What is human handover?

In one sentence

Human handover is when an AI agent stops replying in a conversation and passes it to a person, with the context, because the customer needs judgement, care or a decision the AI should not make.

In more detail

A good handover has three parts: the agent tells the customer a person will reply and roughly when, it alerts the right person, and it stays quiet in that chat until the person is done.

Handover is triggered by rules you write: complaints, refunds, health questions, anything the agent is unsure about, or a customer simply asking for a person.

Without a clear handover, customers get stuck talking to an AI that cannot help, which is the main reason people dislike automated replies.

Handover is not a sign the AI has failed. It is part of the design. An agent that answers routine questions well and passes the rest on quickly is far more useful than one that tries to handle everything itself.

The context matters as much as the timing. A customer who has already explained their problem should never have to explain it again, so the person taking over needs a short summary: who the customer is, what they asked and what the agent has already said.

When to hand over

Customers rarely use the word "complaint". Tell your agent to look for the feeling as well as the words: frustration, repeated questions, capital letters or "this is the third time".

What the handover message should say

A handover message has three jobs: say a person is coming, say roughly when, and say what happens in the meantime. Two examples:

During opening hours:
"I'm sorry about this, and thank you for explaining. I've passed your message and photos to Sam, our manager. Sam will reply here by 4pm today."

Outside opening hours:
"Thanks for letting us know. Our team is back at 9am tomorrow, and the first person in will reply to you here. You won't need to explain it again."

Only promise a time you can keep. A realistic "by tomorrow morning" builds more trust than "in a few minutes" followed by silence.

An example handover

An online homeware shop answers Instagram messages with an AI agent. Names and details are examples.

The agent did the useful early work, gathering photos and details, then stepped aside before any decision about a replacement. The customer saw one continuous conversation and never had to repeat themselves.

Notes to give your agent

You can paste something like this into your agent's instructions and adjust it for your business:

Handover rules
- Hand the chat to a person for: complaints, refunds, damaged or missing items, health questions, anything legal, any customer who asks for a person, and anything these notes do not cover.
- Before handing over, tell the customer a person will reply, and when. Our reply times: Monday to Saturday, 9am to 6pm, within two hours. Outside those hours, say we reply from 9am.
- Never promise a refund, a replacement or compensation. Never say whose fault it was.
- When you hand over, tag the chat and write a summary of up to three lines: who, what happened, what you have already said.
- After handing over, do not reply in that chat until it is handed back.

How the right person finds out

How you are alerted depends on how you work. In Spaces, the agent can tag the chat, give it to someone on your team as its owner and mark it as handed over. Spaces can send a webhook when a chat is handed over, which a tool you already use can turn into an alert, and a routine can email you a summary of unread chats.

Whatever you choose, make sure one named person owns each handover. An alert that goes to everyone is often an alert nobody answers.

Handing the chat back

When the person has finished, the chat can be handed back to the agent so it answers that customer again next time. Many owners only hand back once the issue is fully closed, so the agent never steps into the middle of a sensitive conversation.

If you want the agent to stay out of a particular customer's chats for good, simply leave that chat handed over.

Common mistakes

• No time given. "Someone will be in touch" leaves the customer guessing and often leads to a second, angrier message.
• A time nobody can keep. Quote the reply time you actually manage on a busy day.
• The agent keeps talking after handover. Two voices in one chat confuses everyone.
• The customer has to repeat everything. Always pass on a summary.
• Handover only on keywords. Upset customers rarely say "complaint", so describe the situations, not just the words.
• Nobody owns the alert. Give every handover a named person.

A quick handover checklist

• A written list of situations that always go to a person.
• A handover message for opening hours and one for out of hours.
• A realistic reply time the agent can quote.
• One named person, or a rota, for each kind of handover.
• A short summary written by the agent every time.
• A clear rule for when the chat is handed back.

Why it matters for your business

• Customers with real problems reach a person quickly.
• The agent never makes decisions you have not allowed.
• You can let the agent handle more, knowing the hard cases still come to you.
• Your team spends its time on the conversations where a person makes the difference.

How Spaces handles it

Your AI agent decides when to hand over, using your rules. Spaces keeps the conversation on your own account, so you or a colleague can reply in the same chat from your phone or app, and the customer sees one continuous conversation.

Once a chat is handed to a person, Spaces keeps the agent out of it: no replies, scheduled messages or routines go there until it is handed back. If the agent tries to reply anyway, Spaces refuses the message and tells it that a person has taken over.

For extra caution, Spaces also offers approvals, where a reply waits until a person approves it before it is sent. Approvals are being turned on workspace by workspace.

In short

Human handover is the moment an AI agent steps aside for a person. Done well, it tells the customer who is coming and when, passes on the context and keeps the agent quiet until the person is finished.

Write your handover rules before you switch an agent on, give it the wording to use and make sure someone owns every alert. The guide on handing a chat from AI to a person walks through the setup in more detail.

What should always trigger a handover?

Complaints, refunds, health or legal questions, abuse, and any customer who asks for a person. Add anything your notes do not cover, so the agent never has to guess.

How fast should a person reply after handover?

Within the time the agent promised. Give it a realistic time to quote, with a different one for evenings and weekends if you need it.

Can the agent take the chat back afterwards?

Yes, if you tell it to. Many owners let it resume once the issue is closed.

What if nobody is available to take over?

The agent should say when the team is next available and that the customer will not need to explain again. Then the first person in picks it up from the summary.

Does the customer see that the chat was handed over?

They see the agent's message saying a person will reply, then that person's reply in the same chat. There is no other sign, because everything happens on your own account.

Can I step in without the agent handing over?

Yes. You can reply from your phone at any time. To stop the agent also replying, mark the chat as handed over, for example by asking your agent to hand it to you.

Is handover the same on every channel?

Yes. It works the same way on WhatsApp, Instagram, LinkedIn, Messenger, X and email, because it is about who replies in the chat, not which app it is in.