Live Chat for Sales: Turning a Support Widget Into a Revenue Channel
Most chat widgets are bought by a support team and judged on support numbers. That is why so many of them never make any money. Sales chat is a different job with different rules: qualify in three questions, get a human in front of the buyer fast, and never lose the enquiry that arrives at midnight.
The short version
- Qualify in three questions. Anything longer and the visitor leaves.
- Escalate on buying signals, not on failure. Price and competitor questions go to a person immediately.
- After-hours is free money. The enquiry at 11pm costs you nothing to catch and everything to miss.
- Measure capture, not chatter. A conversation with no email attached cannot be followed up.
Businesses already treat chat as a sales tool
We scanned 4,278 small business websites across seven countries and found that only 7.5% ran a chat widget of any kind. What was more interesting was who did. Estate agents came top at 13.6%, hotels sat at 10.9%, and auto repair shops came last at 1.4%. Read that list again and it is not a ranking of technical skill. It is a ranking of what one enquiry is worth.
Businesses install chat when a single missed message costs real money. An estate agent who loses a viewing request loses a commission. A garage that loses one loses an oil change. Nobody decided this from a best-practice article. They worked it out from their own maths. You can read the full method and the country and vendor breakdowns in our live chat adoption study. The useful question is the same one those businesses asked: what is one enquiry worth to you, and how many are you currently dropping?
Sales chat is not support chat with a different greeting
Support chat is judged on how few conversations reach a human. Sales chat is judged on how many reach the right human, fast. Those two goals pull in opposite directions, and teams that run one widget on both goals usually end up doing neither well.
The practical fix is to split the behaviour by page, not by product. On help pages and order status pages, let the AI resolve as much as it can. On pricing, product and demo pages, treat every conversation as a lead and make the handoff to a person the easiest thing in the widget. Same tool, two rulesets. Our live chat widget lets you set a different greeting and a different escalation rule per page, which is the whole mechanism.
Qualify in three questions
Long qualification forms belong on a landing page, not in a chat window. A visitor opened the chat because they wanted an answer, not an interview. Ask three questions at most, and one of them must be for a contact detail. Here is the set we would keep, in order.
| Ask | What it tells you | Skip it when |
|---|---|---|
| What are you trying to fix? | Real intent, in their own words | Never. This is the whole question. |
| Roughly how many people would use it? | Deal size, in a band | One-off or low-value purchases |
| When do you need it working? | Urgency, and whether to route now or later | They already asked to speak to someone |
| Best email or number for you? | Whether the lead survives the tab closing | Never. Ask before answering the third question. |
| Are you looking at anything else? | Which objection to lead with | They are already on the checkout page |
Two rules make this work. Answer something before you ask for anything, because a bot that demands an email before it helps reads as a toll gate. And ask for the contact detail while the visitor still wants something from you, not at the end when they have already got their answer and gone.
Route to a human fast, and on the right signal
Most teams escalate when the AI fails. For sales, escalate when the visitor succeeds. A person asking about price has already decided they might buy, and handing them a polite bot answer at that moment is the most expensive mistake in the whole channel.
| Signal | What should happen | Why |
|---|---|---|
| Visitor asks about price or a discount | Hand to a human now | Price questions are buying questions |
| Visitor names another tool or supplier | Hand to a human now | Comparison is the moment you win or lose |
| Visitor viewed the pricing page this session | Offer a human straight away | Intent is already high |
| Same question asked twice | Hand to a human now | The AI has missed something |
| Annoyed tone, or a complaint | Hand to a human now | Nobody buys from a bot that argues |
| Opening hours, delivery times, specs | Let the AI answer | A human adds nothing but delay |
The handoff only helps if someone is actually there. This is where most small teams fall down: the alert lands in a dashboard nobody has open. EzyConn pushes the conversation into Microsoft Teams, Slack or Zoom, and your rep replies from the app they already have on screen. If you want the mechanics of a clean transfer, including what to pass across so the customer does not repeat themselves, we wrote that up in the guide to chatbot-to-human handoff.
