing data, waiting on a delivery — rather than announcing that your team exists. "Choosing the right plan?" invites an answer. "How can we help you today?" asks the visitor to do the work of framing their own problem before they have decided you are worth the effort.
The message should also come from a person. In Lavenity the sender's name and picture are resolved from the teammate who created the message, so the greeting carries a real face and a real name that will still be there when the visitor replies. Two variables cover the rest of the personalisation: the recipient's first name and your customer-facing workspace name, each with a sensible fallback when the value is not known.
Target by audience, page, and time on the page
Three rules do most of the targeting work. Audience narrows by relationship: everyone, anonymous visitors, leads who have shared contact details, or existing users. The page path narrows by intent, either an exact path such as /pricing or a prefix such as /docs* for a whole section. Time on the current page narrows by attention, and a delay of roughly ten to twenty seconds is what separates a reader from a bounce.
Combining the three keeps volume low and relevance high. A trial reminder shown to users on the billing page after two seconds is a completely different message from a first greeting shown to anonymous visitors after fifteen. Written as one generic broadcast, both become noise. Written as two narrow rules, each reaches a few dozen people a week who genuinely needed it.
Delivery, repeat protection, and the reply
A message stays a draft until you set it live, and pausing it later parks the copy, audience and schedule without discarding any of it. Delivery begins either the moment it goes live or at a scheduled date and time. Each person receives a given message once: the delivery is reserved for that visitor before the message is shown, so a reload, a second tab or a return visit tomorrow will not repeat it.
What happens after delivery matters more than the delivery itself. A proactive chat message lands in the visitor's conversation as a message from your teammate, takes ownership of that conversation when nobody owns it yet, and steps the AI assistant aside so a person is expected to continue. The reply then arrives in the shared inbox like any other request, with the original message visible as the first turn, so whoever picks it up can see exactly what was promised.
Ship your first proactive message this week
Pick one moment where you already know what people ask. The support archive usually makes the choice obvious: the pricing question that returns every week, the setup step where sessions quietly end, the delivery question that follows every order. Start from a template that matches the format, rewrite the headline in the words your customers actually use, and keep the body to one sentence with one clear next step.
Set the rules deliberately and write down why. Audience, page path and delay are three separate decisions, and each one should have a reason you can state out loud. Then check who is on the other end: a proactive chat message expects a human answer, so publish it for hours when somebody is on shift and route the conversations it creates the way you would route any first contact.
Read the results after a week rather than a day. The number of people reached shows scale, but the questions that matter are whether the conversations were useful and whether they arrived in a queue that could absorb them. When replies are thin, the moment is usually wrong before the copy is — move the delay or the page rule before rewriting the message. When the queue strains, narrow the audience instead of pausing the whole programme.
Frequently asked question
Will proactive chat messages annoy visitors?
They do when they are broadcast to everyone on every page. Targeting is what separates a useful message from an interruption: a narrow audience, a page path that matches real intent, and a delay long enough to show genuine attention. Each person also receives a given message only once, so someone who ignores it is not asked again on the next page or the next visit.