Live chat for your website

The chat widget is the fastest channel to set up and the one most often set up badly. A chat bubble nobody answers is worse than no chat bubble at all — it promises immediacy and then fails to deliver it. Turn it on when you have someone to staff it, or when Tola AI can answer on its own.

A small chat launcher sits in the corner of your website. A visitor clicks it, types, and the message arrives in your Tola inbox within a second. Your agent replies from the same screen they use for WhatsApp and email, and the visitor sees the reply live.

Because it runs on your own website, live chat gives you context no other channel does: which page they were on, how they found you, and — if they are signed in — exactly who they are and what they have bought.

Installing it

Go to Channels → Add channel → Website.
Give it a name, your website address, and a welcome message.
Choose which agents get access to it.
Tola generates a short snippet of code. Send it to whoever manages your website and ask them to add it to every page.

That last step takes a web developer about five minutes, or can be done through a tag manager without a developer at all. Once it is live, open your website, click the bubble and send a test message — it should appear in your inbox immediately.

Ask for the widget to be added to every page, not just the contact page. Most chats start where the customer got stuck, which is rarely the page you would expect.

Recognising signed-in customers

If your visitors log into your product or store, your website can tell Tola who they are — name, email, phone — so the conversation attaches to their existing customer record instead of arriving as an anonymous visitor.

Your website team can also pass through details that matter to your business: what plan they are on, how much they have spent, whether they have an outstanding invoice, when they signed up. Those become custom attributes on the contact, which means you can then route, prioritise and report on them.

This is worth asking for. The difference between an agent seeing "Visitor #4471" and seeing "enterprise customer, two open invoices, signed up in 2023" is the difference between a generic reply and a useful one.

For anything sensitive, ask your team to enable identity validation as well. Without it, a technically-minded visitor could claim to be another customer; with it, that is not possible. If you handle payments, accounts or personal data, it is worth doing before you launch rather than after.

Pre-chat form

The pre-chat form asks for details before the conversation starts — typically a name and email so you can follow up if the visitor leaves.

Ask for as little as possible. Every required field is a reason to close the widget instead of using it. Email alone is usually enough; collect the rest inside the conversation, once the customer is already talking to you.

Business hours

Set your working hours per weekday, with a timezone. Outside them, the widget shows your out-of-hours message and tells visitors when someone will reply, instead of leaving them waiting in an empty chat.

Business hours also feed your SLA measurements, so overnight silence does not count against your response times.

Appearance

The widget is configured to match your brand: colour, position on the page, launcher text, avatar and the shape of the bubble. You can also remove the "powered by" branding entirely, so it reads as part of your own product.

Proactive messages

Live chat is the one channel where it is reasonable for you to start the conversation. An ongoing campaign sends a message when a visitor has been on a particular page for a set time — a pricing page, a half-finished checkout, a documentation page people get stuck on.

Used sparingly this converts well. Used on every page it is the reason people install popup blockers. Pick two pages where a stuck visitor is genuinely likely, and leave the rest alone. See campaigns.

Staffing it properly

Live chat is judged on speed in a way email is not. A visitor who waits four minutes leaves; the same person will happily wait four hours for an email reply.

Three things keep it honest:

Set business hours you can actually meet, and let the widget say so outside them. Under-promising costs you nothing.
Cap concurrent chats per agent at two or three. See assignment and capacity — an agent juggling six chats is doing all six badly.
Consider putting AI on this channel first. It is the lowest-stakes place to start, and it is where instant answers matter most. See Tola AI.