After hours is where the easy money is
People research purchases in the evening. They compare suppliers on a Sunday. They fill in a contact form at 10pm and then buy from whoever replies first on Monday morning, which is usually not the business that took until Monday afternoon.
An AI that answers the basic question at 10pm, takes the email address and books the callback does not need to be clever. It needs to be awake. Set it to be honest about the hour, promise a specific response time, and make sure the morning queue is sorted by when the message arrived rather than by who shouted loudest. The night shift for a two-person business is not a rota problem. It is a configuration setting.
The five numbers worth watching
Chat dashboards are full of vanity metrics. Total conversations tells you nothing about revenue. These five do.
| Metric | How to work it out | Note |
|---|---|---|
| Contact capture rate | Chats where you got an email or number, divided by all chats | Fix this first if it is under half |
| Median time to a human | Seconds from the escalation to the first human reply | Under two minutes in working hours |
| After-hours share | Chats started outside opening hours, divided by all chats | Often a third or more |
| Chat-sourced pipeline | Value of deals where chat was the first touch | The only number finance cares about |
| Cost per captured lead | Monthly chat spend divided by leads captured | Compare it against your ad cost per lead |
Start with contact capture rate. It is the cheapest thing to improve, it does not depend on your sales team, and it is the number that turns chat from a cost line into a pipeline line. Once it is healthy, work on median time to a human. Everything else is reporting.
What it should cost
Sales chat has an awkward billing problem. Most tools charge per seat, and the people you most want in the widget, part-timers, the founder, the one rep who knows the product cold, are exactly the people you will refuse to pay a seat for. Others charge per AI resolution, so the channel gets more expensive precisely when it starts working. Both models tax the behaviour you want.
EzyConn is flat. The free plan is $0 forever with 1 agent seat plus the owner, 500 messages a month and one chatbot, which is enough to find out whether chat produces leads for you at all. Paid plans are Starter at $25 a month with 3 agent seats and Professional at $95 a month with 10, priced per workspace rather than per agent, and 20% off for annual billing. Full details and the message allowances are on the pricing page.
Frequently asked questions
Does live chat actually increase sales?
It increases the number of enquiries you capture, which is not quite the same thing. Chat catches the visitor who has one question standing between them and a purchase, and it catches the visitor who arrives at 11pm. Whether those turn into sales still depends on your product, your price and how fast a human follows up. Treat chat as a capture channel first and judge it on captured leads, not on a conversion rate you cannot attribute.
How fast do you need to reply to a sales chat?
Fast enough that the visitor is still on the page. In practice that means an instant first response from AI and a human within about two minutes when the conversation is a buying conversation. The number that matters is not average response time across every chat. It is the median time from the moment someone asks for a person to the moment a person replies.
Should AI or a human handle sales chat?
Both, in that order. AI is good at answering factual questions instantly, asking two or three qualifying questions and taking a contact detail before the tab closes. It is bad at negotiating, at reading hesitation and at handling an unusual request. Let AI open the conversation and hand over the moment the visitor mentions price, a competitor, a deadline or a discount.
What should I measure to prove chat is making money?
Five numbers cover it: contact capture rate, median time to a human, the share of chats that start outside opening hours, chat-sourced pipeline, and cost per captured lead. Contact capture rate is the one to fix first, because a conversation with no email attached to it cannot be followed up and cannot be counted.
How much does live chat for sales cost?
EzyConn starts at $0 forever with 1 agent seat plus the owner, 500 messages a month and one chatbot. Paid plans are Starter at $25 a month and Professional at $95 a month, priced per workspace rather than per agent. Annual billing takes 20% off. AI is included on every plan rather than metered on top.
Find out what your chat is worth
Run it free for a month: 1 agent seat plus the owner, 500 messages on the widget. Count the leads it captures after 6pm. That number decides whether you ever need a paid plan.
Start freeAdoption figures cited here are from our own August 2026 scan of 4,278 small business websites. EzyConn plan details are current as of October 2026.