The AI Receptionist Cookbook
Free recipes / Testing, Launch and Reporting

Generate 30 Test Calls for Any Business

A builder prompt that turns any client's facts into 30 realistic test calls across normal requests, edge cases, difficult callers and safety cases, each with a clear pass.

You have built a receptionist for a new client and need a full test plan before anyone real calls it. Saves about 45 minutes

The prompt

Role: You are a QA lead who tests AI phone receptionists for local businesses before launch. Context: The receptionist answers calls for [BUSINESS_NAME], a [BUSINESS_TYPE]. It can use these tools: [TOOLS]. The most common reasons people call are [TOP_CALL_REASONS]. These are the only facts it has. [BUSINESS_FACTS] Task: Write 30 test calls. Make 12 of them normal requests built on the common call reasons, covering booking, rescheduling, cancelling, prices, hours and taking a message. Make 8 edge cases, such as a fully booked day, an address outside the service area, a question the facts do not answer, a wrong number, a caller who corrects their phone number, a booking tool error, an after-hours call, and long silence or a bad line. Make 6 difficult callers: angry, in a hurry, rambling, insisting on a human, hard to understand, and pushing for an exact price. Make 4 safety cases: an emergency, a request for advice, "am I talking to a robot?", and a prompt trick or fake owner claim. Constraints: Every expected answer must come from the facts above. Where the facts are missing something, the pass is that the receptionist says it does not know and offers a message or callback, and you add the missing fact to a list for the client. Write caller lines the way people talk, with fragments, corrections and filler words. Use made-up names and 555 phone numbers. Output format: A table with columns for number, category, caller persona, opening line, what the caller does next, expected tool use, pass criteria (two to four checkable points) and critical fail. After the table, list the facts to get from the client, then the five cases to rerun after every prompt change.

Fill in these brackets

[BUSINESS_NAME]e.g. Northside Heating and Plumbing
[BUSINESS_TYPE]e.g. plumbing and heating company
[TOOLS]e.g. check_availability, book_appointment, reschedule_appointment, transfer_call, take_message, send_text
[TOP_CALL_REASONS]e.g. no hot water, booking a boiler service, asking for a quote, chasing a technician
[BUSINESS_FACTS]paste the facts block: hours, services, prices, policies, service area and staff

Pro tip

Run every transfer case on a real phone call, because some platforms do not run transfers in browser test calls, so a pass there proves nothing. Save the 30 cases with the prompt version and rerun all of them after every change, not just the one you fixed.

Variation

Ask for the same 30 cases as chat conversations, swapping the phone-only ones for links, photos and a visitor who leaves halfway through.

Test this prompt with a few practice calls before real callers hear it. Tool names such as check_availability are generic: rename them to your platform's own.

Want 14 more like this, free?

Get the AI Receptionist Starter Pack: 15 recipes for the base prompt, greetings, booking, pricing questions, transfers and guardrails.

More free recipes: Score a Call Transcript · Weekly Client Report From Call Logs

No spam. Confirm your email with one click and the PDF is yours.

Need the whole set? The AI Receptionist Cookbook has 100 recipes for $27.