Testing an AI Receptionist for Locksmith Services — 20 Trial Calls to Place Before You Commit (2026)
A free trial is only useful if you test the calls that actually cost you money. Here are twenty test calls a locksmith owner should place during any AI receptionist trial — the 2 AM lockout, the all-keys-lost push-to-start, the Spanish caller, the price shopper, the angry repeat caller — plus how to score them and review the recordings.

Testing an AI Receptionist for Locksmith Services — 20 Trial Calls to Place Before You Commit (2026)
Most locksmith owners test an AI receptionist the same way. They call once, from the office, in a quiet room, and say something like I am locked out of my car. The receptionist answers politely, asks a couple of questions, and the owner concludes it works. Then the first real Saturday night arrives, with a caller on a highway shoulder, a Spanish speaker, a push-to-start all-keys-lost job, and someone yelling that the technician never showed up, and the owner finds out what the software actually does.
As of September 2026, free trials are standard across AI receptionist products, and the trial period is the only moment where evaluating one costs nothing except your time. That makes the trial the most important part of the buying decision. The problem is not access to a trial. It is that most owners test the easy call and skip the hard ones.
This guide gives locksmith owners a structured test plan: twenty specific test calls covering the scenarios that make or lose money, what a pass and a fail sound like for each, a scoring sheet, and how to review the recordings and transcripts afterward. It applies to any AI receptionist for locksmith services you are evaluating, including ours.
Before you place a single test call
Testing is only meaningful if the receptionist has been set up the way you would actually run it. A few preparation steps:
- Load your real information. Business name, service area, hours, service types, and, if the product quotes, your real pricing. A receptionist tested with placeholder data will give placeholder answers.
- Set your escalation rules. Decide what should transfer to a person, what should create a callback, and what should just be a notification. Then test that those rules fire.
- Recruit other voices. Your own voice, in a quiet office, is the least realistic caller you will ever test. Ask a spouse, a friend, a technician, and ideally a Spanish speaker to call.
- Call from different phones. A cell phone, a different carrier, a phone with a weak signal, and speakerphone.
- Test at different times. Business hours, evening, and the middle of the night, since some systems behave differently by schedule.
- Keep the scoring sheet open. Score immediately after each call while it is fresh.
A realistic first week of setup is covered in detail in the guide to setting up an AI receptionist in your first week. The test plan below assumes that setup is done.
The twenty test calls
Each call below lists the scenario, what to say, and what a pass looks like. Run all twenty if you can. If you only have time for ten, prioritize calls 1, 2, 4, 5, 7, 9, 12, 14, 16, and 19.
Emergency and after-hours
1. Car lockout at 2 AM. Call in the middle of the night. Say you are locked out of your car in a parking lot and your phone battery is low. Pass: answers promptly, asks your location and whether you are safe, collects year, make, and model, gives your after-hours price or a clear price structure from your settings, and books or escalates according to your rules. Fail: long hold, generic voicemail-style greeting, no location question, or no clear next step.
2. All keys lost, push-to-start vehicle. Say you lost the only smart key for a push-button start SUV. Pass: recognizes this as an all-keys-lost smart key job, collects year, make, model, and ideally VIN, mentions proof of ownership, and quotes or escalates correctly from your pricing. Fail: treats it as a simple duplicate, quotes a basic key price, or promises something your shop cannot do for that vehicle.
3. Stranded on a highway shoulder. Call with traffic noise in the background and say your key broke off in the ignition on the highway. Pass: prioritizes safety, suggests contacting emergency services if in danger, captures exact location including direction of travel and nearest exit, and escalates urgently. Fail: proceeds straight to a sales script with no safety question.
4. House lockout with a child inside. Say your toddler is locked inside the house or car. Pass: tells you to call 911 immediately if a child is in danger, and captures details for escalation. Fail: tries to book a normal appointment.
Language and communication
5. A Spanish-speaking caller. Have a Spanish speaker call and speak only Spanish from the first word. Pass: responds naturally in Spanish, collects the same fields as an English call, and sends notes or confirmations the team can use. Fail: responds in English, asks the caller to hold for someone who speaks Spanish with no one available, or loses information in translation.
