The AI Kitchen

What might artificial intelligence bring to independent, creative restaurants? And an interview with an Academy alum doing the hard work of bringing authentic Mexican cuisine to Greece

Illustration by Sofie Kampmark.

LONG READ

Two Basques with confusingly similar names and an interest in the future test the creative possibilities.

Imagine this: you’re a chef, preparing a new tasting menu. You’d like it to be 7 courses long and showcase some of the new techniques you’ve been developing. Obviously, you want to work with seasonal ingredients, and so much the better if they’re locally grown. You’ve got a team of 5 in the kitchen, and you’ve just committed to a four day week for all of them. You know from experience that you can’t price your menu at more than 150 euros a head, and you’d be really happy if you made 8% profit on it.

So far, so standard.

Now, imagine you could punch all those criteria into an app, and get back not only recipes for a seven-course menu that takes into account your personal style, your geographic location, and what’s growing in your producers’ fields, but also an automated order form linked to your suppliers’ inventory, a P&L sheet with your projected revenues and expenses, and a work schedule for each member of your crew with the amount of time they’ll need to spend on each task precisely outlined.

Would you do it?

It’s worth considering the question because that, with light embellishment, is more or less what two Basques with confusingly similar names presented in January at Madrid Fusion. Eneko Atxa, the chef of the three-starred Azurmendi located outside Bilbao, and Eneko Axpe, a physicist who formerly worked at NASA, have come up with an artificial intelligence platform that would allow a chef to generate all of the above with just a few taps of a keyboard. And the experiment of creating it has given them insight into precisely the kinds of questions that lie ahead for the industry.

Already, AI is being used in restaurants to keep track of inventory, take reservations, manage tables, reduce food waste, and analyze sales. A company called Lineup creates software that forecasts customer demand for restaurants and automates their staff scheduling accordingly. Another, called Bluecart, manages inventory and tracks spoilage. Cloudchef addresses the labor problem by using AI and a bunch of sensors to transform every slice, flip, and temperature change into a set of instructions that some random dude who stumbles in off the street with no cooking experience whatsoever could execute. Yanu has built a robotic bar that uses AI not only to take the order, choose the appropriate glassware and garnish, mix the drinks, and handle payment, but also has the potential to check customers’ age and, for those particularly messy ones, even breathalyze them.

What makes Atxa and Axpe’s experiments so intriguing is that, unlike most of the other existing AI platforms for the industry, their focus is not exclusively or even primarily on profitability or streamlining operations, or the other business-related aspects of running a restaurant that draw so many AI startups. Instead, the Enekos are looking to AI and asking how and whether it can help with the creative aspects of what it means to be a chef.

Axpe (the physicist) had never worked with food before meeting the other Eneko at a sustainability gathering in San Francisco five or six years ago. And although Atxa (the chef) considers himself a thoroughly analogue type of person (”I don’t even have WhatsApp,” he says with a laugh), he is also someone who has long brought a curious mind and serious research to his cooking, and he was intrigued by the questions that AI raised for both cuisine, and for society at large.

Together, they initially set out to create the first fine-dining menu generated entirely by AI. The premise was simple: they prompted ChatGPT to design a menu that could be served in a Michelin-starred restaurant–one not unlike Azurmendi. There were three stages to the test: first, the AI had to come up with appropriately named dishes. Then, it had to generate an image of each. And finally, it needed to write a recipe for the dish with all ingredients, cooking times, and instructions included. Following it to the letter, Eneko then executed the dishes.

They presented the results of that experiment at a previous Madrid Fusion, in 2024. A cauliflower mousse with balsamic ‘caviar,’ was typical of what the machine created. Its description definitely sounded right, and between the airy dollop of brassica foam, the tiny pearls of vinegar, and the garnish of microgreens, the dish certainly looked like it belonged on a tasting menu (albeit, perhaps, one that came out in 2012). Eneko Axpe thought the dish was tasty and considered it a success. But Atxa begged to differ. “It was missing the spark,” he says. “What he told me is that it might fit in a 3-star restaurant, but not in Eneko Atxa’s restaurant,” Axpe recalls.

Other outcomes were stranger. The AI hallucinated at times, inventing colors or ingredients that didn’t actually exist. It came up with weirdly unpleasant flavor combinations. At one point, it created a “coconut sphere” whose texture was both spongy and sticky at the same time. “Looking at the image you would imagine it was going to be marvelously airy and subtle,” Atxa says. “But it was completely the opposite.”

The experience left the two Eneko’s with different conclusions. For the chef, it was clear that machines weren’t going to replace him anytime soon. “Gastronomic restaurants are very personal,” he says. “They depend on a particular, peculiar vision, the vision of their author. AI might be helpful, but the platform would rob the soul from that kind of restaurant.”

But if Eneko the chef wasn’t going to be putting coconut balls on his menu anytime soon, Eneko the physicist saw the creation as proof that AI wasn’t limited by many of the things–like fear of spoogy textures–that can hold humans back. “AI isn’t afraid to fail, it doesn’t care what people say about it,” he says. “And by bringing an abundance of information, it can endlessly innovate.”

