Let AI Count the Mozzarella, Not Greet the Customer

Grant Calloway

Andy's Pizza runs 14 shops across Washington D.C., Maryland, and Virginia. Its owner, Andy Brown, has a rule about artificial intelligence, and he states it plainly. Quote: "We will never replace a cashier with a kiosk. You will never call and get AI phone ordering at Andy's Pizza." End quote.

Grant Calloway

Now, you might expect that to be the end of the story. Pizza man refuses robots, everybody goes home. But here's the twist. Andy's Pizza uses AI extensively. Customers just never see it.

Grant Calloway

According to reporting from the Harvard Business School Institute for Business in Global Society, Brown used AI coding tools to build internal applications. One updates the kitchen screens. One tracks phone orders. One helps optimize staff schedules. He built them with Anthropic's Claude and YouTube tutorials, not with an enterprise software contract and a team of consultants.

Grant Calloway

The one I like best is the inventory app. A restaurant lives or dies on knowing what's in the walk in freezer. So Brown built a voice operated tool, and it works in English and in Spanish. An employee walks around the storage room, talks into a phone, and says how much of each ingredient is on the shelf. Brown's description: "almost like having a sous chef follow you around." The creator of this episode's research notes that this takes the place of clipboard counts, which is a chore nobody ever wrote a poem about.

Grant Calloway

So there's the strategy in one sentence. Put AI behind the counter, and keep human beings in front of it.

Grant Calloway

Let me put that against the bigger tape. The U.S. Chamber of Commerce reports that 58% of small businesses were using generative AI in 2025. That's up from 40% in 2024 and 23% in 2023. A separate survey from the Small Business Expo, 521 respondents, found 78.9% of owners say AI is more useful than it was a year ago. The top areas they named were research at 28.8% and advertising at 26.9%.

Grant Calloway

Plenty of people are trying it. Trying it and running a business on it are different animals. McKinsey reports that 88% of organizations use AI in at least one function. Only 1% describe their rollout as mature. Only 6% qualify as high performers seeing meaningful financial returns. That's a lot of folks in the lobby and very few at the table.

Grant Calloway

Why? Now I'm interpreting, so take it as my read and not a finding. I suspect many owners buy a pile of disconnected tools, or they hand the customer conversation to a bot and then wonder why buyers drift away. Andy Brown did neither. He picked a problem he could see with his own eyes.

Grant Calloway

Which brings us to a tool for deciding where AI belongs. Harvard Business School professor Andy Wu and NYU professor Bharat Anand developed a simple matrix, described in their Harvard Business Review work on a generative AI playbook. It rests on two questions.

Grant Calloway

First: how costly would a mistake be? Think money, safety, legal trouble, a bruised reputation. Second: what kind of knowledge does the task need? Is it explicit, meaning written down, structured, sitting in a database? Or is it tacit, meaning judgment, nuance, reading a person's mood?

Grant Calloway

Cross those two questions and you get four zones. Low stakes with explicit knowledge, that's where AI can largely run on its own. Think voice to text inventory logs, or a draft schedule you'll glance at. High stakes with explicit knowledge, AI can gather and draft, but a human checks and signs. Cash flow projections live here. Tammy Halevy of Reimagine Main Street notes that owners hesitate to hand finances to AI over accuracy and privacy concerns, and honestly, that hesitation looks sensible to me.

Grant Calloway

Low stakes with tacit knowledge, AI becomes a sandbox. Ten versions of a promo for the neighborhood street fair, you pick one and rewrite it in your own voice. And high stakes with tacit knowledge, keep it human. Greeting a regular. Calming an upset customer. Deciding who fits your crew.

Grant Calloway

Professor Wu explains why a small owner has an edge here. Quote: "The CEO in the smaller organization would have a more direct outlook of all the tasks that are taking up time in the organization." End quote. He adds that such a leader is equipped with more of the information needed to see where the AI opportunities are. You don't need a committee to find the bottleneck. You're standing in it.

Grant Calloway

Here's another example, from Houston. Becker's Books is a family owned used and rare bookstore with about 500,000 books. Books printed before ISBNs became widely used in the 1970s are hard to catalog. Manager Charlie Becker, backed by a $100,000 fellowship from O'Shaughnessy Ventures, is working with a software developer on an AI tool that scans a book's title and copyright pages, captures the details, and finds matching listings online. It's still a prototype, so I won't promise you a miracle. Early experiments suggest it can help with pricing and filling out online listings. Notice what it doesn't touch: the browsing, the conversation at the shelf. Same pattern as the pizza shop.

Grant Calloway

So here's your homework, and it takes one sitting. Take a sheet of paper and list everything your business does in a typical week. Then split the list in two columns. Front of house: anything touching customer trust, rapport, or hospitality. Back of house: anything repetitive, like tracking, logging, drafting, counting.

Grant Calloway

Leave the front column alone. Run the back column through the matrix. Pick one task that's low stakes and runs on explicit information. Try an AI tool on it for a few weeks, and keep a human watching the output. Start narrow. Then decide whether it earned a bigger job.

Grant Calloway

One last point. People are weary of phone trees and chatbots that can't answer a simple question. A real person picking up the phone is becoming a feature. You can advertise that. And AI behind the curtain is how you keep your costs down while you do it.

Grant Calloway

The market's loud with talk about replacing people. Andy Brown went the other way: let the machine count the mozzarella, and let a human say hello. That's a position I can trade on.