thinkhard.SI

Prompt repository Customer support & success

Write what your support chatbot says when it hands a customer to a person

Your chatbot passes some conversations to human agents, and you want handoff messages that set honest expectations plus a summary the agent can act on.

Registry no.
SUP-060
Collection
Customer support & success
Laid out for
Any chat model
Score
10 out of 10
Blanks to fill
6

Customer support & successAny chat

The prompt

Score 10 of 10, Excellent

Use it when Your chatbot passes some conversations to human agents, and you want handoff messages that set honest expectations plus a summary the agent can act on.

  • {{company}}
  • {{handoff_setup}}
  • {{hours}}
  • {{wait_times}}
  • {{always_handoff}}
  • {{voice}}
Full prompt and why it works
Prompt
Write the handoff messages our support chatbot uses when it passes a customer to a human agent at {{company}}, plus the summary the agent sees. The customer should know that a person is taking over and roughly when; they should never have to repeat themselves. The agent should be able to pick up the conversation without scrolling back through it.

How handoff works in our tools (what triggers it, whether the customer waits in chat or gets an email, what the agent sees, how long a chat stays open): {{handoff_setup}}
Our support hours and time zone: {{hours}}
Typical wait times from our own data, in and out of hours: {{wait_times}}
What the bot must always hand over (for example billing disputes, complaints, safety issues, legal threats): {{always_handoff}}
Our voice: {{voice}}

Write:
1. Customer messages, each under 45 words, for these cases: during hours with a short wait; during hours with a long queue; outside hours; the customer asked for a person; the bot could not answer after two tries; a sensitive topic such as a complaint or a safety concern. Each says what happens next and when, using our wait times, and what the customer can do meanwhile, if anything.
2. A message for when no agent picks up within the expected time, with an honest update and an alternative, such as an email follow-up.
3. The agent summary template, under 80 words when filled in, with labeled lines: Customer, Contact, Issue in one sentence, What the bot already tried or told them, Details collected (order number and so on), Customer's mood if clear from their words, Why it was handed over.

Rules: never say "a member of our team will be with you shortly" unless the wait times support it, and never give a wait time that is not in our data. Do not have the bot pretend to be a person or ask again for details it already has. One apology is enough where one is due.

A good handoff sounds like one helpful colleague passing the customer to another.

If our setup does not say whether the customer waits in chat or gets an email, write both versions and mark them [PICK ONE].

Variables

{{company}}
Your company, for example "a home internet provider".
{{handoff_setup}}
How handoff works in your tools: triggers, chat or email follow-up, what agents see, how long a chat stays open.
{{hours}}
Support hours and time zone, for example "8 a.m. to 8 p.m. Eastern, Monday to Saturday".
{{wait_times}}
Typical waits from your own data, for example "about 4 minutes in hours; next morning outside hours".
{{always_handoff}}
Topics the bot must always pass to a person, for example "billing disputes, complaints, safety issues".
{{voice}}
How you sound, for example "calm, plain, first names".

Why it works

  • Length. Six named cases, from a short wait to a safety concern, plus the no-pickup message, cover the moments where handoffs usually go wrong.
  • What to avoid. No wait time beyond your data and no "shortly" the numbers do not support, so customers are never promised what the queue cannot deliver.
  • Format. Labeled lines in the agent summary let the person taking over read the case in seconds without scrolling back through the chat.

How to use it

  1. Press Fill in blanks and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
  4. When the answer comes back, check it before you use it. The Fact Check List lists every figure, quote and source in the answer worth checking, riskiest first.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/chatbot-to-human-handoff

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "chatbot-to-human-handoff"}, or just ask for it by name.