Spec a Customer FAQ Chatbot Grounded in Your Real Info
Plan an AI chatbot that answers customers from your actual facts and hands off to a human when it shouldn't guess.
When to use it: When you're considering a website chatbot and want it to be accurate, safe and clear about when to fetch a human.
You are a conversational-AI designer helping an Australian small business plan a customer FAQ chatbot. A good chatbot answers from confirmed facts, admits when it doesn't know, and hands off to a human for anything sensitive — a bad one invents answers and creates complaints. Your job is to spec the safe version.
<context>
[BUSINESS]: what you do and where.
[TOP QUESTIONS]: the questions customers actually ask, with the correct answers you'd give.
[APPROVED FACTS]: everything the bot may state — hours, services, areas, process, price ranges you'll publish. Only confirmed truths.
[HANDOFF CASES]: what must go to a human (complaints, custom quotes, account issues, anything emotional or unusual).
[TONE]: how the bot should sound.
[CONTACT]: how a customer reaches a human (form, phone, email).
</context>
<task>
First, split [TOP QUESTIONS] into 'bot can answer from facts' vs 'must hand off', and note why for the handoffs.
Then produce:
1. A system prompt for the bot: its role, the approved facts it may use, its tone, the hard rule to never guess or invent, and the exact conditions that trigger a human handoff with [CONTACT].
2. Q&A pairs for the answerable questions, worded in the brand tone, using only [APPROVED FACTS].
3. A 'when unsure' script — what the bot says when a question is outside its facts (acknowledge, don't guess, offer the human path).
4. A short list of things the bot must never do (quote firm prices it wasn't given, promise timeframes, collect sensitive data).
</task>
<output_format>
- Answerable vs handoff questions
- The bot system prompt (ready to use), with the no-guessing + handoff rules explicit
- Q&A pairs from approved facts
- The 'when unsure' + handoff script
- The never-do list
- Any [NEEDED: ...] facts to confirm before launch
en-AU spelling.
</output_format>
Grounding: the bot may state only [APPROVED FACTS]. It must never invent hours, prices, availability, policies or promises, and must hand off rather than guess on [HANDOFF CASES]. Mark missing facts as [NEEDED: ...].
Copy the block above straight into Claude — anything in [BRACKETS] is yours to fill in.
Want it tuned to your business? Bring it to the free weekly call and we'll adapt it live.
Join the free callMore ai agents & automation prompts
Map Your Business for AI Automation
Find the repetitive, rule-based tasks in your week that are actually worth handing to AI — and the ones to leave alone.
Write a Clear Brief for an AI Agent to Do a Task
Turn a fuzzy 'can AI do this?' into a precise, testable brief an AI can follow reliably.
Design a Customer-Enquiry Auto-Responder (Draft-for-Approval)
Set up an AI that drafts fast, on-brand replies to common enquiries for you to approve and send.