Latest / Restaurant Business with Fexingo: Food Service, Hospitality, and Independent Operators / How Restaurants Are Using AI to Predict Inventory
Transcript
- Lucas: Luna, I want to talk about something that's quietly reshaping how restaurants order food — and it's not just for big chains anymore. Luna: I'm guessing this is the AI inventory prediction thing that's been popping up in trade press lately? Lucas: Exactly. And I've got a concrete case. There's a 12-unit restaurant group in Chicago — casual midscale, think gastropubs and taquerias — that started using a predictive inventory system from a company called Winnow about six months ago. Lucas: They cut food waste by 22 percent. That's not a rounding error. That's real money when your food cost runs 30 to 35 percent of revenue. Luna: Twenty-two percent in six months? That's impressive. How does it actually work — is it like a smarter spreadsheet, or is it genuinely machine learning? Lucas: It's genuine ML. The software ingests historical point of sale data, but it also layers in external signals — weather forecasts, local event calendars, even day-of-week and holiday patterns. It then predicts, for each ingredient, how much you'll need for the next week down to the pound. Luna: So it's not just a fancy order guide — it's actually learning from what happened last time it rained on a Tuesday during March Madness. Lucas: Exactly. The Chicago group told me that before Winnow, their head chef was essentially guessing based on gut feel and last year's numbers. That works okay for a single location, but across 12 kitchens with different traffic patterns, the variance eats into margin. Lucas: The system also flags anomalies — like if a supplier sent you 50 pounds of avocados when you ordered 40, it catches it. That kind of leak adds up fast. Luna: I can see how that helps waste, but what about the cost of the system itself? For a small independent with one or two locations, does the math pencil out? Lucas: That's the key question. Winnow and competitors like LeanPath and BlueCart charge on a subscription basis — typically a few hundred to a couple thousand dollars a month depending on volume. For the Chicago group, the 22 percent waste reduction translated to about 18,000 dollars saved per location annually, after the subscription cost. Lucas: But for a single restaurant doing maybe 1.2 million in revenue with a 32 percent food cost, the savings might be closer to 6 or 7 thousand a year — which is still a positive ROI, but it takes longer to feel meaningful. Luna: So the ROI is real, but the upfront friction isn't just money — it's also the time to set it up and train staff. What's the integration like with existing POS systems? Lucas: Most of these platforms have APIs that connect to the major POS systems — Toast, Square, Aloha, Micros. The Chicago group used Toast, and they said the integration took about two weeks to get the historical data flowing and another two weeks to calibrate the model. Lucas: The bigger hurdle, honestly, is getting the kitchen staff to trust the system over their own instincts. One chef told me he ignored the system's recommendation to order less romaine because he 'had a feeling' — and ended up throwing away 40 percent of the delivery. Luna: That's human nature, right? It's hard to hand over that decision to an algorithm when you've been doing it by feel for 15 years. Lucas: Right. And this is a theme we see in a lot of industries — AI is only as good as the adoption. But the data is pretty clear that for high-volume, high-turnover ingredients, the machine wins on consistency. Luna: You know, this reminds me of something I've been thinking about lately. We do these episodes because we genuinely believe these conversations help operators and entrepreneurs make better decisions. And I think that's worth saying out loud. Lucas: Absolutely. Look, we deliberately don't run ads on these shows. No sponsors, no promos. If you've found value in episodes like this one — maybe it's saved you a few hundred bucks on food cost or made you think differently about a supplier relationship — the only way we keep this ad-free is through listener support. Luna: Yeah, it's a simple model. If these conversations have moved your work forward in some small way, you can support us at buy me a coffee dot com slash fexingo. That's the whole ask — no tiers, no perks, just keeping the signal clean. Lucas: Exactly. And now back to those avocados. Because the other thing the Chicago group learned was that their avocado waste wasn't just about over-ordering — it was about ripeness timing. Luna: Tell me more — is the AI predicting ripeness windows too? Lucas: Not directly, but it can model your usage rate. If you typically use 30 pounds of avocados on a Saturday but you're ordering them Thursday, the system can recommend adjusting the order day or the quantity to align with when they'll actually be used. It's that kind of operational nuance that human planners often miss. Lucas: And this is where I think the technology really shines — not in replacing the chef, but in providing a second set of eyes on the data. The chef still decides the menu, the quality specs, the supplier relationships. The AI just says, 'hey, based on the last eight weeks, you're going to need 15 percent less spinach this week.' Luna: So it's more of a decision support tool than a replacement. That seems like a smart framing. Are there any downsides or risks we should flag? Lucas: The biggest risk is over-reliance. If the system gets stale because you change the menu significantly — say you swap out a burger for a salmon bowl — the historical data won't reflect that. You need to retrain the model, which takes a few weeks of new data. Lucas: Also, these systems work best with standardized recipes. If your kitchen is loose with portioning — one cook puts 6 ounces of protein, another puts 7 — the predictions get noisy. Luna: That's a good point. The tech forces you to tighten up your processes, which is probably a good thing anyway. Lucas: Exactly. And I think this is where the industry is headed. We're seeing more and more mid-market groups adopt this — not just the big chains like Darden or Chipotle. The cost of the sensors and software is dropping, and the cloud infrastructure makes it accessible. Luna: What about the next frontier? Are we going to see AI that automatically places orders with suppliers based on predictions? Lucas: Some systems already do that — automated ordering through integrations with distributors like Sysco or US Foods. But most operators I've talked to want a human in the loop to review the order before it goes through. Trust, again. Lucas: But give it another year or two, and I think full automation will be common for high-volume, predictable items. The chef will still order the specialty ingredients and the seasonal stuff. Luna: So we're not talking about a robot chef any time soon, but the back office is getting a serious upgrade. Lucas: Exactly. And that 22 percent waste reduction? That's not just good for the bottom line — it's also good for the planet. Less food in landfills, less water used to grow food that gets thrown away. Lucas: It's one of those rare wins where operational efficiency and sustainability actually align. Luna: I love that. So for our listeners who are operators — what's the one takeaway? Should they be looking into this now, or wait another year? Lucas: I'd say if you have three or more locations, the math starts to get interesting today. For single units, the subscription cost might be a barrier, but there are cheaper alternatives — even using a spreadsheet with some basic forecasting formulas can get you part of the way. Lucas: The key is to start tracking your waste systematically. If you don't know how much you're throwing away, you can't fix it. Luna: Right. Measure first, then optimize. That's solid advice. Lucas, great conversation as always. Lucas: Thanks, Luna. Next time we'll look at how restaurant groups are using AI for menu engineering — figuring out which dishes to push based on margin and popularity. That's a whole other layer.