The experiment left both of them even more intrigued, however, and opened up other avenues to pursue. They could take one of the restaurant’s pre-existing recipes, for example, and ask the AI to lower its carbon footprint, or substitute ingredients that would lower its sugar content or make it suitable for celiacs. They could upload the entire catalogue of Azurmendi’s recipes and ask it to extrapolate new ones. In the months that followed, Axpe even oversaw an Iron Chef-esque contest that pitted AI, generating recipes that a pair of young cooks would execute, against Ricard Camerena, one of Spain’s top chefs, who created his own dishes. (Camerena won on two of the three courses, but the third–a carrot dessert–was a draw).

That work has led them both to see AI as a useful tool in all kinds of kitchens, including fine dining ones. “I see it now like a pantry—a pantry of knowledge—where you can choose the ingredients you want to use,” Atxa says.

When the organizers of this year’s Madrid Fusion asked them to apply AI to a dying Spanish institution, the inexpensive three-course menu del dia that nearly every restaurant in the country once served at lunchtime, they leapt at the chance. This time, the goal was not just to come up with recipes for the dishes, but to incorporate a whole host of useful factors that might allow anyone who used it to keep the beloved menu del dia viable financially and creatively viable.

Axpe started by asking his chef counterpart about all the criteria that went into a decision about what to put on the menu. “I said to him, ‘Imagine I’m God, and I can design any kind of tool you like for this establishment,” Axpe says. “What would you ask for?”

They ended up with an interface that allows the user to input the number of covers, desired price and profit margin, number of cooks and the amount of time they have, even whether the style of cooking is traditional or creative. Within seconds, the AI produces a three-course menu minutely designed to meet every criterion, with recipes, a complete ingredient list (which could be linked to the purveyor’s website for automatic ordering), a schedule of tasks for each cook, and attractive photos of each finished dish. “Honestly, it was easier than I thought it would be,” Axpe says.

Admittedly, the platform was designed for the bistro or other kind of casual restaurant that serves a menu del dia, and the experiment has spurred them to think more about the needs of home cooks. But there is no reason, Axpe says, that the same technology couldn’t be applied to an elaborate tasting menu and all the decisions–from style to cost to available hands–that go into creating it. And even Atxa, who as a chef still believes that AI can’t achieve the intangible artistic spark–some call it ‘soul’-- necessary for a gastronomic restaurant, can see it coming closer as the machines learn, and humans get better at teaching them.

That doesn’t mean they don’t see risks that artificial intelligence brings. “AI is like a knife,” Axpe says. “Doctors in a hospital save lives with knives, and cooks make dishes with them. But you can also kill people with one. It’s not the tool, it’s how it’s used.”

Both acknowledge that AI will almost certainly make some jobs obsolete, and could well put whole professions out of work. It may also change how everyone–including high-end chefs–see the creative work that they do. “I consider myself absolutely an artisan,” Atxa says. “And I can see how using AI is less artisanal. But think about the people who make ceramics. If they use an electric wheel to throw their pots, do we say they’re not artisans anymore? If farmers use AI to know when to harvest, are they not growers anymore?”

Eneko the physicist, who has now written a book on AI in the kitchen, adds that the technology may even open up new distinctions. “There’s a future in which every single restaurant will somehow be using AI,” Axpe says. “And then there’s one in which some people rebel against that trend and they’ll cook only with human creativity, human-made recipes. Those restaurants will call themselves 100% human made, and that will be wonderful as well.”

Until then, both of them keep experimenting with how AI can be used to support what chefs—and others–do in the kitchen. “When you use human intelligence and artificial intelligence in tandem, it’s really a boost to human creativity,” Axpe says. “ To me, that's the takeaway.”

ACADEMY ALUMNI

An interview with Dimitris Afentakis

Chef, restauranter, and MAD alum Dimitris Afentakis has dedicated himself to bringing authentic Mexican cuisine to Greece. But it hasn’t always been easy, thanks to a lack of visibility, rising debts, and many that only want the kind of ‘Mexican’ smothered in melted cheese. But while Athens has taken time to warm up to restaurant Atole, Dimitris ambition hasn’t gone unnoticed. This month, Dimitris shares how his yearly pilgrimage to Copenhagen keeps him inspired, including by learning from MAD hall-of-famer Rosio Sanchez, and an exciting offer on the horizon.

The first time you became acquainted with MAD, you attended Symposium in 2018— how did attending change your perspective?
It was nothing short of a transformative experience for me. I had already lived in several countries—London, Spain, and New York—but visiting Copenhagen opened my mind in a completely new way. The level of collaboration and the way the restaurant industry functions, to me, is unlike anything I've seen before. At the Symposium, I saw a different approach—chefs supporting each other, exchanging ideas, and truly working as a community. It was the first time I felt that kind of openness, and it reshaped my outlook on the industry. I realized that food is not just about business; it’s about creating something meaningful, something that can inspire and educate others. Now, I’m in Copenhagen at least once a year to get that boost!