Chatbot-to-Human Handoff: When and How to Escalate to a Live Agent
The five triggers for escalating a chatbot conversation to a person, what context to pass along, how to handle after-hours handoffs, and example messages you can adapt.

في هذه المقالة
What happens in your chat when the bot is out of its depth? If the honest answer is "it loops" or "it tells the customer to email support," your chatbot is creating work rather than saving it.
A good chatbot human handoff happens at the right moment, for the right reasons, and carries the full conversation with it. The customer is told what happens next, and the agent who picks it up can reply without asking a single question the bot already asked.
Here's how to decide when to escalate, what to pass along, and what to say.
Key takeaways
- Escalate on five triggers: low confidence, sensitive topics, an explicit request, repeated failure, and high-value customers.
- Pass the transcript, intent, tags and collected details so the customer never has to repeat themselves.
- After hours, set an honest expectation and gather what the agent will need, instead of leaving the customer in silence.
5 triggers to escalate a chatbot to a live agent
| Trigger | Example | What the bot should do |
|---|---|---|
| Low confidence | The question isn't covered by the knowledge base | Say so plainly and offer a person instead of guessing |
| Sensitive topic | A reaction to a product, a legal threat, a data deletion request | Hand off right away without trying to resolve it |
| Explicit request | "Can I talk to a human?" | Hand off, every time, without arguing |
| Repeated failure | The customer rephrases the same question twice, or says "that's not what I asked" | Stop trying and escalate |
| High-value customer | A large B2B quote, a VIP account, an order above a threshold you set | Route to a person even if the bot could answer |
Two of these deserve extra attention.
Low confidence is your main defense against wrong answers. An assistant that's allowed to say "I'm not sure, let me get a colleague" makes far fewer confident mistakes. We go deeper in how to prevent AI chatbot hallucinations.
Explicit requests should never be blocked. Some bots reply "Are you sure? I can help with..." three times before giving in. Customers notice, and they remember.
What context to pass in an AI to human handover
The agent receiving the conversation should understand the situation within seconds. That means the handoff includes:
- The full transcript, not just a summary
- Intent: what the customer is trying to do, such as "return a damaged item"
- Tags: topic, channel, language, urgency
- Collected details: name, email or phone number, order number, product, anything the bot already asked for
- Why it escalated: which trigger fired and what answer, if any, the bot already gave
If your chatbot and your agents work in different tools, this is where context gets lost. A shared inbox that holds conversations from every channel solves most of it; our post on omnichannel vs. multichannel support explains why that continuity matters.
Never make the customer repeat themselves
The biggest complaint about bot-to-human transfers is having to start over. Compare these two agent openers, after the bot has already collected an order number and a description of a cracked mug:
| Weak opener | Strong opener |
|---|---|
| "Hi, how can I help you today?" | "Hi [name], I'm [agent] from the support team. I can see order 10482 arrived with a cracked mug, and I'm sorry about that. Could you send a photo so I can arrange a replacement?" |
The strong version proves the agent read the conversation, apologizes once and moves straight to the next step. Train agents to read the last few messages before typing anything.
After-hours handoff: set expectations honestly
A handoff at 10 p.m. on a Saturday is still a handoff, but nobody is there to pick it up. Don't pretend otherwise:
- Tell the customer when a person will reply, using your real hours: "Our team is back Monday at 9:00 CET."
- Get a way to follow up if you don't have one: email, phone number, or continuing on the same channel.
- Gather what the agent will need, such as an order number or photos, so Monday's reply solves the issue instead of asking questions.
- Keep answering simple follow-ups. If the customer then asks about delivery times, the bot can still help.
- Work the overnight queue first in the morning, most urgent and oldest first.
On WhatsApp and Instagram, remember the 24-hour customer service window. If a customer's last message was Saturday night, a Monday-morning reply may fall outside it, so plan weekend coverage or an alternative follow-up channel.
Example handoff messages
Wording matters more than most teams expect. A few you can adapt:
- Low confidence: "I don't have a reliable answer to that, and I'd rather not guess. I'm passing this to our team, and someone will reply here shortly."
- Explicit request: "Of course. I'm connecting you with a colleague now. They'll see everything we've discussed."
- Sensitive topic: "Thank you for telling us. This needs a person, so I've passed it to our team as a priority."
- After hours: "Our team is offline right now and back at 9:00 tomorrow. I've saved your details and order number, so you won't need to explain again."
Avoid "please hold" when you can't promise a time, and never say "right away" if nobody is online.
How chatbot human handoff works in Conviro
In Conviro, you set handoff conditions and confidence thresholds on the AI assistant itself. When confidence is low, the topic is sensitive or the customer asks for a person, the conversation moves to the Live Inbox with the transcript, intent, tags and collected details attached. Website chat, WhatsApp and Instagram DM conversations share the same inbox, so agents work from one view whatever the channel. Human handoff is always possible: the assistant is there to support your team, not to wall customers off from it.
Frequently Asked Questions
When should a chatbot transfer to a human?
When the bot isn't confident in its answer, the topic is sensitive, the customer asks for a person, the same question fails twice, or the customer is high value. Write these rules down before launch, then read real conversations during the first weeks and tune them.
What is a warm handoff in customer service?
A warm handoff passes the customer to a person together with full context: the transcript, what they want and the details already collected. The agent continues the conversation instead of restarting it. A cold handoff only transfers the customer, who then has to explain everything again.
Can customers always ask to speak to a human?
They should be able to. Blocking or delaying an explicit request for a person damages trust quickly, even when the bot could technically answer. If nobody is available, say when someone will reply and collect the details your team needs to follow up properly.
How do I know if my handoffs are working?
Track how often conversations escalate, how long escalated customers wait for a person, and satisfaction for escalated versus fully automated chats. Read a sample of transcripts weekly. Our guide to measuring CSAT in chat shows how to separate those scores.
Next steps
List your five escalation triggers, write the handoff message for each, and agree on who picks up after-hours conversations. Then test the full path from bot to agent before launch. Start free with Conviro and try handoff with your own inbox.
نشرت من قبل
كونفيرو
مشاركة الرؤى حول دعم العملاء المدعوم بالذكاء الاصطناعي وأتمتة روبوتات الدردشة واستراتيجيات نمو SaaS.
صمم عملية التسليم، وليس الروبوت فقط
قم بإحضار محادثات الذكاء الاصطناعي والمحادثات البشرية في صندوق بريد وارد واحد
قم بتوجيه المشكلات المعقدة إلى فريقك مع السياق بينما تتعامل الأتمتة مع الأسئلة المتكررة.
مقالات ذات صلة
المزيد من المشاركات

Customer Support
Self-Service Customer Support: How to Let Customers Help Themselves
6 دقيقة قراءة
اقرأ المقال
Customer Support
Unified Inbox for WhatsApp, Instagram, Web Chat, and Email: How to Choose One
7 دقيقة قراءة
اقرأ المقال