Guide · Bots & AI
Chatbot to human handover: getting it right
The moment a chatbot hands a conversation to a person decides how customers remember the whole interaction. A good handover feels seamless: the person already knows the question and the details, and replies quickly. A bad one makes the customer repeat everything, wait without information or end up back at the bot's menu. This guide explains how to design handovers that work.
[ 01 ]
In short
In short
- 01
Define clear triggers: the customer asks, the bot fails repeatedly, the topic is sensitive or the opportunity is valuable.
- 02
Pass the full conversation and collected data to the person, so the customer never repeats themselves.
- 03
Tell the customer what happens next and how long it will take, honestly.
- 04
Outside working hours, collect details and promise a reply at a specific time.
- 05
Measure time to human reply and outcomes of handed-over conversations.
[ 02 ]
Guide
When to hand over
- The customer asks for a person — in any words, at any step.
- Repeated failure: the bot did not understand twice, or the customer is going in circles.
- Sensitive topics: complaints, refunds, health, legal or financial issues.
- High-value opportunities: large orders, corporate clients, custom requests.
- Emotion: frustration or anger detected in the conversation.
- Out of scope: questions the bot is not designed to answer.
Do not make reaching a person difficult. Customers who feel trapped leave or complain publicly.
Transfer the context
The person taking over should see:
- the full conversation history;
- data the bot collected: name, contact, order number, selected service;
- the reason for handover;
- the customer's language;
- previous interactions from the CRM, if available.
This is the main difference between a helpful handover and a frustrating one. See CRM integration.
Set expectations
When handing over, the bot says what happens next: "A manager will reply here within 10 minutes" during working hours, or "Our team works from 9:00 to 18:00; we will reply tomorrow morning" outside them. Only promise times you can keep.
Routing
- By topic: sales, support, billing.
- By language: Azerbaijani, Russian, English speakers.
- By customer type: new or existing, corporate or individual.
- By workload: to the least busy available agent.
Routing rules can be simple at first and refined based on data.
During the handover
- Agents are notified immediately — in the inbox and in a messenger they watch.
- Unanswered conversations escalate after a set time.
- The bot stays silent while a person is handling the conversation, so it does not interrupt.
Returning to the bot
After the person resolves the issue, the conversation can return to the bot for follow-up tasks — for example, sending a booking confirmation or collecting feedback. The transition should be explicit.
Outside working hours
The bot can still help with routine tasks, collect details for the team, and set a clear expectation for when a person will reply. Morning shifts should start with the overnight queue.
Measure
- Share of conversations handed over, by reason.
- Time from handover to first human reply.
- Resolution rate and customer satisfaction after handover.
- Repeated handovers for the same topic — candidates for improving the bot or the knowledge base.
[ 03 ]
FAQ
Frequently asked questions
Should the bot hand over on every "operator" request?
Yes. Customers who ask for a person should get one, possibly after the bot collects a couple of details that will help the person reply.
What if no one is available?
Say so honestly, collect contact details and the question, and give a specific time for the reply. Then meet it.
Can AI help human agents after handover?
Yes. Suggested replies, summaries of long conversations and quick access to knowledge base articles speed up agents, with the agent in control of what is sent.