6. A caller who switches languages mid-call. Start in English, then switch to Spanish partway through. Pass: follows the switch smoothly without restarting the intake. Fail: gets confused or repeats questions.
7. A noisy roadside caller. Call from outside near traffic or wind. Pass: asks for clarification when it cannot hear, confirms critical details back, and does not guess at addresses. Fail: records a garbled address confidently or loops.
8. A caller who interrupts constantly. Interrupt every question with a new question or a side comment. Pass: handles interruptions gracefully, answers what it can, and steers back to the required fields. Fail: talks over the caller, loses track, or skips required information.
9. Address given in pieces. Give the street number first, then later say the street name, then the city, then mention it is actually the apartment complex behind the gas station. Pass: assembles the address correctly and reads it back for confirmation. Fail: records only part of it or never confirms.
Sales and pricing
10. The price shopper. Call and say you just want a price for a key for a common sedan, nothing else. Pass: asks the minimum needed to quote accurately — year, make, model, key type — then gives the price from your settings and asks for the booking. Fail: refuses to give any price when you have loaded pricing, invents a number, or never asks for the job.
11. The caller who does not know the key type. Say you need a spare key but have no idea what kind of key you have. Pass: asks simple identifying questions such as whether the car has push-button start or buttons on the key, and uses the vehicle to narrow it down. Fail: gives up or quotes without identifying the key. How to handle this call type well is covered in identifying key type from the vehicle on the phone.
12. The price objection. After receiving a quote, say that is too expensive and another company quoted less. Pass: stays professional, explains what is included, does not negotiate beyond your rules or invent a discount. Fail: argues, caves to a price you did not authorize, or ends the call abruptly.
13. A residential or commercial request. Ask about rekeying a house or a business lock change. Pass: routes correctly according to the services you offer, either handling it or explaining clearly that you focus on automotive work. How to handle mixed service types is covered in the guide to commercial, residential, and automotive routing.
Difficult callers
14. The angry repeat caller. Call as a customer whose technician did not arrive yesterday, and be frustrated. Pass: acknowledges the problem, does not argue, captures the details, and escalates to a person or queues a priority callback per your rules. Fail: tries to sell a new appointment, gets defensive, or loses the complaint.
15. The caller who asks for a human. Say at the start that you want to speak to a real person. Pass: follows your rules — transfers if someone is available, or explains clearly and queues a callback with your details. Fail: ignores the request or traps you in a loop.
16. The out-of-area caller. Give an address well outside your service area. Pass: recognizes it, tells the caller politely that it is outside the area, and does not book a job you will not drive to. Fail: books it anyway.
17. The wrong number. Call asking for a pizza order or a doctor's office. Pass: politely identifies the mistake and ends the call without creating a fake lead. Fail: tries to run a locksmith intake.
18. The suspicious request. Ask for a car key to be made but say the car is your ex's and you do not have the title. Pass: explains that proof of ownership is required and does not book a job without it. Fail: books it anyway.
Operations
19. The existing appointment question. Call about an appointment you already booked and ask when the technician will arrive. Pass: identifies you, looks up or acknowledges the appointment according to what the product supports, and gives a status or escalates. Fail: does not recognize you and tries to book a duplicate.
20. The after-call follow-through. After any booked test call, check what happens next. Pass: confirmation text arrives in the right language, the call shows up in your system with a recording, transcript, and summary, and the right person is notified. Fail: no confirmation, missing record, or no notification.
What pass and fail really mean
It is tempting to score calls on how natural the voice sounds. Voice quality matters, but it is rarely what loses money. The four things that matter most are:
- Did it capture the job correctly? Name, phone, vehicle, key type, service, and exact location.
- Did it do the right next thing? Book, quote, escalate, transfer, decline, or tell the caller to call 911.
- Did it avoid saying something wrong? Invented prices, false promises, or services you do not offer.
- Did the information reach your team in usable form?
A receptionist that sounds slightly robotic but captures every field and escalates correctly is worth more than a charming one that books out-of-area jobs.
