10 test calls to make before you buy an AI phone agent
A demo always sounds good. These ten calls show how an AI voice agent for property management handles the messy calls your properties actually get.
The Utter team · October 6, 2026 · 3 min read
Every AI voice agent for property managers sounds impressive in its own demo. The demo is a script. Your tenants and renters are not.
Before you sign up for anything, including Utter, call it yourself and try to make it fail. Here are ten calls that tell you more than any sales deck. Make them from your own phone, at night if you can, and take notes.
1. The real emergency
"There's water coming through my ceiling." Listen for three things. Does it treat this as urgent straight away? Does it get your address and unit first? And does it actually put you through to a person, or only promise that "someone will call you"?
An emergency is not the time for troubleshooting or a long list of questions. Afterwards, ask the vendor what happens when your on-call person doesn't pick up. "We'll keep trying" isn't an answer.
2. The fake emergency
"My dishwasher won't drain." This one should not wake anyone up. A good agent asks whether water is reaching the floor, may suggest a simple check, decides it's routine, and logs it for your team.
3. The mumbled address
Give your unit and street name quickly, the way people really talk. Then check what was written down. Street names are hard for any speech system. A good agent reads the address back to you to confirm.
4. The renter with conditions
"Is the two-bedroom still available? I have a dog and I'd like to move mid-month." It should answer from your real listing, apply your pet policy, and offer a tour time that actually exists. Then check your calendar to see that the tour is really there.
5. Spanish
Start in Spanish. Does it switch, and does it stay switched for the whole call? Then check that the details it took down, like your name and number, came out right.
6. The interruption
Talk over it halfway through a sentence. Does it stop and listen, or keep reading its script? Does it lose track of what it already asked?
7. "I want to talk to a person"
Ask for a human. During business hours, does it transfer you? After hours, does it tell you what happens instead, such as taking a message for a callback? It should never pretend nobody exists.
8. The caller who isn't a tenant
Call as an owner asking about a statement, or a vendor running late. An agent that only knows leasing and maintenance will get stuck. One that answers your main line should route you to the right person or take a clear message.
9. The question it can't answer
Ask something that isn't in your information: "Do you allow home businesses?" A good agent says it will find out, and makes sure someone follows up. A bad one makes something up. Made-up answers are the most dangerous failure, so test for them on purpose.
10. The hang-up and call back
Hang up mid-call, then call again. Is there a record of the first call that your team can see? Does the second call go smoothly, or does the agent get confused?
What to check afterwards
The call is only half of it. After each test, look at what landed on your side:
- Was the right person reached, and how fast, for the emergency?
- Is there a record with the details you'd need to act, not just a recording?
- Did the renter get a confirmation text with the right time and address?
- What did each call cost? Note how long each test call ran and multiply by the per-minute rate. What an AI receptionist really costs a property manager walks through the math.
If a vendor won't let you make these calls before paying, that tells you something too. Our demo line is open any time: (984) 363-1437 in the US, or (604) 706-8349 in Canada. Try to break it.