Handing a chat from bot to a person without losing the customer

Short answer: A good handover has five parts: clear triggers, an honest message to the customer, routing to the right person with the full conversation attached, a plan for nights and weekends, and a few numbers you review every month. On WhatsApp, timing matters more than on your website, because once 24 hours pass after the customer's last message you can only reply with a pre-approved template. Design the handover first and the bot's clever answers second.

Most chatbot complaints are not about wrong answers. They are about being stuck: typing "agent" three times, getting the same menu back, or being told someone will reply and then hearing nothing. If your business in Dubai, Abu Dhabi or elsewhere in the Gulf runs a bot on its website or WhatsApp number, the handover is where you keep or lose that customer.

Step 1: Decide the triggers

Write down, in plain language, every situation where a person should take over. Helpdesk tools turn these into rules, but the list is a business decision.

The customer asks. "Human", "agent", "call me", or the Arabic equivalents your customers actually type. Never make someone ask twice. Intercom's documentation for its AI agent lists a clear request to speak to a human as one of the default escalation cases, alongside strong frustration and a customer stuck in a repetitive loop.

Repeated failure. The bot has failed to answer or misunderstood twice in a row. Two is usually enough; by the fourth attempt the customer has gone.

Negative mood. Many tools detect sentiment. Intercom's escalation rules article gives examples of rules based on a "Sentiment" attribute being negative, an issue type, an order total above a set value, or a custom flag such as a VIP customer.

High-value or sensitive topics. Large orders, complaints about a completed job, refunds, damaged deliveries, anything involving a customer's health or safety. These should go to a person even if the bot could technically answer.

Known customers. Repeat clients or account holders can skip the bot entirely if your CRM recognises their number or login.

Keep a separate list of what the bot must never do, such as promising refunds or confirming bookings it cannot see.

Step 2: Write the handover message

The customer should know three things the moment the bot steps aside: a person is taking over, roughly when they will reply, and that they do not need to repeat themselves.

Compare "Transferring you now" with "I've passed this to our bookings team with everything you've told me. Someone usually replies within 15 minutes during opening hours, here in this chat." Only promise times you can actually keep. Intercom notes its bot posts a brief handoff message by default before routing; most tools let you rewrite it, so make it sound like your business, in the customer's language.

Step 3: Route it with the context attached

A handover that arrives as a blank new ticket undoes the bot's work. Check that your setup passes:

the full transcript, not just the last message
what the bot already collected (name, order number, location, preferred language)
the reason for the handover (customer asked, frustration, VIP, topic)
the channel, so the reply goes back on WhatsApp rather than by email

Then route by skill. In the Gulf that often means language first: an Arabic-speaking customer should reach someone who can reply in Arabic. After that, route by topic (sales, bookings, support) and by customer value.

Know how your tool hands the conversation back, too. Zendesk's handoff and handback article explains that once a conversation is handed to a live agent, the AI agent can no longer respond in it, and the agent remains first responder until the related ticket is closed. By default a solved ticket closes after four days, which can leave a returning customer talking to a human queue instead of the bot; Zendesk describes changing that automation or adding a trigger.

Step 4: Plan nights, weekends and the WhatsApp clock

Out of hours, the honest option beats the hopeful one. The bot should say the team is offline, give the next opening time, collect the details a person will need, and confirm the customer will be contacted then.

On WhatsApp there is a hard technical limit to plan around. Meta's service messages documentation explains that each message or call from a customer opens a 24-hour customer service window, reset by every new message. While it is open you can send free-form replies; once it closes, you can only send pre-approved template messages.

In practice:

A handover at 23:00 that nobody picks up until the afternoon of the following day may fall outside the window. Have an approved template ready, such as "Hello, this is [business] following up on your enquiry about [topic]. Reply here to continue."
Ask the bot to collect everything before the team logs off, so the morning reply can be complete in one message.
Watch the window in your inbox tool; many show a countdown or flag conversations close to expiry.

Meta's pricing for business messages has changed several times in 2025 and 2026, including how replies inside the window are categorised and charged, so check its current pricing page rather than relying on older guides.

Step 5: Measure the handovers

Review these monthly:

Handover rate and the reasons behind it. A rising rate on one topic usually means the bot's knowledge is missing something.
Time to first human reply after a handover, split by opening hours and out of hours.
Repeat questions: how often agents ask for details the bot already collected.
Customers who leave between handover and first reply.
Conversations the bot closed that the customer reopened soon after.

If you use a tool that bills per outcome, read how handovers are counted. Intercom's outcomes article explains that escalations triggered by default behaviour or workspace-level rules are not billed, while handoffs that follow a configured procedure can be.

FAQ

Should the bot always offer a "talk to a person" button?

For most small businesses, yes. Hiding it rarely saves real work; it moves the frustration into reviews instead. A visible option that the bot offers after a failed answer is a reasonable middle ground.

Can the same person handle website chat and WhatsApp?

Usually, if both channels feed one shared inbox. Keep the reply on the channel the customer used, and make sure the inbox shows which one that was.

What if nobody is free during opening hours?

Treat it like out of hours: tell the customer honestly, collect details, and give a realistic time. A queue position or estimated wait is better than silence.

How do I stop agents and the bot replying at the same time?

Pick a tool where handover clearly switches the bot off for that conversation, as described in the Zendesk article above, and test it before launch.

If your chatbot keeps customers stuck where a person should step in, we can help you map the handover.