The scoring sheet
Score each call from zero to two in each column: zero for a fail, one for partial, two for a clear pass.
| Test call | Captured job correctly | Right next action | Avoided wrong statements | Reached team usably |
|---|---|---|---|---|
| 1. 2 AM car lockout | ||||
| 2. All keys lost, push-to-start | ||||
| 3. Highway shoulder | ||||
| 4. Child locked inside | ||||
| 5. Spanish caller | ||||
| 6. Language switch | ||||
| 7. Noisy roadside | ||||
| 8. Constant interruptions | ||||
| 9. Address in pieces | ||||
| 10. Price shopper | ||||
| 11. Unknown key type | ||||
| 12. Price objection | ||||
| 13. Residential or commercial | ||||
| 14. Angry repeat caller | ||||
| 15. Asks for a human | ||||
| 16. Out of area | ||||
| 17. Wrong number | ||||
| 18. Suspicious request | ||||
| 19. Existing appointment | ||||
| 20. After-call follow-through |
Twenty calls across four columns gives a maximum score of 160. The total is less important than where the zeros land. A zero on calls 3, 4, 14, 16, or 18 is a more serious problem than several ones scattered across the price calls, because those calls carry safety, reputation, or liability risk.
How to review recordings and transcripts
The call itself tells you half the story. The record tells you the rest. After each test call, open the recording, transcript, and summary, and check:
- Does the transcript match what was said? Pay special attention to numbers, addresses, vehicle years, and names.
- Does the summary reflect the real outcome? A summary that says booked when the call was escalated, or omits a complaint, will mislead your team.
- Were the right fields filled in? Vehicle, key type, location, service, and price.
- Did escalation rules trigger? Check whether the transfer, callback, or notification actually happened, and whether it reached the right person.
- Where did the call lose time? Long pauses, repeated questions, and loops are easier to spot in a transcript than live.
- Did lead-source tracking record where the call came from, if you use tracking numbers.
Keep a list of every problem you find and fix it in settings before retesting. Many issues in a trial are configuration issues rather than product limitations, and a good trial includes a second round of the same calls after changes.
Answer rate matters before any of this
All twenty calls assume the receptionist answers. It is worth measuring that separately. Place several calls in quick succession, or have a few people call at once, and confirm every call is picked up without busy signals or voicemail. A receptionist that performs brilliantly but misses simultaneous calls on a busy morning has not solved the problem you bought it for. The connection between a high call answer rate and lead-generation ad performance is covered in the guide to call answer rate.
Common trial mistakes
- Testing only once. One call proves almost nothing.
- Testing only in English. If your market has Spanish-speaking callers, a Spanish test is essential.
- Not loading real pricing. You cannot evaluate quoting with fake numbers.
- Not checking after-call follow-through. A booked call with no confirmation text or team notification is a broken process.
- Letting the trial expire. Schedule the tests in the first few days so there is time to adjust and retest.
- Comparing products with different tests. If you are evaluating several options, run the same twenty calls on each. A comparison of options is on the alternatives page.
Hearing TheKeyBot before a trial
If you want to hear how an AI receptionist for locksmiths handles calls before you start any trial, there are two quick options. The instant locksmith demo has the AI call your phone in about thirty seconds, answering as a locksmith company. You can also call the live demo line at +1 (716) 350-6391, which answers as a demo locksmith shop, and run a few of the calls above yourself — the price shopper, the unknown key type, or the Spanish caller are good starts.
TheKeyBot's full platform quotes jobs by year, make, and model from your shop's pricing database, books appointments, handles GPS dispatch, sends deposit links and SMS confirmations in the caller's language, records and transcribes calls with summaries and lead-source tracking, and applies escalation rules for transfers, callbacks, and shop notifications. The locksmith product overview is on the AI receptionist for locksmiths page.
A hypothetical trial week
This is an illustrative example, not a real shop. A two-van automotive locksmith starts a fourteen-day trial on a Monday. On Tuesday, the owner loads real pricing and escalation rules. On Wednesday evening, the owner, a technician, the owner's bilingual cousin, and a friend place all twenty calls across the evening and a late-night slot.
The scoring shows strong results on lockouts, price shopping, and Spanish calls, but a zero on the out-of-area call because the service radius was entered incorrectly, and a one on the angry repeat caller because the escalation notification went to an old email address. The owner fixes both settings on Thursday and reruns those calls on Friday, which now pass. The following week, real calls confirm the patterns, and the owner makes the buying decision with a week of trial left.
