Chatbot vs live chat: when to use each in support

A chatbot answers customers automatically, while live chat connects them with a person on your team in real time. For most support teams the answer to chatbot vs live chat is both: an AI chatbot answers the questions your help content covers, and a teammate takes over in the same conversation when a question needs a person.

This guide covers what each handles well, how they compare side by side, the handoff rules that make them work together and what each costs.

Chatbot vs live chat: the short answer

Three different things share the chat window:

• A rule-based chatbot follows the buttons and paths you design, such as “Track my order” or “Talk to sales”. It answers only what you scripted.
• An AI chatbot reads the customer’s question and writes an answer from your help content and the other sources you connect.
• Live chat puts a person from your team in the conversation, in real time.

Use an AI chatbot for the first reply to questions your content answers, and live chat for the conversations that need judgment, access or empathy. Customers expect that path. In a Gartner survey of 3,566 customers run in February and March 2026, 87 percent said companies that use generative AI in customer service must offer a way to reach a human agent (Gartner, August 4, 2026).

What an AI chatbot handles well

An AI chatbot replies in seconds at any hour. It does best with:

• Repeat questions. How to invite a teammate, reset a password, find an invoice or connect an integration, answered from your documentation.
• Nights, weekends and holidays. The first reply does not wait for someone to log in.
• Peaks. A launch or an outage brings many conversations at once, and the AI answers them all without a queue.
• Other languages. Many AI chatbots answer in the customer’s language.
• Details before a handoff. It can collect the email address, the account and a description, so a teammate starts with context.

Its limits come from the same place. An AI chatbot is only as good as the content it answers from, and outdated or missing articles lead to wrong or vague answers. It cannot make an exception to a policy, judge an account-specific problem or calm a customer who is already upset.

A bad first experience is expensive. In another Gartner survey, only 27 percent of customers said they would try a chatbot again after a negative experience (Gartner, September 2, 2026). For more on where AI chatbots save time and where they fail, see how AI chatbots help customer support.

What live chat handles well

Live chat brings judgment, account access and empathy, which an AI chatbot lacks:

• Exceptions. Refunds, contract questions and anything where the policy needs interpretation.
• Account access. Looking at the customer’s setup, data or billing history.
• Emotion. Frustrated or worried customers, and cancellations.
• Complex problems. Bugs that need follow-up questions, screenshots and a look at the logs.
• High-value conversations. Key accounts and sales conversations where a person builds trust.

Its limit is staffing. Live chat needs people online during the hours you promise, queues grow at peak times, and the cost rises with the number of conversations your team answers.

Chatbot vs live chat side by side

How to run both: handoff rules

The handoff decides whether the two work as one. Set when the AI hands over, what it passes along and what the customer sees.

When to hand off

1. The customer asks for a person. Hand over right away. If you ask for details first, do it once, and hand over when the customer asks again.
2. The AI finds no answer in your content, or the customer says the answer did not help.
3. The topic is sensitive: billing disputes, refunds, cancellations, security, legal questions or data deletion.
4. The customer repeats the question or shows frustration.
5. The problem needs access or an investigation, such as a bug or an error on one account.

What to pass along

The full conversation, what the AI already tried, who the customer is and, for technical questions, the page, device and any screenshots. The teammate should be able to reply without asking the customer to repeat anything.

What the customer sees

A short message that a teammate is joining, and when to expect a reply. Outside your hours, say when the team is back, and let the AI keep answering what it can.

What to review

Read the week’s handoffs. Repeated handoffs on one topic usually point to a missing or unclear article, and late handoffs point to a rule that needs changing. The full playbook, with the metrics to track, is in how to design AI-to-human handoffs.

What each costs

AI chatbots and live chat are billed differently:

• AI chatbot. In September 2026, AI chatbots for customer support list at $0.05 to $2 per conversation or resolution, depending on the vendor and the unit it bills, plus the plan they run on. AI chatbot pricing compares the units and vendors, with worked totals for three team sizes.
• Live chat. Live chat tools are mostly priced per seat. LiveChat, for example, costs $19 to $79 per person a month billed annually (LiveChat pricing, checked September 27, 2026). The larger cost is people: every hour of coverage needs someone online.
• Both. The AI line grows with the conversations the AI takes, and the people line grows with the conversations it hands over. With a good setup, the AI answers what your content covers and your team spends its time on the conversations that need a person.

Compare the total for your volume, not the unit price. For live chat tools and what they cost, see the best live chat software for SaaS.

How Gleap runs AI answers and live chat together

In Gleap, live chat and Kai Support, the AI support agent, share one widget, which runs on websites and in web, iOS, Android, React Native and Flutter apps. Both are on every plan.

• Kai Support answers first. It answers from the knowledge sources you connect, cites them and labels its answers as AI. On Hobby, those are snippets, websites, files and your other connected sources; help center articles need Growth.
• A person when needed. When a customer asks for a person or taps Talk to our support team, the conversation lands in your inbox with the whole thread, and a teammate replies in the same chat under their own name. If a customer asks for a person again after Kai Support asked for details, it hands over.
• Rules you set. In the Let Kai answer step of your workflow, you can forward the conversation to your team when Kai Support finds no answer, describe in your own words when Kai Support should pass a conversation on, choose after how many answers the Talk to our support team button appears and send your reply times with the handoff.
• Office hours. The widget shows your reply time during office hours and when your team is back outside them. Kai Support keeps answering around the clock.
• More channels on Growth. Hobby covers live chat. On Growth and up, email, WhatsApp, Instagram, Messenger, Slack and Telegram land in the same inbox with the same Kai Support.
• Technical questions on Growth. On Growth and up, a conversation Kai Support cannot answer in the widget goes to a board. Where your team turned on Kai investigates here, Kai Support investigates it with the logs and your code, and what it cannot settle reaches your team with its diagnosis.

Commonplace, a Gleap customer, reports that Kai Support handles 50 to 60 percent of its inbound support. That is its result, not a benchmark.

Hobby costs $29 per month billed annually ($35 monthly) with one seat, and Growth costs $99 per month billed annually ($119 monthly) with 4 seats included, then $25 per extra seat per month. Kai Support costs about $0.05 per conversation (estimate); the other agents are paid from the same AI balance. Every plan starts with a 7-day free trial, and no credit card is needed. The live chat page shows the widget and the handoff in detail.

Gleap publishes this guide.