Somebody Sold My Pizza Place an AI Phone System

My favorite pizza place replaced the person answering the phone with an AI. My orders used to take about a minute: say what I wanted, get the total, find out when to pick it up. I’m not going to name the place, because the pizza is still good and I still eat there. But I timed a call with the new system.

Forty seconds of greeting before I could say anything. I asked what pizzas they had, and it asked me to repeat the question. Twice. Ninety seconds in, I finally heard that they make four styles. Then came the order, read back to me with a pause and a “just to confirm” in front of every line, and read back wrong. The system kept getting it wrong, so I said the same order more than once.

The call took three minutes. I remember previous orders taking about one, though I don’t have a timed average for those calls.

I don’t know what the shop pays for it, how many calls it used to miss, or whether anyone there has compared the two systems. It might be saving enough staff time to justify a slower order. As a customer, though, I got a longer greeting and more mistakes.

What happens when the staff can’t pick up?

If the shop is missing orders during the dinner rush, I can see a reason to buy an automated phone system. A customer who gets through may put up with a slower call. That is a different comparison from putting the system in front of someone who could have reached an employee immediately.

A service business has the same decision to make. An HVAC company might want help catching no-heat calls after hours. During the day, its dispatcher may already know the customer and need only a gate code or a model number. I’d test both kinds of call before deciding which ones to automate.

Street names, accents, background noise, and corrections to an order all belong in that test. A quiet sales demo doesn’t tell you how the system handles a caller standing next to a running truck. It also doesn’t tell you how often regular customers will tolerate the same long greeting.

How I’d try it at the shop

I’d start with the current phone service. Sample calls during busy and quiet hours, and note how long an order takes, whether it’s correct, and how much employee time it needs. Check the phone logs for missed and abandoned calls too. Call length alone won’t tell you whether the business is losing orders.

Then I’d have staff place test orders through the vendor’s system, using a test setup that won’t send food to the kitchen or charge anyone. Include a changed topping, an address the system has to ask about, and a caller correcting something it got wrong.

Before a customer pilot, agree on what would justify keeping it. That might be more completed orders during the rush, less staff time on the phone, or fewer missed calls. Set a limit on the mistakes and abandoned calls you’re willing to accept in exchange. Have a way for a caller to reach a person when the order gets stuck, and test that as well.

I’d first try it as overflow when staff can’t answer. If that works, the shop can decide whether to give it more calls. And I’d shorten the greeting before either test. Forty seconds was a lot to sit through before I could ask about a pizza.

If you’re considering a phone system like this, send me the quote and describe the calls you need it to handle. I can help work out what to test and whether the results support buying it.