Live chat
What is live chat and when does it reduce work?
Live chat is a conversation in a window on your own site. The channel lowers the threshold to ask, and that's exactly why volume goes up because of it.
In short
Live chat is a text conversation between a visitor and an agent or AI, held in a window on your own website or in your app. It differs from email in being synchronous: the visitor stays on the page waiting for an answer. That difference in expectation determines nearly everything about how you set the channel up.
Why chat raises your volume and can still save work
A chat window lowers the threshold. Questions someone would never have put in an email do get typed. Teams that switch on chat therefore nearly always see more contacts than before, and that's not a setup error.
The gain is in what those questions are. Many chat questions come from someone standing on your product page or in the checkout right now. A thirty-second answer there prevents an abandoned cart or a return because someone picked the wrong size. That same question by email would have arrived three days later, when the order is already wrong.
So don't look at the number of chats, but at the number of emails and calls alongside. If those stay flat while chat is added, you're serving a group that used to simply click away.
When do you switch the window on, and when not?
A chat button nobody staffs costs more than it returns. Someone typing at 9 in the evening who gets an email back an hour later received a promise you didn't keep. So outside your staffing don't show chat but a form, or let the AI handle the conversation, with the response time you actually meet stated.
Where the window sits matters too. On a product page and in the checkout, chat answers buying questions. On a shipping information page the same window mainly attracts WISMO questions you could have prevented with a tracking block on the page itself.
How many conversations can one person handle at once?
Two to three at once is the ceiling for most teams where quality holds up. Above that, waiting time appears inside the conversation: the customer can see someone is there and still waits two minutes for an answer. That feels worse than an email arriving tomorrow.
The staffing calculation is the same as for phone, only you divide by the number of simultaneous conversations. Don't forget a chat can carry after-work: a task, a return, a note on the customer card.
What AI should and shouldn't do in chat
Chat is the channel where AI pays off fastest, because the questions are short and the context is on the page. An agent that knows which product the visitor is viewing and which order belongs to the email address answers a large share of questions without intervention.
Where it goes wrong is with commitments. An AI promising an exception to your return policy, or naming a delivery time it didn't pull from a system, creates the most expensive kind of ticket: a customer with a screenshot. Define which topics the AI handles alone and which it hands over, and measure how often agents correct an AI answer before you move that line.
Live chat or chatbot?
| Live chat | Chatbot | |
|---|---|---|
| Who answers | An agent, optionally with an AI draft | Software, following a script or model |
| Availability | Within your staffing | Always |
| Exceptions | Can deviate and commit | Only what's been defined |
| Cost per conversation | Labour cost divided by simultaneous conversations | Nearly zero after setup |
| Risk | Waiting time at peak | A wrong answer nobody sees |
In practice the choice is rarely one or the other. What works is an AI handling the common questions and handing over on doubt or on anything involving money, with the whole conversation attached so the customer doesn't retype their question.
Where it goes wrong
Popping the window before the visitor did anything
A chat that opens by itself after three seconds with "Can I help you?" interrupts someone who's reading. Open the window on behaviour instead, for example on a second visit to the same product page or a long pause in the checkout.
Opening hours that don't match staffing
If chat is on until 5.30 and the team leaves at 5, the last conversations wait on nobody. Tie the window's visibility to who's actually online.
No transcript to the customer
What's agreed in chat disappears the moment the window closes. Send the conversation by email afterwards, otherwise in a week there's no record of the commitment you made.
Measuring chat with email numbers
A four-hour response time is fine in email and worthless in chat. Set a separate timer on first reply for chat, in seconds, and also measure how many conversations are abandoned without an answer.
Setting up chat in five steps
Pick the pages, not the whole site
Start with product pages and checkout. That's where questions have direct revenue impact and stay short.
Enter your real opening hours
Tie it to staffing, not shop hours. Outside those hours show a form with the response time you actually meet.
Give the AI the fifteen questions you get most
Take them from last month's mailbox. This is setup based on your own data, not on a template list.
Define the handover rules
Money, complaints and exceptions go to a human. Write that down before you go live, not after the first incident.
After two weeks, measure what came in
Group the conversations by topic. The biggest category is nearly always something that could have been on your site.
What to do this week
Read fifty chat conversations from the past month and note per conversation whether the answer was somewhere on your site. For those where it was, look at where the customer searched before they started typing. That's the page you change, and that change saves more work than a faster chat reply.
What this looks like in a system
In Cuego the chat widget is part of the same workspace as email and phone. A conversation shows which page the visitor has open and, once the email address is known, the orders that belong to it. The AI agent runs the conversation within the boundaries you set per topic and hands over to an agent with the full history attached. Close the conversation and you can email the transcript and start a task or return straight from it.
ChatRelated terms
Abandon rate
Abandon rate is the share of customers who drop off before reaching anyone. Those questions don't disappear, they return by email tomorrow.
Read onAI agent
An AI agent understands a question in plain language, looks up information in your systems and carries out the action behind it. That's something other than a chatbot with choice buttons.
Read onWISMO
WISMO is the question of where the order is. It's rarely a location question and nearly always a certainty question, and that difference determines what you can do about it.
Read on