Nothing in that example depends on luck. It depends on testing the hard calls early enough to fix what is fixable.
What this costs
The full platform for locksmiths comes in three plans: Core at $500 a month with 500 AI minutes and 45 cents a minute after, Pro at $750 a month with 1,000 minutes and 40 cents a minute after, and Elite at $1,200 a month with 2,500 minutes and 35 cents a minute after. Every Core, Pro, and Elite plan includes a 14-day free trial, which is enough time to run all twenty calls twice, and every plan includes unlimited users with no per-seat fees. Plan details are on the pricing page.
If you only need message-taking to start, KeyBot Lite is $149 a month for a 24/7 bilingual AI message-taking receptionist with 100 calls included, 50 cents a minute after, and the first 5 answered calls free on a 7-day trial. Lite takes messages and delivers them by Telegram; it does not quote, book, or dispatch, so tests 10, 12, and 19 apply differently to it.
The bottom line
Evaluating an AI receptionist for locksmith services by placing one easy test call is how owners end up surprised on their first busy Saturday night. A useful trial runs twenty specific calls: emergencies and after-hours lockouts, all-keys-lost smart keys, highway and child-safety situations, Spanish callers and language switches, noisy and interrupting callers, addresses given in pieces, price shoppers and objections, unknown key types, angry repeat customers, requests for a human, out-of-area and wrong-number calls, suspicious ownership stories, existing-appointment questions, and after-call follow-through. Score each on capture, next action, avoided mistakes, and team handoff. Review recordings, transcripts, and summaries after every call, fix what is configuration, and retest before the trial ends. Do that and you will know what the receptionist does on the calls that cost you money, not just the calls that are easy to answer.
Frequently asked questions
How should a locksmith test an AI receptionist during a free trial?
A locksmith should test an AI receptionist by placing a structured set of realistic calls, not a single easy one. Load real pricing and escalation rules first, then have several people call from different phones and times, covering lockouts, all-keys-lost smart keys, Spanish callers, price shoppers, angry repeat customers, out-of-area addresses, and requests for a human, and score each call immediately.
What are the most important AI receptionist test calls for locksmiths?
The most important test calls are the 2 AM lockout, the all-keys-lost push-to-start job, a Spanish-speaking caller, a noisy roadside caller, a price shopper, an angry repeat caller, an out-of-area address, an address given in pieces, and after-call follow-through. Safety calls, such as a child locked inside, and suspicious ownership requests should also pass because they carry the most risk.
What does a failed AI receptionist test call look like?
A failed AI receptionist test call is one where the job was captured incorrectly, the wrong next action happened, or the receptionist said something false. Examples include booking an out-of-area job, quoting a basic key price for an all-keys-lost smart key, ignoring a request for a human, responding in English to a Spanish caller, or sending no confirmation and no team notification after a booking.
What should I check in the recordings and transcripts?
In recordings and transcripts, check that numbers, addresses, vehicle years, and names were captured accurately, that the summary matches the real outcome, and that required fields were filled in. Also confirm escalation rules actually fired and reached the right person, look for pauses and repeated questions, and verify lead-source tracking if you use tracking numbers.
How much does TheKeyBot cost for a locksmith shop?
TheKeyBot's full platform for locksmiths costs $500 per month for Core with 500 AI minutes, $750 per month for Pro with 1,000 minutes, and $1,200 per month for Elite with 2,500 minutes, each with a 14-day free trial and no per-seat fees. KeyBot Lite, a message-taking receptionist, is $149 per month with 100 calls included. Details are at https://www.thekeybot.com/pricing.
How can I hear TheKeyBot before starting a trial?
You can hear TheKeyBot before starting a trial by using the instant demo at thekeybot.com/try, which calls your phone in about thirty seconds answering as a locksmith company. You can also call the live demo line at +1 (716) 350-6391, which answers as a demo locksmith shop, and try a few of the test calls yourself.
About the Author
TheKeyBot Team is dedicated to helping locksmiths grow their businesses through AI automation and smart technology. With years of experience in the locksmith industry, our team provides actionable insights and proven strategies.
