Inside a restaurant AI hackathon
Last month in Boston, a group of restaurant leaders, technologists, and enthusiasts got together to figure out what’s possible.
The idea for a restaurant AI hackathon came to restaurateur
after a particularly bad Saturday brunch service at a new location of his Boston restaurant, Shy Bird. Four months in, service was going great, but brunch remained tricky — especially on the last Saturday of June.Feldman explained it like this:
“By 9:45AM… it was clear we were going down. By noon, our average ticket time had gotten to over 30 minutes. Our longest single ticket clocked in at 45 minutes. Guests were understandably frustrated. Comps piled up. Our team, both FOH and BOH, were demoralized.”
Feldman is a tech-savvy operator, no stranger to new and trendy tech tools. (He was wearing a pair of Meta Ray-Bans during our interview last week.) This includes the resoundingly popular ChatGPT, an AI-powered chatbot that’s helped make generative AI the fastest-adopted tech tool in history.
Artificial intelligence has much potential to help restaurants.
The tech industry has promised, for well over a decade, to deepen guest relationships, automate repetitive tasks, provide valuable, applicable data and insights, and more. Now, as AI proliferates across the industry, those transformational opportunities finally feel real.
What’s less clear is how restaurants want to — and can — leverage those opportunities. Huge corporate chains are testing high-profile AI integrations to help take orders, predict periods of particularly high (or low) demand, make suggestions, handle customer service, and cut kitchen waste. But costs can add up, and, as we’re learning, these tools aren’t a sure solution.
In search of answers about that bad brunch, Feldman tried something novel: he fed data about the day’s service, plus photos and videos of the kitchen and an annotated menu, into ChatGPT. Then he prompted it in refreshingly natural language: Here's our data and some images from today's brunch service. WTF went wrong?
Less than five minutes later, he understood exactly WTF went wrong. Ingredients were set up inefficiently for the day’s most popular dish, ChatGPT told him, which caused a backup at the grill station. There, the layout prevented extra help from stepping in during crunch time. It was shockingly clear; ChatGPT provided a “profound and immediate” solution to a cascading problem, he said.
He soon shared details in his industry newsletter, Elided Opinions, and other operators enthusiastically responded. They seemed excited and inspired by ChatGPT’s practical, personalized, and shockingly accessible triage, and Feldman saw an opportunity to go deeper with an in-person event.
“I had wanted to do this for a while, but that [response] was kind of a catalyst for really pushing forward,” Feldman told me. “I've been getting a lot of outreach from restaurant tech people, but not from colleagues in the industry.”
On a Tuesday in August, Feldman partnered with Agent.ai to host a group of about 50 restaurant operators, employees, and industry technologists to explore what’s possible for independent restaurants. The hackathon started with education and ended with a few viable prototypes. Here’s how it went down.
Our conversation has been lightly edited for length and clarity. (Feldman also summed it up here.)
Expedite: Sounds like you got a great response after publicly detailing that particularly bad brunch.
Eli Feldman: “That was the first time that I think people saw a truly practical use for AI. Like, ‘I've had a terrible brunch, too. I've stood in the kitchen or stood at the pass and watched my KDS screen light up.’ A lot of my colleagues reached out and said, ‘Hey, I want to talk about this.’”